Conference program

Abstract: We systematically scrutinise all the facets of privacy in the 5G delegated-authentication procedure called AKMA (Authentication and Key Management for Applications based on 3GPP credentials in the 5G Systems). We define, in general terms, a privacy-threat model and privacy requirements for this protocol. Using these definitions, we find numerous privacy failings in the AKMA protocol. We propose a patch, called AKMA-p, which imposes minimal changes on AKMA, yet it attains all our privacy requirements. We also formalise and analyse all of this in terms of formal privacy-verification in the Dolev-Yao model; to this end, we use the Tamarin prover to systematically carried out our formal analyses of AKMA and AKMA-p.

Meta Review: The paper "A Systematic Study of Practical & Formal Privacy in the 5G AKMA Procedure" presents a formal analysis of the 5G AKMA protocol using the Tamarin prover. Authors discuss different weaknesses and flaws of the protocol with a specific focus on privacy properties. Authors propose new fixes which unlike the existing related work, aims at minimizing the changes in the current standard to keep backward compatibility and ease acceptability. Even if verifying privacy properties remains a challenge in symbolic models, authors prove the correctness of their fixes using the Tamarin prover. In summary, the paper makes a significant contribution to the standardization of an important 5G protocol, and deserves to be included in the EuroS&P'25 program.

Abstract: Federated Learning (FL), a distributed learning mechanism where data is decentralized across multiple devices and periodic gradient updates are shared, is an alternative to centralized training that aims to address privacy issues arising from raw data sharing. In spite of the expected privacy benefits, prior research showcases the potential privacy leakage that advents from overfitting, exploited by passive attacks for membership and attribute inference purposes. However, limited attention has been given into understanding and defending against active threats, that is attacks that increase model leakage by interfering on the training process. This work addresses this gap by introducing Active Attribute Inference (AAI), a novel active attack that encodes sensitive attribute information by making any targeted training sample leave a distinguishable footprint on the gradient of maliciously modified neurons. Results, using two real world datasets, show that it is possible to successfully encode sensitive information incurring in a small error in terms of neuron activation. Inference on a practical scenario reveals that AAI can improve upon a state-of-the-art approach by achieving over 90% of restricted ROC AUC. Moreover, as a defense against such active attacks, this work introduces several attack detection strategies tailored for different levels of the defender's knowledge. Including the novel White-box Attack Detection Mechanism (WADM) that detects abnormal changes in weights distribution, and two black-box strategies based on model performance monitorization. Results show that the detection rate can be 100% on both datasets. Remarkably, WADM reduces any attack to random guessing while preserving model utility, offering significant improvements over existing defenses. Hence, this research contributes for an enhancement of privacy of FL systems, ensuring an appropriate privacy-model utility tradeoff.

Meta Review: This paper presents a novel inference attack in the context of Federated Learning. The proposed attack reveals a new threat without relying on overfitting, i.e., inferring attribute information by actively modifying the model gradients. The evaluation of the proposed attacks is comprehensive over large datasets. The results demonstrate the effectiveness and generality of the attacks. Besides, the paper also evaluates black-box and white-box defence mechanisms to show how the attacks can be detected and mitigated. In summary, the paper makes a new contribution to security and privacy of federated learning, and identifies a new attack surface that can compromises the privacy claims of FL systems. The findings will help the community build more secure and privacy-preserving FL systems.

Abstract: Anonymous token schemes are cryptographic protocols for limiting the access to online resources to credible users. The resource provider issues a set of access tokens to the credible user that they can later redeem anonymously, i.e., without the provider being able to link their redemptions. When combined with credibility tests such as CAPTCHAs, anonymous token schemes can significantly increase user experience and provider security, without exposing user access patterns to providers. Current anonymous token schemes such as the Privacy Pass protocol by Davidson et al. rely on oblivious pseudorandom functions (OPRFs), which let server and user jointly compute randomly looking access tokens, and hence their performance decreases with the number of issued tokens. In this work, we propose a new approach for building anonymous token schemes. Instead of relying on two-party computation of a pseudorandom function, we propose to offload token generation to the user by using group verifiable random functions (GVRFs). GVRFs are a new cryptographic primitive that allow users to produce verifiable pseudorandomness. Opposed to standard VRFs, verification is anonymous within the group of credible users. We give a construction of group VRFs for the Dodis- Yampolskiy PRF based on pairings and a new Diffie-Hellman inversion assumption that we analyze in the generic group model. Our construction does not require generic zero- knowledge proofs and enjoys compact public keys, compact proofs and fast verification. By replacing the OPRF in Privacy Pass with our group VRF, we obtain a highly scalable anonymous token scheme with constant communication during token issuance, and where the number of issued tokens can be decided adaptively even after the credibility check, e.g., based on the current threat level of the provider. Our tokens are further countable and publicly verifiable.

Meta Review: The paper introduces a new cryptographic primitive, a Group Verifiable Random Function (GVRF), and applies it to construct an anonymous token system that addresses certain shortcomings of Privacy Pass. While the reviewers expressed doubts about the direct applicability of the proposed system as a drop-in replacement for Privacy Pass, it was agreed that the primitive and its construction represent valuable scientific contributions. Notably, they enable a new trade-off: more efficient token issuance (with constant communication cost in the proposed system) at the expense of a more costly presentation. It is worth noting that the GVRF primitive relies on a new hardness assumption, but this assumption is appropriately formalized and analyzed in the generic group model. Finally, the work provides a proof-of-concept implementation of good quality.

Abstract: Automation apps enable seamless connection of IoT devices and services to provide useful functionality for end-users. These apps are typically executed on cloud-based Trigger-Action Platforms(TAPs) such as IFTTT and Node-RED, supporting both single- and multi-tenant architectures. These architectures raise security and privacy concerns in the face of cloud attackers and malicious app makers, resulting in massive exfiltration of sensitive user data. To address these concerns, we design TAPShield, an architecture that uses confidential computing and language-level sandboxing to protect users’ sensitive information against untrustworthy TAPs and malicious apps. TAPShield targets JavaScript-driven TAPs built on the Node.js environment and uses trusted execution environments via Intel SGX to protect against cloud attackers. It further uses language-level sandboxes such as vm2 and SandTrap to protect against malicious apps. We implement TAPShield for two popular TAPs, Node-RED and IFTTT, and report on the security, performance, and compatibility trade-offs on a range of real-world apps. Our results show clear security benefits with acceptable performance overhead while adhering to existing development practices of production-scale TAPs

Meta Review: This paper provides a system for protecting security and privacy of cloud-based IoT apps in the presence of attackers that may control the cloud environment or a malicious app in a multi-tenant architecture. It makes a significant contribution by providing guarantees under a strong adversary model. The TEE-backed approach and system design provides a clean solution for a host of problems associated with deploying TAP services on untrusted multi-tenant cloud services. The paper provides a comprehensive evaluation of the solution using real-world IoT applications on the IFTT and NodeRed platforms. To summarize, reviewers unanimously agree that the paper addresses a key problem in the security of trigger action programs under a strong threat model, and provides a convincing evaluation of the proposed system, thereby advancing the state of the art.

Abstract: In many places, a surge of micro-mobility sharing systems, as for instance e-scooters, can be observed. Shared micro-mobility is a cost-efficient and flexible alternative to owning vehicles and, furthermore, leads to reduced traffic and air pollution. However, sharing information about vehicles impacts the privacy of the individuals using such vehicles as changes in vehicle state are linked to an individual's mobility pattern. Malicious exploitation of knowledge on mobility patterns of individuals may assist in criminal activities such as stalking or burglary. Thus, it is very important that micro-mobility sharing platforms do not leak sensitive data about the mobility patterns of their users, resulting in a tradeoff between sharing and privacy. To characterize the privacy leakage in one specific instance of shared micro-mobility, we conducted a large-scale, long-term data collection from scooters run by an e-scooter company in a European university town. Indeed, the data reveals several privacy issues: For instance, we were able to reconstruct work and school schedules of various individuals. Furthermore, we could infer interests and hobbies by visits to, e.g., sports facilities. Our initial discovery of such leakages was aided by the fact that the specific e-scooter company does not comply with existing privacy standards, in particular the use of dynamic IDs. Yet, an a-posteriori analysis of our data shows that even with dynamic IDs, we are able to re-construct 80\% of the trips, which still constitutes a substantial privacy leakage.

Meta Review: The paper "They See Me Scooting - A Long-Term Real-World Data Analysis of Shared Micro-Mobility Services and their Privacy Leakage" investigates privacy risks in micro-mobility services, which have been gaining popularity over the last years. This work advances the state of knowledge by showing that a previously suggested and implemented countermeasure to privacy leaks - dynamic scooter IDs - still allow for determining users' movement patterns from information about scooter availability learned from publicly available APIs. These worrisome results are presented in a well-written and illustrated manner, providing concrete examples of movement paths that illustrate how micro-mobility users' personal information can be leaked. The paper provides suggestions for actionable defenses, and in their disclosures and reactions to reviewer feedback the authors show commitment to research ethics and reproducibility. Overall, the direct real-world implications of the findings that show a clear need for further privacy protections add to the state of knowledge in micro-mobility privacy and make this paper a great addition to Euro S&P 2025.

Abstract: Tag Management Systems were developed in order to support website publishers in installing multiple third-party JavaScript scripts (Tags) on their websites. Google developed its own TMS called “Google Tag Manager” (GTM) that is currently present on 42% of the top 1 million most popular websites. However, GTM has not yet been thoroughly evaluated by the academic research community. In this work, we study, for the first time, the Tags provided within the GTM system. We propose a new methodology called “detecting privacy leaks in isolation” and apply it to multiple Tags to analyse the types of data that Tags collect and contrast them to the legal and technical documentation, in collaboration with a legal expert. Across three studies – in-depth analysis of 6 Tags, automated analysis of 718 Tags, and analysis of Google “Consent Mode” – we discover multiple hidden data leaks, incomplete and diverging declarations, undisclosed third-parties and cookies, personal data sharing without consent and we further identify potential legal violations within EU Data Protection law.

Meta Review: This paper investigates the use of Google Tag Manager (GTM) across websites, offering both technical and legal perspectives on its data collection practices and transparency. The authors conducted an in-depth study of six popular tags and present a methodology for testing them in isolation. The reviewers recognize that the paper makes a significant contribution by shedding light on an underexplored aspect of web tracking, raising awareness about potential legal violations and value the authors' effort to address reviewer concerns (e.g., limited number of tags analyzed and claims of methodological novelty). The reviewers believe that the paper merits inclusion in the EuroS&P'25 program.

Abstract: During police investigations, the possibility of extracting coordinates data of a vehicle without relying on GPS is vital because the targets may know about possible bugs using this technology and take action against them. To this aim, extracting non-encrypted data from the Controller Area Network (CAN) provides a valid solution for officers. Indeed, CAN exchanges non-encrypted data, including physical information about the car's movement. In this paper, we present On Path Diagnostic - Intrusion & Inference (OPD-II), a novel path inference attack leveraging a physical car model and a map-matching algorithm to infer the path driven by a car based on CAN bus data. Unlike available attacks, our approach only requires the attacker to know the initial location and heading of the victim's car and is not limited by the availability of training data, road configurations, or the need to access other victim's devices (e.g., smartphones). We implement our attack on a set of four different cars and a total number of 41 tracks in different road and traffic scenarios. We achieve an average 95% accuracy in reconstructing the coordinates of the recorded path by leveraging a dynamic map-matching algorithm that outperforms other state-of-the-art proposals while removing their set of assumptions.

Meta Review: This paper suggests an important attack with direct real-world applicability. It proposes a novel path inference attack method called OPD-II (On Path Diagnostic-Intrusion & Inference), which realizes high-precision reconstruction of vehicle driving paths by utilizing vehicle CAN bus data and OBD-II protocol. This method is novel and innovative, bringing new research perspectives to the field of vehicle location privacy protection. This paper outperforms previous works. The authors make their tool/toolchain publicly available to facilitate reproducibility and open science. This paper demonstrates that the proposed algorithm works in various road types and environmental conditions. It allows corresponding parties (companies, developers) to have a time to consider the reported issues and possibility of fixing issues later. The authors did a responsible disclosure with the automotive companies involved in this paper's evaluations.

Abstract: Industrial control systems have enabled the digitalization and automation of industrial production and services, such as electric powerhouses, the electric grid, and water supply networks. Due to their critical role, any exposure to the public Internet makes them vulnerable to attacks that may have catastrophic implications. In this paper, we report that the readily available application-layer scanning on all ports opens new avenues to assess the exposure of devices that run industrial control protocols that were not possible with previously proposed active port scanning. We consider 17 widely used industrial control system protocols, and we develop a methodology that unveils that up to 140 thousand industrial control systems are exposed around the globe. Our study shows that many allegedly exposed industrial control systems are honeypots that emulate industrial protocols. We develop a methodology to infer the presence of honeypots and classify them into three tiers based on the confidence that these act as honeypots: low-, medium-, and high-confidence. We classify them thanks to large-scale application-layer scanning on all ports and multiple independent attributes, including network information, number of open ports, and known honeypot signatures. Our results show that around 15\% of the exposed industrial control systems are honeypots (with two-thirds of them belonging to the medium- or high-confidence categories). Our results challenge previous reports on the prevalence and distribution of exposed industrial control systems. The developed methodology enables industry operators to assess exposed assets and aids protection teams in creating stealthier honeypots.

Meta Review: The paper analyzes the problem of detecting ICS honeypots on the Internet. Detecting honeypots is difficult, mostly because ground truth does not exist. The authors propose an easy-to-apply approach for detecting and classifying honeypots into low, medium, and high confidence. As a plausibility check, the authors themselves deploy honeypots and wait until they are included in the Censys data set. The responsible disclosure efforts are particularly interesting as they describe interesting information "from the trenches" when doing this type of scanning. The "Oddballs" sections are also quite interesting as they explicitly describe future research directions. The ethical considerations and the statement about open science are convincing. Overall, the reviewers found that enhances our understanding of detecting ICS honeypots on the Internet.

Abstract: Scripting languages like Python or JavaScript are extremely popular among developers, in part due to their massive open-source ecosystems that enable smooth code reuse. However, recent work shows that a lot of scripting code runs C/C++ code under the hood, via native extensions. This might introduce subtle security issues that can surprise the users. Prior work in this domain relies on simple, intra-procedural, flow-insensitive data flow analysis to detect such problems, but it is unclear if a more holistic polyglot static analysis could be feasible, and if so, what are its costs and benefits. In this work, we propose CHARON, the first inter-procedural, polyglot static analysis for detecting vulnerabilities in scripting languages. Our approach advocates for linking together the code property graphs of the different languages and performing cross-language data flow analysis by switching between code representations, when cross-language function calls are encountered. In this way, CHARON supports data flows that cross several times the language boundary, spanning multiple functions on either side. We evaluated CHARON on 11.8K polyglot packages from npm and PyPI, containing 896M lines of code. CHARON identified 5,813 manually-confirmed, vulnerable data flows in 116 packages. We performed a baseline comparison of CHARON with single-language analysis on native code, showing a ~6x increase in true positives and ~4% less false positive alerts. We demonstrated exploitability of the discovered vulnerabilities by creating 63 PoCs across 34 packages, showing, among others, how we can escalate a buffer overflow vulnerability in native extensions to arbitrary code execution, which we believe to be the first of its kind. Overall, our results show that inter-procedural, polyglot analysis is both feasible and effective for detection of native extension vulnerabilities.

Meta Review: This work presents the concept of Polyglot Property Graphs (PPGs), which are an extension of Code Property Graphs. PPG's leverage cross-language edges that allow us to perform end-to-end static analysis in the setting of, e.g., script code calling into native code modules. The reviewers believe PPG's to be a valuable building block that has clear utility in advancing the field of static analysis of data flows across language boundaries. The reviewers also believe that the utility of such an analysis was well analyzed on a significant corpus of Python and NodeJS packages, leading to 63 PoC exploits.

Abstract: The use of security standards is common in the corporate world to enhance organizational security, but the level of their effectiveness remains uncertain. Budgetary restrictions, skills shortages, existing infrastructure, difficulties in adoption, and usability challenges may hinder the implementation of complete security measures. As a result, organizations may opt to adopt only part of the available security procedures. Among the different security techniques and validations available, many companies choose to follow security standards such as NIST Cybersecurity Framework or ISO 27001:2013, which consist of many individual guidelines. However, some of these guidelines may be outdated, unclear, or incapable of addressing new technologies, resulting in insufficient practical and actionable security improvements. We conducted two studies to understand the perceptions and pain points of practitioners. In the first study, 30 security experts evaluated the perceived effectiveness of 113 guidelines from ISO 27001 and 108 guidelines from the NIST Cybersecurity Framework, rating them as High, Medium, or Low. They also provided their perspectives and insights behind the total 1326 ratings and what areas of improvement existed in the standards. Building on the initial study's recommendations, we engaged 14 additional security experts to explore the pain points and potential changes in security systems through in-depth interviews. Our findings indicate that in order to enhance the perception and effectiveness of the standards, timely updates, flexibility, maintainability, cost-effectiveness, and the ability to impact technical security are needed.

Meta Review: The paper examines the effectiveness of ISO 27001 and NIST CSF standards from the practitioners' perspective, focusing on their strengths, weaknesses, and alignment with real-world security requirements. It employs a structured, multi-phase research approach, incorporating quantitative surveys where experts initially rate the perceived efficacy of the two frameworks and qualitative interviews with experts to further identify challenges associated with these frameworks. The insights from people directly involved in the design of these standards offer a unique perspective on the issue. The paper highlights significant gaps in current research, specifically addressing the disconnect between standard design practices and practitioners. The survey instrument and the interview guideline are made open-source thereby promoting open science. Overall, this paper has important implications for practitioners, given the recent updates to ISO 27001 and NIST CSF, assessing how these revisions tackle current challenges.

Abstract: The modern web ecosystem relies heavily on the inclusion of third-party scripts as they offer useful, and often necessary, functionality. This inclusion leads to the ``blending'' of code from different origins, which has significant ramifications. Specifically, the inability to effectively and robustly disambiguate between first-party and embedded third-party scripts can severely undermine the security and privacy guarantees of existing defenses (e.g., blocking trackers or preventing vulnerabilities such as DOM XSS), as well as the validity of web measurement studies. To address that gap we propose StyxJS, a system that is able to provide real-time attribution of third-party scripts while preventing evasive tactics that can be employed by malicious scripts. This is achieved through an automated pipeline consisting of stack walking, script rewriting, browser API overriding, and tamper-proofing mechanisms. Crucially, our system does not require any developer input or prior knowledge about the website and can, thus, be readily incorporated into any countermeasure or web measurement apparatus that requires robust script attribution. We conduct an extensive experimental evaluation of our system and demonstrate that it accurately captures more script inclusion techniques compared to prior work, while incurring a negligible performance overhead, and effectively maintains page-deployed security mechanisms (e.g., CSP). We also detail the straightforward process and benefits of retrofitting a varied set of existing defenses on top of StyxJS, as well as leveraging it to analyze the web ecosystem. We will release our system as an open source project, to allow security researchers and practitioners to benefit from StyxJS' capabilities.

Meta Review: The authors proposed StyxJS, a new framework designed to attribute the sources of JS scripts in execution. The reviewers appreciate the authors' contribution to summarizing a comprehensive list of factors that disrupt source attribution while considering potential adversarial attempts to bypass StyxJS. Moreover, the authors' commitment to releasing StyxJS as an open-source project is highly appreciated, as it provides a foundational framework upon which follow-up research can be built. The paper and its artifact make solid contributions to our community by providing up-to-date insights into the comprehensive factors that complicate the attribution of JS script sources. Overall, this paper contributes to improving our understanding of attributing JS scripts in execution along with the open-source artifact.

Abstract: Small and medium enterprises (SMEs) are increasingly vulnerable to cyber threats due to limited resources and cybersecurity expertise, in addition to an increasingly hostile cyber threat environment at national and international levels. This study aims to improve the cyber resilience amongst SMEs by developing a risk assessment tool. A comprehensive review of existing risk assessment tools and international advice and guidance was carried out. Through iterative collaboration with experts and SMEs, a user-friendly risk assessment tool was developed. This tool simplifies risk for non-expert users and also raises an SME’s awareness of their cyber risk just through answering the questions in the tool. The tool has led to insights into the current state of participating SME’s cybersecurity.

Meta Review: The paper reports on a multistage lifecycle development of a risk assessment tool for SMEs in Ireland. In each stage the tool was envisioned to address a double goal of assessing pertaining risk to a SME but also continuously raise awareness about threats, risks, and protections aspects relevant for the involved stakeholders. The development included design work with experts, walk-through and think-aloud studies, as well as focus groups that provided qualitative responses to the utility, scope, and implementability of the tool. A comparison of the tool in conjunction with an extensive thematic analysis of the qualitative responses offered evidence of the tool's versatility and adaptability for any SME, likely also outside of Ireland, and customizable to the domain of an SME's primary operation.

Abstract: More and more immersive environments support third-party apps, leading to concerns about the trustworthiness of user interfaces (UIs), which attackers could exploit, endangering users. Although security warnings effectively safeguard users by highlighting risks, most studies primarily target security indicators for app transitions in virtual reality. Our research broadens this focus by providing a systematic, data-driven investigation of security warnings for third-party applications in immersive environments. We analyzed a decade's worth of top VR interactions and security conference findings and assessed the top 10 free VR apps from two leading stores each. From our design process, we implemented four warnings. Through two user studies involving 61 participants, we measured their responses to these warnings during virtual object interactions. Our findings indicate that a red glow on an object was the most effective warning, frequently associated with danger, while pop-up warnings were the least effective.

Meta Review: This paper examines security warnings for third-party apps in immersive environments, demonstrating that a red glow on an object might be the most effective warning while pop-up warnings are less effective. This work provides valuable empirical evaluation of security warning designs in immersive environments, and these designs could be especially foundational to future work. The methodology is very comprehensive, first analyzing a decade of VR interactions, then assessing VR apps, followed by user studies to evaluate the efficacy of a different variety of warnings in the VR space. The results are also interesting, showing how some of the designs are effective, while others less so.

Abstract: The current Cyber Insurance environment provides insufficient incentives for companies to maintain high-security readiness, paralleling ``good driver'' discounts in auto insurance. Additionally, navigating rigid protocols, given the focus on timely compliance in the aftermath of data breaches, can complicate claim approval. Additionally, rigid protocol navigation in the aftermath of data breaches can complicate claim approval. This study explores the potential benefits of customized Cyber Liability Insurance (CLI) policies tailored to an organization's specific security posture. Our hypothesis suggests that such tailored insurance solutions can enhance cost-effectiveness and reduce coverage gaps. Three key research questions guide our investigation. The work focuses on how organizations make decisions regarding CLI providers and coverage, revealing non-methodical selection processes often led by non-security personnel like CFOs. We further identify practical pain points across CLI's lifecycle, highlighting CISOs' concerns about coverage costs and clarity. Finally, we investigate customization solutions to align CLI with organization-specific risk, emphasizing the limitations of a one-size-fits-all approach. In summary, this research advances the field of CLI, offering insights to improve cost-effectiveness, transparency, and alignment with organization-specific risk, contributing to the enhancement of cyber liability insurance in an evolving landscape of cyber threats and vulnerabilities.

Meta Review: The paper examines how organizations select Cyber Liability Insurance (CLI), the challenges they encounter, and the potential advantages of customized CLI policies designed to align with an organization's specific security posture. The research employs a rich study design that integrates a survey, a policy comparison analysis done in the wild, and interviews with CISOs to gain a deeper understanding of CLIs. A significant contribution of the paper is the establishment of an LLC to conduct a comparative analysis of the CLIs and the purchase of two CLI policies to understand how these systems function in the real world. This approach provides firsthand insights into the various issues and pain points in selecting CLIs, while also highlighting the mismatches between the needs of organizations and offerings of providers in practice. The entire survey, policy analysis, and interview instruments are made open-source, thereby promoting open science. Overall, this paper has important implications for organizations choosing CLIs, particularly those with limited amount of cybsersecurity know-how.

Abstract: Recent growth and proliferation of malware has tested practitioners' ability to promptly classify new samples according to malware families. In contrast to labor-intensive reverse engineering efforts, machine learning approaches have demonstrated increased speed and accuracy. However, most existing deep-learning malware family classifiers must be calibrated using a large number of samples that are painstakingly manually analyzed before training. Furthermore, as novel malware samples arise that are beyond the scope of the training set, additional reverse engineering effort must be employed to update the training set. The sheer volume of new samples found in the wild creates substantial pressure on practitioners' ability to reverse engineer enough malware to adequately train modern classifiers. In this paper, we present MalMixer, a malware family classifier using semi-supervised learning that achieves high accuracy with sparse training data. We present a novel domain-knowledge-aware technique for augmenting malware feature representations, enhancing few-shot performance of semi-supervised malware family classification. We show that MalMixer achieves state-of-the-art performance in few-shot malware family classification settings. Our research confirms the feasibility and effectiveness of lightweight, domain-knowledge-aware feature augmentation methods and highlights the capabilities of similar semi-supervised classifiers in addressing malware classification issues.

Meta Review: The authors propose MalMixer, a method for malware family classification using semi-supervised learning that achieves high accuracy with sparse training data. All reviewers agree that the data augmentation strategy is very interesting on its own, and a very solid approach to improve the accuracy of models. The data augmentation presents an elegant way for generating new training samples while preserving semantic relations between features of binaries. The authors also conducted a very thorough experimental evaluation of their propose methodology, and can perform well even in few-shot learning scenarios with few examples of a malware families. They demonstrates solid performance improvements over the state of the art, and an ablation study which confirms usefulness of the MalMixer components. While the paper is limited to a closed-world evaluation and static features (limiting the impact in consideration of, e.g., malware packers), the reviewers found sufficient merit in the proposed data augmentation methodology. Furthermore, all reviewers noted that the paper is very well written and easy to follow, which, together with releasing their code and artifacts, improves accessibility of this research to future researchers.

Abstract: We conduct the first comprehensive security study on representative port forwarding services (PFS), which emerge in recent years and make the web services deployed in internal networks available on the Internet along with better usability but less complexity compared to traditional techniques (e.g., NAT traversal techniques). Our study is made possible through a set of effective and efficient methodologies, which are designed to uncover the technical mechanisms of PFS, experiment attack scenarios for PFS protocols, automatically discover and snapshot port-forwarded websites (PFWs) at scale, and classify PFWs into well-observed categories. Leveraging these methodologies, we have observed the widespread adoption of PFS with millions of PFWs distributed across tens of thousands of ISPs worldwide. Furthermore, 32.31% PFWs have been classified into website categories that serve access to critical infrastructure or data, such as, web consoles for industrial control systems, IoT controllers, code repositories, and office automation systems. And 18.57% PFWs didn't enforce any access control for external visitors. Also identified are two types of attacks inherent in the protocols of Oray (one well-adopted PFS provider), and the notable abuse of PFSes by malicious actors in activities such as malware distribution, botnet operation and phishing.

Meta Review: The paper presents the authors' findings on the first security study of port-forwarding services (PFSs). Their study analyses two popular PFSs and sheds light on the ecosystem along different dimensions: a content-focused analysis (which websites are being served), a methodological perspective (how the ecosystem is evolving), and a security assessment (which authentication measures are in place). The work further outlines possible attacks on PFS protocols, including a man-in-the-middle attack, which the authors disclosed responsibly. The reviewers agreed that focusing on this previously overlooked ecosystem is a significant strength of the work and may pave the (methodological) road for more elaborate, up-to-date assessments and studies in the future.

Abstract: With the increasing prevalence of progressive web applications, web workers have found themselves in the spotlight. Indeed, workers have drastically changed the Web attack surface. Prior work has demonstrated unique flaws enabled by the introduction of service workers, e.g., their computation, persistence, caching, or ability to process web push messages. As for XSS (cross-site scripting), its treatment in web worker contexts by various Web stakeholders is largely unsatisfactory and insufficient. Content injection attacks are still largely framed from the perspective of webpages, including the defense mechanisms. In this work, we undertake the first comprehensive security analysis of content injection attacks in all web workers, with a focus on XSS. To do so, we start by defining a clear threat model, considering that workers are dedicated to code execution, but lack a DOM, meaning that the ways they are infected by attackers are different from web pages. Then, we devise a rigorous methodology we applied to a large-scale dataset consisting of 4,757,077 workers collected from 56,945,781 websites in the wild. As a result, we confirmed, through extensive manual vetting, different forms of server and client XSS in at least 89,945 workers. We reported our findings to many of the vendors (covering 99.8% of the vulnerable workers), and many of them acknowledged the issue, and at the time of this writing, the issue has been fixed for at least 82.3% of the vulnerable workers. From a defensive side, we engaged in an extensive discussion with proposals and demonstrated how to extend the Content Security Policy and the importScripts function implementations to allow the expression and support of finer-grained policies like nonces and hashes which we believe can help thwart most of the attacks discussed in this work.

Meta Review: The paper presents a large-scale study of cross-site scripting (XSS) vulnerabilities in web worker contexts. By examining XSS risks in dedicated, shared, and service workers, the authors apply a semi-automated methodology to a massive dataset of over 56 million websites, identifying 89,945 risky workers. The study also proposes extending Content Security Policy (CSP) and Subresource Integrity (SRI) to mitigate these risks and discusses their proposals with the W3C Web Application Security Working Group. Overall, this paper makes a timely and valuable contribution toward understanding and addressing XSS in web workers.

Abstract: Kubernetes has become the dominant platform for managing containerized applications, but its native Secrets management mechanisms introduce security vulnerabilities, especially in environments where third-party applications may have excessive permissions. In this paper, we present KubeKeeper, a comprehensive solution for protecting Kubernetes Secrets against leakage due to excessive permissions. KubeKeeper automatically encrypts Secrets and ensures that only explicitly authorized Pods can access their decrypted form. This is achieved by leveraging Kubernetes’ dynamic admission control capabilities to operate transparently without requiring changes to infrastructure or application code. We evaluated KubeKeeper on a diverse set of 335 Kubernetes applications and demonstrate that it successfully protects Secrets against all identified excessive permissions, without introducing performance degradation during execution or any significant overhead during creation and deployment.

Meta Review: The paper presents KubeKeeper, a protection mechanism for Kubernetes designed to mitigate secrets leakage caused by overpermission vulnerabilities. The reviewers agree that the problem addressed by the paper is both timely and relevant, and that KubeKeeper offers a practical and effective solution. Notably, the solution integrates with existing Kubernetes deployments, requiring minimal changes to infrastructure or application code. The paper also introduces a static analysis tool to detect overpermission vulnerabilities in Kubernetes applications. KubeKeeper was evaluated on a dataset of 335 Kubernetes applications, with the static analysis tool revealing widespread security issues.

Abstract: Trusted Execution Environment (TEE) has become increasingly popular in privacy-protected cloud computing, and its rapid development has led to the availability of various heterogeneous TEE platforms on cloud servers. To facilitate portable TEE applications on heterogeneous TEE platforms, portable languages or intermediate representations (IRs) with platform-dependent TEE runtimes are adopted. However, existing remote attestation solutions for portable TEE applications follow a nested attestation pattern, i.e., attesting only the TEE runtime and relying on the TEE runtime to measure the loaded portable application, leading to potential security issues. On the other hand, directly packing the TEE runtime and the portable application into an enclave for secure attestation undermines the portability of the portable TEE applications. In this paper, we introduce the concept of portable identities to identify portable TEE applications, and propose a layered attestation framework, Latte, achieving both security and portability in attesting portable TEE applications. We provide a prototype implementation of Latte to validate its practicality, with WebAssembly as the portable IR, and Intel SGX and RISC-V Penglai as the exemplar heterogeneous TEEs. The evaluation demonstrates that Latte introduce minimal performance overhead compared with the nested attestation pattern.

Meta Review: The paper presents LATTE, a framework for layered attestation of portable TEE applications, implemented for both SGX and Penglai. It includes a rigorous description of the proposed security primitives. Building on MAGE (USENIX Security 2022), LATTE primarily decouples attestation between runtimes and applications through the use of portable identities. It is designed to integrate with existing attestation mechanisms, and the implementation is planned to be open-sourced.

Abstract: While greybox fuzzing is routinely applied in production environments with great success, directed greybox fuzzing has struggled to gain real-world adoption—despite the great (intuitive) promise and the many optimizations proposed in literature. In practice, directed fuzzers struggle for three critical issues. First, popular implementations build on and compare to ancient baselines, often derived from AFLGo. Unfortunately, none of the optimizations that are essential for performance in modern greybox fuzzers are available in these baselines. As a result, we find reported improvements in directed fuzzing are often only “imaginary” and do not lead to better performance on a modern baseline. Second, directed fuzzing evaluations commonly ignore or misinterpret important factors affecting fuzzing overhead—such as build times and timeouts. As design decisions now build on unreliable data, we find the directed fuzzers perform worse than expected in practice. Third, while almost all directed fuzzers rely on (expensive) analysis stacks, such as points-to and reachability analysis components, they often opt for very different implementations. Since these implementations have their own unique benefits and drawbacks, we find performance differences of directed fuzzers are frequently due to these components rather than the proposed directed fuzzing optimization. In this paper, we investigate the practical impact of these issues by means of an analysis and evaluation of a representative set of popular directed greybox fuzzers. As a way forward, we then present LibAFLGo, a modular directed fuzzing framework that addresses all three issues and allows one to directly compare different directed fuzzing policies on top of a modern fuzzing stack. Our experimental results on state-of-the-art directed fuzzing policies provide two main insights. First, the original AFLGo policies outperform more recent directed fuzzing policies when testing on a modern fuzzing stack. Second, none of the directed fuzzing policies can favorably compete with (nondirected) LibAFL, which scored better overall performance across benchmarks. As such, the quest for efficient directed fuzzing policies must continue.

Meta Review: This paper identifies three important issues with the implementation and evaluation of directed greybox fuzzers in the literature. From their analysis, the authors conclude that the current evaluation practices are insufficient and lead to misleading results, and propose a unified platform, LibAFLGo, which can be used to build and fairly evaluate directed fuzzers. LibAFLGo is based on LibAFL, a state-of-the-art (nondirected) fuzzing library, enabling comparison of directed and nondirected fuzzers. The authors also reimplement existing directed fuzzing approaches on this platform, revealing that they do not perform better than nondirected approaches and that their performance as reported in the original papers might have resulted from the identified issues of evaluation practices. The reviewers agree that this analysis and the released open-source platform and modernized reimplementations of existing approaches are important motivators and enablers of future work in the area of directed fuzzers.

Abstract: Speculative loading of memory, called hardware prefetching, is common in modern CPUs and may cause microarchitectural side-channel vulnerabilities. As prior work has shown, prefetching can be exploited to bypass process isolation and leak secrets. However, to this date, no effective and efficient countermeasure has been presented that secures software on affected systems. Often, disabling prefetching permanently is considered the only reasonable defense, despite the significant performance penalties this entails. In this work, we propose PreFence, a fine-grained and scheduling-aware defense against prefetching-based attacks for any platform where the prefetcher can be disabled. PreFence extends the process scheduler to be aware of security requirements of individual processes and to manage prefetcher state to protect from malicious parallel processes, even on SMT-enabled platforms. This allows to efficiently disable the prefetcher only during security-critical operations, with a single system call. Library and application developers can protect their code with minimal changes, users can protect entire legacy applications with a wrapper. We implement our countermeasure for an x86_64 and an ARM processor. We evaluate PreFence on two prior-work attacks and find that it reliably stops prefetch leakage. In addition, we observe that PreFence causes only negligible performance impact when no security-relevant code is executed. Finally, we evaluate the performance of a real-world web-server application that uses PreFence to protect security-critical code for HTTPS handling. We find that our countermeasure allows the application to significantly benefit from the prefetcher (running up to 15.8% (Intel) and 7.2% (ARM) faster on average), while at the same time achieving the same security compared to disabling the prefetcher permanently.

Meta Review: This paper proposes PreFence, a fine-grained scheduling-aware defense against prefetcher-based attacks. The process scheduler is modified to control the prefetcher based on the security constraints of a given process. The defense requires minimal changes in user software, as PreFence implements a fine-grained control using system calls. The countermeasure is implemented on both x86 and ARM processors to show its applicability to different types of processors. An additional contribution is the discussion on the granularity of data memory-dependent prefetchers and their security implications. The mitigation's runtime overhead is less than 3%, which is practical in real-world scenarios. Overall, the paper provides an efficient security-aware prefetcher mechanism applicable across different processors.

Abstract: Microarchitectural attacks like Rowhammer and Spectre rely on precisely targeting specific memory page frames, despite the inherent unpredictability of memory allocation. Due to the lack of a generic mechanism to accurately place the target data in the pages of interest, these attacks resort to spraying their target or scanning the entire physical memory for it. These approaches, however, suffer from unreliability and inefficiency. In contrast, the deterministic behavior of page allocators presents an opportunity to enhance existing attacks or enable new ones. This paper introduces Rubicon, a novel technique for page-granular physical memory massaging within the Linux kernel’s Zoned Buddy Allocator (ZBA). Rubicon leverages three new primitives that enable placing a page frame at the head of any chosen ZBA free list, ensuring it is prioritized for allocation regardless of its initial state or per-CPU freelist association. Using Rubicon, we build the first deterministic privilege escalation Rowhammer exploit on x86 with a success rate of 100%. Our integration of Rubicon into a recent Spectre attack shows that the root hash of /etc/shadow can now be leaked in 27.8 and 9.5 seconds on AMD and Intel systems — a 6.8× and 284× speedup over the original attack, respectively. We also propose and evaluate practical mitigations for Rubicon, which limit page movement between ZBA lists with negligible performance and fragmentation impact.

Meta Review: The paper is the first work dedicated solely to the topic of memory massaging, the practice of manipulating the system's memory layout to facilitate attacks. It presents a principled technique that allows reliable and efficient placement of victim pages at vulnerable locations. Beyond the contribution of new techniques, the reviewers appreciate the quality of the writing as well as the promise of future source code availability, which would allow future research to easily apply the current results and to build on them.

Abstract: Decompilers play a crucial role in the detailed analysis of malware or firmware, particularly because control-flow structuring allows the recovery of high-level code that is more readable to human analysts. Despite the ongoing debate over their usage of gotos to work around constraints during control-flow structuring, pattern-matching approaches remain prevalent among both commercial and open-source decompilers. With the emergence of pattern-independent restructuring techniques, various attempts have been made to overcome readability limitations, especially concerning the use of gotos. However, despite these advancements, recent approaches often fail to thoroughly address several inherent challenges of control-flow structuring, thereby impacting output quality or practicality. In this paper, we systematize the intrinsic challenges of control-flow structuring that every approach must address. In addition, we review existing methods, comparing them, while highlighting both their advantages and limitations with respect to these challenges. Specifically, we emphasize the practicability issues of current pattern-independent restructuring techniques and discuss whether and how future methods might overcome them. Finally, we explore the theoretical potential to mitigate some of these challenges by suggesting methodology ideas for various aspects of control-flow structuring. Overall, this paper enables other researchers to make informed decisions when developing or enhancing control-flow structuring methods, thereby preventing negative side-effects arising from the interdependence of challenges.

Meta Review: The paper attempts a systemization of the techniques available for making the recovered code of a program, as produced by a decompiler, more readable. More precisely, this SoK paper explains existing control-flow structuring techniques and how they address a set of identified challenges; a work that can be beneficial to someone that seeks for an overview on the field. Currently, there are two major approaches for this: (a) using known patterns to reconstruct the control-flow graph, and (b) using a pattern-independent methodology. The paper points to possible improvements over existing techniques, which can lead to further research. The summary of the challenges related to control-flow structuring can guide further research into discussing the various tradeoffs.

Abstract: The widespread adoption of EMV (Europay, Mastercard, and Visa) contactless payment systems has greatly improved convenience for both users and merchants. However, this growth has also exposed significant security challenges. This paper provides a comprehensive analysis of security vulnerabilities in EMV contactless payments, particularly within the open-loop systems used by Visa and Mastercard. We categorize attacks based on their objectives across three key areas: application selection, cardholder authentication, and transaction authorization. We replicate the attacks on Visa and Mastercard protocols using our experimental platform to determine their practical feasibility and offer insights into the current security landscape of contactless payments. Our study also includes a detailed evaluation of the underlying protocols, along with a comparative analysis of Visa and Mastercard, highlighting vulnerabilities and recommending countermeasures.

Meta Review: This paper provides a thorough analysis of security vulnerabilities in EMV contactless payment systems, particularly within the open-loop systems used by Visa and Mastercard. It categorizes attacks into different categories and replicates these attacks on Visa and Mastercard protocols to assess their practical feasibility. The paper also includes a comparative analysis of the protocols, offering valuable insights into their respective vulnerabilities and recommending countermeasures. The work makes a valid contribution by testing the practical applicability of known attacks and providing actionable recommendations for enhancing security in contactless payment systems. Overall, the paper is a valuable addition to the discourse on EMV payment security.

Abstract: Android Protected Confirmation (APC) is an authentication protocol designed by Google. It leverages the extra security of the Trusted Execution Environment (TEE) to secure transactions even in the presence of a compromised OS. The intended security guarantee for APC is that if a transaction has been signed under APC, then the user must have previously given its explicit consent, even if an attacker has gained root access to the victim’s Android OS. In this paper, we present a security analysis of APC in the Universal Composability (UC) framework. We uncover two attacks on the design of the protocol which allow a root adversary to issue transactions without the user consenting to them. We provide an attack implementation on a Google Pixel phone, and propose light-weight fixes. Finally, we specify the ideal UC functionality capturing the intended security guarantees for APC, and prove that the fixed protocol UC-realizes it.

Meta Review: The paper "Attacking and Fixing the Android Protected Confirmation Protocol" investigates the APC protocol, which is used to secure critical mobile transactions using the Trusted Execution Environment (TEE). It presents the first formal analysis of the protocol and uncovers two vulnerabilities, demonstrating their impact through proof-of-concept attacks. The authors propose effective mitigations for these issues. Notable for its clear structure and editorial quality, the paper is both informative and accessible. Overall, the paper deepens our understanding of the APC protocol and provides general insights relevant to other security protocols.

Abstract: The proof-of-stake (PoS) protocols aim to reduce the unnecessary computing power waste seen in Bitcoin. Various practical and provably secure designs have been proposed, like Ouroboros Praos (Eurocrypt 2018) and Snow White (FC 2019). However, the essential security property of unpredictability in these protocols remains insufficiently explored. This paper delves into this property in the cryptographic setting to achieve the ``best possible'' unpredictability for PoS protocols. We first present an impossibility result for {\em all} PoS protocols under the \emph{single-extension} design framework, where each honest player extends one chain per round. The state-of-the-art permissionless PoS protocols (e.g., Praos, Snow White, and more), are all under this single-extension framework. Our impossibility result states that, if a single-extension PoS protocol achieves the best possible unpredictability, then this protocol cannot be proven secure unless more than $73\%$ of stake is honest. To overcome this impossibility, we introduce a new design framework called \emph{multi-extension} PoS, allowing each honest player to extend {\em multiple} chains using greedy strategy in a round. This strategy allows us to construct a class of PoS protocols that achieve the best possible unpredictability. Additionally, we design a new tiebreak rule for the multi-extension protocol to choose the best chain that can be extended faster, ensuring that the adversary cannot slowdown the chain growth of honest players. It is noteworthy that these protocols can be proven secure, assuming a much smaller fraction (e.g., 57\%) of stake to be honest. For a comprehensive security analysis in the cryptographic setting, we develop several new techniques. Analyzing chain growth becomes highly non-trivial as players can extend multiple chains. We introduce a new analysis framework using the Markov chain to assess the chain growth of a multi-extension protocol. To prove the common prefix property, we introduce a concept called ``virtual chains'' and present a reduction from the regular version of the common prefix to ``common prefix w.r.t. virtual chains.''

Meta Review: This paper provides a formal study of predictability in Proof-of-Stake (PoS) protocols, presenting two main contributions: an impossibility result for single-extension protocols and a new multi-extension protocol that addresses this limitation. The authors show that no single-extension protocol can be predictability-secure with less than 73% honest stake, establishing a lower bound in this setting. To overcome this, they propose a multi-extension protocol using a D-distance greedy strategy that achieves security with only 57% honest stake. The analysis includes key blockchain properties such as chain growth, common prefix, and chain quality. The paper also introduces analytical techniques that may be useful more broadly in PoS research. In summary, the paper presents a meaningful advancement in PoS research. It is a valuable contribution to the literature on blockchain consensus protocols.

Abstract: Validators in permissionless, large-scale blockchains, such as Ethereum, are typically payoff-maximizing, rational actors. Ethereum relies on in-protocol incentives, like rewards for correct and timely votes, to induce honest behavior and secure the blockchain. However, external incentives, such as the block proposer's opportunity to capture maximal extractable value (MEV), may tempt validators to deviate from honest protocol participation. We show a series of commitment attacks on LMD GHOST, a core part of Ethereum's consensus mechanism. We demonstrate how a single adversarial block proposer can orchestrate long-range chain reorganizations by manipulating Ethereum's reward system for timely votes. These attacks disrupt the intended balance of power between proposers and voters: by leveraging credible threats, the adversarial proposer can coerce voters from previous slots into supporting blocks that conflict with the honest chain, enabling a chain reorganization. In response, we introduce a novel reward mechanism that restores the voters' role as a check against proposer power. Our proposed mitigation is fairer and more decentralized – not only in the context of these attacks – but also practical for implementation in Ethereum.

Meta Review: This paper investigates reorganization attacks on the LMD GHOST protocol, offering an analysis inspired by selfish-mining strategies. It presents novel and interesting attacks on Ethereum's reward mechanism, providing multiple perspectives through game-theoretic reasoning, incentives, and attacker motivation. The incremental design of the attacks enhances clarity and illustrates how adversaries with different capabilities may exploit the system. The paper also suggests a potential alternative to the current reward mechanism, called DAG votes. The paper analyzes the DAG votes mechanism, showing its effectiveness through a theoretical equilibrium result and a practical overhead estimate. Overall, the paper makes a valuable and timely contribution to the understanding of Ethereum's consensus security.

Abstract: Local Differential Privacy (LDP) mechanisms consist of (locally) adding controlled noise to data in order to protect the privacy of their owner. In this paper, we introduce a new cryptographic primitive called LDP commitment. Usually, a commitment ensures that the committed value cannot be modified before it is revealed. In the case of an LDP commitment, however, the value is revealed after being perturbed by an LDP mechanism. Opening an LDP commitment therefore requires a proof that the mechanism has been correctly applied to the value, to ensure that the value is still usable for statistical purposes. We also present a security model for this primitive, in which we define the hiding and binding properties. Finally, we present a concrete scheme for an LDP staircase mechanism (generalizing the randomized response technique), based on classical cryptographic tools and standard assumptions. We provide an implementation in Rust that demonstrates its practical efficiency (the generation of a commitment requires just a few milliseconds). On the application side, we show how our primitive can be used to ensure simultaneously privacy, usability and traceability of medical data when it is used for statistical studies in an open science context. We consider a scenario where a hospital provides sensitive patients data signed by doctors to a research center after it has been anonymized, so that the research center can verify both the provenance of the data (i.e. verify the doctors’ signatures even though the data has been noised) and that the data has been correctly anonymized (i.e. is usable even though it has been anonymized).

Meta Review: The reviewers agreed that this paper makes a strong and well-motivated contribution by introducing a new cryptographic primitive as LDP (local differential privacy mechanisms ) commitments. It combines cryptographic commitments with different local privacy mechanisms. The authors provide formal definitions, a detailed security model, and proofs for this new concept e.g., LDP-hiding and binding. The work is based on standard cryptographic assumptions and demonstrates practicality through an efficient Rust implementation. Finally, they show that its relevance to anonymized data sharing, especially in medical contexts, and the introduction of new security concepts make it valuable in this area.

Abstract: Cybersquatting refers to the practice where attackers register a domain name similar to a legitimate one to confuse users for illegal gains. With the growth of the Non-Fungible Token (NFT) ecosystem, there are indications that cybersquatting tactics have evolved from targeting domain names to NFTs. This paper presents the first in-depth measurement study of NFT cybersquatting. By analyzing over 220K NFT collections with over 150M NFT tokens, we have identified 8,019 cybersquatting NFT collections targeting 654 popular NFT projects. Through systematic analysis, we discover and characterize seven distinct squatting tactics employed by scammers. We further conduct a comprehensive measurement study of these cybersquatting NFT collections, examining their metadata, associated digital asset content, and social media status. Our analysis reveals that these NFT cybersquatting activities have resulted in significant financial impact, with over 670K victims falling to these scams, leading to a total financial exploitation of \$59.26 million. Our findings demonstrate the urgency to identify and prevent NFT squatting abuses.

Meta Review: The reviewers generally agree that this paper presents a valuable measurement study on cybersquatting in the NFT ecosystem, identifying seven distinct naming tactics used by scammers, analyzing their social media presence, and quantifying financial impacts. The paper's comprehensive approach to examining multiple dimensions of NFT cybersquatting, including detailed analysis of scammer behavior and potential coordinated campaigns, represents a timely contribution to understanding security threats in the growing Web3 space. The research is strengthened by its scale, analyzing a large dataset that enables meaningful conclusions, and by the authors' decision to open-source a subset of their collected data. Ethical considerations were also highlighted, with questions about whether appropriate disclosure was made to affected platforms, especially given the financial impact uncovered—scammers collectively profiting over $59.26 million from these activities, affecting more than 670K victims. Following the submission of revisions, the paper has been accepted, with recommendations for the camera-ready version to further address methodology transparency, particularly regarding validation metrics and automation potential. The paper contributes valuable insights into emerging security threats in the Web3 ecosystem and will benefit the community's understanding of cybersquatting in the context of NFT.

Abstract: Lattice-based cryptography is at the basis of most NIST-recommended post-quantum cryptography (PQC) schemes, required to thwart the threat posed by the eventual construction of large-scale quantum computers. At the same time, lattices enable more advanced cryptographic constructions such as fully homomorphic encryption (FHE), which is increasingly being used for privacy-preserving machine learning as governing bodies impose more restrictions on analyzing user data. In this work, we delve into the efficiency and trade-off assessment of polynomial multiplication algorithms and their applications to PQC, FHE, and other schemes. Such algorithms are at the core of lattice-based cryptography and may become a critical bottleneck when deploying PQC- and FHE-based solutions on resource-constrained edge devices, such as microcontrollers and embedded devices. We propose a systematic and formal approach for optimizing polynomial multiplication in a quotient ring, which builds upon trade-offs between the so-called incompleteness of the Number Theoretic Transform (NTT), associated ring prime modulus, target platform constraints, and required security level. The proposed method strives to outperform conventional polynomial multiplication algorithms like regular NTT, Karatsuba, and Schoolbook. Besides efficiency, the presented family of incomplete NTT settings supports a larger set of prime moduli. This property enables new trade-offs for algorithms like the FIPS-approved module-lattice-based key encapsulation mechanism (ML-KEM), and faster amortized bootstrapping in FHE schemes. In particular, we show how to obtain shorter ciphertexts in ML-KEM with only a modest hit in performance compared to the current standard and save 6-42% in the NTT computation of a state-of-the-art FHE solution.

Meta Review: The paper studies incompleteness of the number theoretic transformation (NTT) in the context of lattice based cryptographic primitives, specifically key encapsulation mechanism (KEM) and fully homomorphic encryption (FHE). The paper provides a comprehensive performance analysis of Kyber and FHEW/TFHE-like schemes, reporting their improvements in both. The authors also provide concrete parameter selections for both applications of their family of incomplete NTT, with new trade-offs. Overall, the topic itself is fundamentally important in lattice based-cryptography and the chosen use-cases are popular post-quantum cryptosystems, hence their analysis would give helpful insight for both lattice based scheme designers and implementors.

Abstract: Distributed social networking services (SNSs) recently received significant attention as an alternative to traditional, centralized SNSs, which have inherent limitations on user privacy and freedom. We provide the first in-depth security analysis of Nostr, an open-source, distributed SNS protocol developed in 2019 with more than 1.1 million users. We investigate the specification of Nostr and the client implementations and present a number of practical attacks allowing forgeries on various objects, such as encrypted direct messages (DMs), by a malicious user or a malicious server. Even more, we show a confidentiality attack against encrypted DMs by a malicious user exploiting a flaw in the link preview mechanism and the CBC malleability. Our attacks are due to cryptographic flaws in the protocol specification and client implementation, some of which in combination elevate the forgery attack to a violation of confidentiality. We verify the practicality of our attacks via Proof-of-Concept implementations and discuss how to mitigate them.

Meta Review: The paper "Not in The Prophecies: Practical Attacks on Nostr" studies the security of Nostr a widely deployed distributed Social Networking Services protocol. The reviewers unanimously agree on the contributions of the paper which are three-fold: first, a security analysis of the protocol at the specification level but also regarding some of its implementations. Second, the description of several uncovered and severe flaws found at both levels. Authors followed a clear responsible disclosure process to report these attacks. Third, a description and discussion of different fixes, whose effectiveness have been evaluated by conducting concrete experiments (not reproduced because the code was not open source at time of the review, but will be upon publication). In summary, the paper makes a significant contribution to the domain of "distributed Social Networking Services", and deserves to be included in the EuroS&P'25 program.

Abstract: We present a system, Sandi, for creating trust through accountability. Concretely, we focus on online communication scenarios, where the communicating parties do not know each other, yet would benefit from a degree of initial trust. Sandi can be seen as a reputation system that measures bad behavior, with strong integrity protections and resistance to manipulation. Unlike most reputation systems, Sandi is entirely based on ``downvotes'' and therefore requires strong privacy guarantees to prevent retaliation. It utilizes a ticket-based reporting mechanism to limit who can report. We also prove that Sandi incentivizes good behavior in a well-defined sense. Sandi is by design unidirectional, so that message senders have Sandi scores and receivers can report them for inappropriate communication, but it is designed to benefit both senders and receivers. Senders benefit, as receivers are more likely to react to communication with the added trust signal. Receivers benefit from seeing senders' scores, allowing them to make more informed decisions about which senders to trust. Receivers do not need registered accounts and neither senders nor receivers need long-term keys. Sandi guarantees score integrity, communication privacy, reporter privacy to protect reporting receivers, and sender unlinkability. Sandi can be implemented on top of any communication system that allows for small binary data transfer.

Meta Review: The paper paper is well-motivated and the framework that it proposes is simple, efficient, and based on standard cryptographic assumptions--which lends itself to scalability and deployment. Furthermore, the proposed framework can be easily implemented on a wide range of platforms and communication protocols, hence it is versatile for practical applications. The paper presents well-argued and clearly defined security and privacy properties, which are essential to gain users' trust in sensitive communication contexts. Beyond that, Sandi presents significant technical differences compared to prior works and supplements the theory with a game theoretic analysis.

Abstract: Encrypted data structures are essential for designing efficient encrypted search algorithms and secure databases. However, a critical aspect that has not been adequately addressed is the concurrent nature of modern databases, which allow multiple operations to be executed simultaneously. Agarwal, Kamara, and Moataz (Asiacrypt 2024) recently initiated the study of concurrent encrypted data structures, introducing formalisms for their design and analysis. Building on their foundational work, we adapt their security definitions to support sequential consistency instead of linearizability. While linearizability offers a strong correctness guarantee by ensuring operations appear to occur instantaneously, sequential consistency allows operations to be executed in a consistent order without immediate synchronization across clients, making it more efficient for concurrent environments. We present a new concurrent encrypted multimap (EMM), denoted as SCM, which achieves sequential consistency and provides significant improvements in both asymptotic and empirical efficiency compared to their linearizable EMM scheme, TST. Additionally, we develop a benchmarking tool specifically designed to assess the performance of concurrent EMMs, extending the widely used YCSB benchmark to accommodate multimaps that allow multiple values to be associated with a single key. Our results demonstrate that SCM outperforms TST across various workloads and datasets, especially as the number of concurrent operations increases. In our experiments with 16 concurrent clients, SCM has 46x-357x faster P95 read latency (with the best read performance as the overall percentage of reads decreases) and 1x-69x faster P95 write latency (with the best write performance as the percentage of writes approaches 50%) than TST, demonstrating how SCM effectively balances efficiency and correctness.

Meta Review: Encrypted data structures are essential for modern secure databases. This work extends prior work on concurrent encrypted data structures by incorporating a weaker model of concurrency called sequential consistency. Reviewers were impressed by the authors' careful and detailed review response, which made a significant difference to the final decision.

Abstract: Messaging Layer Security (MLS) is a Secure Group Messaging protocol that uses for its handshake a binary tree – called a Ratchet Tree – in order to reach a logarithmic communication cost in the number of group members. This Ratchet Tree represents users as its leaves; therefore any change in the group membership results in adding or removing a leaf in the tree. MLS consequently implements what we call a tree evolution mechanism, consisting of a user add algorithm – determining where to insert a new leaf – and a tree expansion process – stating how to increase the size of the tree when no space is available for a new user. The tree evolution mechanism currently used by MLS is designed so that it naturally left-balances the Ratchet Tree. However, such a tree structure is often quite inefficient in terms of communication cost. Furthermore, one may wonder whether the binary tree used in that Ratchet Tree has a degree optimized for the features of MLS. Therefore, we study in this paper how to improve the communication cost of the handshake in MLS – realized through an operation called a commit – by considering both the tree evolution mechanism and the tree degree used for the Ratchet Tree. To do so, we determine the tree structure that optimizes its communication cost and we propose optimized algorithms for both the user add and the tree expansion processes, that allow to remain close to that optimal structure and thus to have a communication cost as close as possible to the optimum. We also determine the Ratchet Tree degree that is best suited to a given set of parameters induced by the encryption scheme used by MLS. This study shows that when using classical (i.e. pre-quantum) ciphersuites, a binary tree is indeed the most appropriate Ratchet Tree; nevertheless, when it comes to post-quantum algorithms, it generally becomes more interesting to use instead a ternary tree. Our improvements do not change the TreeKEM protocol and are easy to implement. With parameter sets corresponding to practical ciphersuites, they reduce TreeKEM’s communication cost by 5 to 10%. In particular, the 10% gain appears in the Post-Quantum setting – when both an optimized tree evolution mechanism and a ternary tree are necessary –, which is precisely the context where any optimization of the protocol’s communication cost is welcome, due to the large bandwidth of PQ encrypted communication.

Meta Review: This paper thoroughly investigates the influence of the underlying tree structure on the performance of the Secure Group Messaging protocol "Messaging Layer Security" (MLS). With optimization for post-quantum algorithms in mind, the paper proposes a new tree structure and a tree evolution mechanism that considerably decreases the communication cost. The paper thus focuses on a newly identified problem of a real-world application protocol and provides a thoroughly evaluated improvement.

Abstract: The open-source nature of the Android Open Source Project (AOSP) allows Original Equipment Manufacturers (OEMs) to customize the Android operating system, contributing to what is known as \textit{Android fragmentation}. Google has implemented the Compatibility Definition Document (CDD) and the Compatibility Test Suite (CTS) to ensure the integrity and security of the Android ecosystem. However, the effectiveness of these policies and measures in warranting OEM compliance remains uncertain. This paper empirically studies for the first time the nature of OEM customizations in the Android TLS protocol stack, and their security implications on user-installed mobile apps across thousands of Android models. We find that approximately 80% of the analyzed Android models deviate from the standard AOSP TLS codebase, and that OEM customizations often involve code changes in functions used by app developers for enhancing TLS security like end-point and certificate verification. Our analysis suggests that these customizations are likely influenced by factors such as manufacturers' supply chain dynamics and patching prioritization tactics, including the need to support legacy components. We conclude by discussing potential root causes and highlighting the need for stricter policy and supply chain controls, and more effective ecosystem-wide patching processes.

Meta Review: This paper presents a large-scale empirical analysis of Android TLS protocol stack customizations based on more than 53,000 firmware images from 370 OEMs, using a solid differential analysis approach. The study identifies a range of security risks, including outdated protocols, support for weak ciphers, and problems in certificate validation. These findings are supported by proof-of-concept demonstrations that highlight their potential real-world impact. In addition, the paper analyzes potential OEM motivations behind and provides practical suggestions for improving compliance. The limitations of Google's CDD and CTS frameworks are also discussed. By releasing the dataset and analysis tools, the work promotes reproducibility and facilitates future research. Overall, this is a valuable contribution to the security analysis of customized Android firmware, offering comprehensive empirical insights for improving the security of the Android ecosystem.

Abstract: Controller Area Networks (CANs) are the backbone for reliable intra-vehicular communication. Recent cyberattacks have, however, exposed the weaknesses of CAN, which was designed without any security considerations in the 1980s. Current efforts to retrofit security via intrusion detection or message authentication codes are insufficient to fully secure CAN as they cannot adequately protect against masquerading attacks, where a compromised communication device, a so-called electronic control units, imitates another device. To remedy this situation, multicast source authentication is required to reliably identify the senders of messages. In this paper, we present CAIBA, a novel multicast source authentication scheme specifically designed for communication buses like CAN. CAIBA relies on an authenticator overwriting authentication tags on-the-fly, such that a receiver only reads a valid tag if not only the integrity of a message but also its source can be verified. To integrate CAIBA into CAN, we devise a special message authentication scheme and a reactive bit overwriting mechanism. We achieve interoperability with legacy CAN devices, while protecting receivers implementing the AUTOSAR SecOC standard against masquerading attacks without communication overhead or verification delays.

Meta Review: This paper presents CAIBA, a CAN bus authentication mechanism that mitigates masquerading attacks using an authenticator node. Unlike traditional group key-based MACs, CAIBA ensures source authentication by having each ECU share a unique source key with the authenticator and a group key with other nodes. The sender transmits an XORed MAC, which the authenticator modifies in real time using reactive bit flipping. This enables authentication without modifying receiving nodes, maintaining compatibility with existing CAN systems. Lightweight and real-time, CAIBA was tested on a CAN bus, demonstrating low latency and strong security against compromised ECUs. Reviewers appreciated several aspects of the CAIBA paper. They found it well-written and easy to follow. Additionally, reviewers highlighted the novelty of using Reactive Bit Flipping (RBF) to address CAN masquerading attacks, which provides an innovative approach to source authentication without major modifications to the existing CAN infrastructure. Another strong point was CAIBA’s backward compatibility with CAN 2.0. Reviewers also acknowledged the potential for integration with SecOC, which could further strengthen CAN security in modern vehicle architectures. While reviewers acknowledged the strengths of the CAIBA they also suggested some ways to improve it, including adding a formal security proof and refining claims about SecOC. They also encouraged authors to explore whether simpler MAC-based schemes could provide similar security with lower complexity. While the need for hardware modifications and the 40kbps evaluation were noted, these aspects present opportunities for further development. Reviewers recommended adding synchronization data, clarifying timing, and comparing computational overhead with existing solutions to enhance CAIBA’s impact and applicability.

Abstract: Internet censorship is a well-explored area, typically focusing on Web censorship enacted by powerful nation-states like China and Russia. In this paper, we diverge from the norm and study the unexplored VoIP censorship prevalent in the Middle East for over a decade. We present the first research analyzing the VoIP filtering mechanics deployed nationwide. Based on extensive on-the-ground experiments, our investigation reveals novel censorship techniques for distinctly filtering VoIP traffic originating from specific apps. \r\n\r\nWe meticulously analyze the VoIP traffic of nine popular apps and reveal the presence of sophisticated \"middleboxes\" placed within the local ISPs. These middleboxes possess the unique capability to selectively block the calling facility, leaving the rest of the apps' features, such as texting and media sharing, completely unhindered. Unlike traditional Web filtering techniques that often require inspecting DNS and HTTP(s) traffic, our research demonstrates that the middlebox scrutinizes the STUN protocol requests and response packets to individually identify the application-specific VoIP call flow. \r\nOur analysis of widely used apps such as WhatsApp, Signal, and Facebook Messenger reveals a uniquely characterizable flow of STUN packets. We experimentally confirm that the censor exploits such fingerprints to block VoIP calls made through these apps. \r\nFurthermore, our experiments unveil additional nuances in censorship tactics. For apps like LINE, the middlebox searches for the VoIP server IP address in some selective packets. On a successful match, it drops the packets, effectively disrupting the call.\r\n\r\nIn essence, our findings provide valuable insights into the intricate mechanisms of sophisticated VoIP filtering techniques, \r\npaving the way for more informed approaches to combating such censorship practices.

Meta Review: The paper tackles an important topic and demonstrates a well-presented study, providing useful insights into the status quo of VoIP censorship in two Middle Eastern countries. The paper studied an extensive list of popular apps/services that could be affected by the censorship policies, and examined a reasonable level of technical details in different cases. The research team took personal trips to those understudied countries and conducted the on-the-ground experiments, which is also highly appreciated by the reviewers. Overall, the paper advances the understanding of VoIP censorship and makes impactful observations for the interested community

Abstract: 4G and 5G represent the current cellular communication standards utilized daily by billions of users for various applications. Consequently, ensuring the security of 4G and 5G network implementations is critically important. This paper introduces an automated fuzzing framework designed to test the security of 4G and 5G attach procedure implementations. Our framework provides a comprehensive solution for uplink and downlink fuzzing in 4G, as well as downlink fuzzing in 5G, while supporting fuzzing on all layers except the physical layer. To guide the fuzzing process, we introduce a novel algorithm that assigns probabilities to packet fields and adjusts these probabilities based on coverage information from the device-under-test (DUT). For cases where coverage information from the DUT is unavailable, we propose a novel methodology to estimate it. When evaluating our framework, we first run the random fuzzing experiments, where the mutation probabilities are fixed throughout the fuzzing, and give an insight into how those probabilities should be chosen to optimize the Random fuzzer to achieve the best coverage. Next, we evaluate the efficiency of the proposed coverage-based algorithms by fuzzing open-source 4G stack (srsRAN) instances and show that the fuzzer guided by our algorithm outperforms the optimized Random fuzzer in terms of DUT's code coverage. In addition, we run fuzzing tests on 12 commercial off-the-shelf (COTS) devices. In total, we discovered vulnerabilities in 10 COTS devices and all of the srsRAN 4G instances.

Meta Review: This paper presents CovFUZZ, an over-the-air (OTA) fuzzer for cellular protocols, applied to the 4G and 5G Attach procedures. In contrast to previous, fully black-box OTA testing approaches, CovFUZZ uses coverage information from the device-under-test when available. In cases where this is not possible, CovFuzz utilizes open source cellular stack implementations, such as srsRAN, to approximate reached coverage. The reviewers appreciated this approach to OTA fuzzing of cellular protocols. They positively highlighted the open source-driven approach of CovFuzz, the experimental validation of different fuzzing parameters, and the discovery of multiple vulnerabilities on COTS devices, demonstrating the practical impact of this work. Although the work does not experimentally compare against other OTA fuzzing approaches, the reviewers agreed that this paper provides a valuable step forward for testing the security of cellular protocol implementations.

Abstract: Telecommunication services are essential in ensuring the operation of numerous critical infrastructures. Emerging concepts, such as the Open Radio Access Network (O-RAN), are transforming the operation of Radio Access Networks (RANs) while simultaneously introducing new attack vectors. In this study, we identify that relatively simple Denial of Service (DoS) attacks leveraging subscription mechanisms can compromise O-RAN implementations. We present a novel DoS attack specifically targeting the Near Real-Time (Near-RT) RAN Intelligent Controller (RIC). By deploying a malicious xApp, we demonstrate how an adversary can flood the Near-RT RIC with excessive subscription requests, leading to service disruption. This attack exploits the lack of rate-limiting mechanisms within the Service Model (SM), a critical component of the Near-RT RIC responsible for handling E2 subscription requests. We systematically evaluate various attack scenarios and investigate the underlying vulnerabilities exposed. Furthermore, we propose and assess countermeasures to safeguard publicly accessible O-RAN systems from such threats.

Meta Review: This paper presents the concept for a Denial of Service attack against Open RAN mobile network setups. Through the deployment of a malicious xApp it is possible to flood the Near Real-Time RAN Intelligent Controller of the network that leads to a rate limiting in the network. Due to the increasing adoption of O-RAN, this paper addresses a very relevant and timely security issue that led to the assignment of three individual CVEs. Alongside the attack concept, the paper also proposes an effective countermeasure that can help to protect against the identified DoS threat. The experimental evaluation covers 7 out of 8 major O-RAN open source platforms and produces interesting results for them. Overall, this paper contributes to the security of mobile network implementations and addresses an important security threat for the specific context of O-RAN communication.

Abstract: The emergence of softwarized network devices, like programmable switches and smart NICs, has brought about new and advanced network functionalities. By offloading network functionality from the network control-plane to the programmable data-plane, intelligent decision-making becomes possible at line rate. In this paper, we offload fine-grained Distributed Denial of Service (DDoS) attack detection to the data-plane. The state-of-the-art in this regard mainly aims to embed Machine Learning (ML) models into the data-plane without compromising on inference accuracy. Besides accuracy, we must consider multiple other factors, like traffic feature availability and false positive rates. To that end, we propose O'MINE: One Model is Not Enough, a novel collaborative detection mechanism comprising lightweight ML models. This maximises the detection accuracy while keeping the false positive rate (FPR) low. We use three state-of-the-art datasets to evaluate the O'MINE algorithm and its ML models. Our results show that O'MINE can detect DDoS attacks with high accuracy (~ 94%) and low FPR (~0.2%), outperforming the state-of-the-art. Lastly, O'MINE only consumes a few device resources (~6% of LUT and ~4% of FF) on the Xlinx Alevo U250 FPGA we have used for our inference at line rate.

Meta Review: The reviewers unanimously agreed that the paper presents a compelling solution for DDoS mitigation using programmable data planes. Unlike previous studies that rely on a single model, this work introduces a novel approach leveraging multiple lightweight ML models for collaborative detection. The coordination among multiple programmable switches to jointly reduce false alarms is both innovative and effective. Additionally, the system's implementation on real-world hardware, including FPGA and Intel Tofino chips, further strengthens the paper's contribution.

Abstract: Irrespective of the security and isolation guarantees offered by the mobile operating system, the Mobile Application Security Verification Standard (MASVS) recommends app developers to implement hardening techniques for self-protection—to prevent tampering and leakage, detect jailbreaks, etc. Despite increasing regulation to incentivize developers toward implementing self-protection, our understanding of the use of hardening techniques on different mobile platforms is very limited—especially regarding differences, if any, between the two main mobile ecosystems. In this paper, we systematize knowledge on the use and analysis of hardening techniques, covering, for the first time, both Android and iOS apps. To this end, we present HALY, a framework to analyze the adoption of hardening techniques. Using HALY’s static and dynamic analysis, we analyze 2,646 popular apps available on both Android and iOS, and measure the prevalence of hardening techniques. Contrary to expectation, apps on iOS underperform in self-protection, implementing only half of the recommended hardening techniques compared to their Android counterparts—challenging the long-held belief that iOS is simply “more secure.” Equally surprising, while privacy-sensitive apps implement more self-protection, many apps implement hardening techniques on only one of the two OSes. Furthermore, as many common techniques are easy to individually bypass, the additional security is questionable. Overall, almost all apps implement some hardening techniques, but as many as 24.1% (Android) and 85.0% (iOS) implement fewer than half of the recommended ones, and only 4.7% of apps on Android and 0.2% on iOS adopt all.

Meta Review: The paper "SoK: Hardening Techniques in the Mobile Ecosystem - Are We There Yet?" introduces the HALY framework to analyze different hardening techniques and mitigations used in mobile ecosystems such as Android and iOS. A study on malware samples complements a somewhat limited study of a few thousand apps (of millions of existing apps). Open source apps served as ground truth for the exploration and provide a solid baseline. As several other frameworks already exist, HALY finds an interesting niche in exploring and estimating the hardening features to provide a great overview of the field. In summary, the reviewers enjoyed the broad scope of HALY and the comparison to other existing work, providing a great overview and systematization of the field.

Abstract: To combat eavesdropping and injection attacks, wireless networks widely adopt encryption to provide confidentiality and integrity guarantees. In this paper, we present a novel and generic attack, termed LenOracle, which can hijack the TCP/UDP connections over encrypted wireless networks (e.g., 5G/4G/3G and Wi-Fi) via packet injections from the Internet. Due to the design nature of wireless networks and stream ciphers they used, the length of IP packets being transmitted can be acquired by radio sniffing. It thus provides a side channel for adversaries. We found that adversaries could utilize this side channel with TCP features to infer the presence of a connection, infer the protocol state (sequence number, acknowledge number) of the connection, and finally hijack TCP/IP connections over wireless networks. Through real-world experiments in commercial LTE networks and real Wi-Fi networks, we demonstrated that the LenOracle attack is practical and severe against both TCP and UDP connections. For the former, we successfully injected a fake short message into a victim TCP connection; For the latter, we were able to inject a fake DNS response into a UDP connection and poisoned the DNS cache of the victim device. Following the responsible disclosure policy, we have reported our findings and mitigation recommendations to GSMA and Wi-Fi Alliance. The GSMA acknowledged that the issue affects 5G/4G/3G, notified all its members (operators and vendors worldwide) of this issue, and highlighted the mitigation we proposed.

Meta Review: The paper "The Danger of Packet Length Leakage: Off-path TCP/IP Hijacking Attacks Against Wireless Networks" presents a clever attack that crosses multiple abstractions and has the potential for real impact. The work leverages a combination of low-level cellar/wifi radio behaviors and IP-level probing to compromise a variety of protocols such as DNS and RCS messages in a number of specific scenarios. The reviewers agreed on the novelty and importance of the work, and appreciated the real-world exploration, clarity of the paper, and the responsible disclosure process executed by the authors.

Abstract: In an anonymous credential system, users collect credentials from issuers, and can use their credentials to generate privacy-preserving identity proofs that can be shown to third-party verifiers. Since the introduction of anonymous credentials by Chaum in 1985, there has been promising advances with respect to system design, security analysis and real-world implementations of anonymous credential systems. In this paper, we examine Hyperledger AnonCreds, an anonymous credential system that was introduced in 2017 and is currently undergoing specification. Despite being implemented in deployment-ready identity system platforms, there is no formal security analysis of the Hyperledger AnonCreds protocol. We rectify this, presenting syntax and a security model for, and a first security analysis of, the Hyperledger AnonCreds protocol. In particular, we demonstrate that Hyperledger AnonCreds is correct, and satisfies notions of unforgeability and anonymity. We conclude with a discussion on the implications of our findings, highlighting the importance of rigorous specification efforts to support security evaluation of real-world cryptographic protocols.

Meta Review: The paper provides an analysis of the anonymous credential system Hyperledger AnonCreds that is currently being standarised. For this, the paper provides a clean formal presentation of the protocol, including revocation for which no formal write-up exists. It defines the syntax and security model for any AnonCreds system. Additionally, it describes the HLAC protocol in detail, breaking it down into its building blocks and the three core algorithms: setup, issuance, and credential presentation (proving and verification).

Abstract: In this paper, we introduce an Oracle version of the Restricted Syndrome Decoding Problem (RSDP) and propose novel authentication protocols based on the hardness of this problem. They follow the basic structure of the HB-family of authentication protocols and later improvements but demonstrate several advantages. An appropriate choice of multiplicative subgroup and ring structure gives rise to a very efficient implementation with up to 50\% lower hardware costs compared to other \emph{Learning Parity with Error} based approaches. In addition, the new protocols also have lower key size, lower communication costs, and potentially better completeness/soundness compared to learning-based alternatives. This is appealing in the context of low-cost, low-powered authenticating devices such as radio frequency identification (RFID) systems. Lastly, we show that with additional assumptions, RSDP can be naturally adapted to Lyubashevski et al. framework of Man in the Middle secured authentication protocol.

Meta Review: This paper proposes new light-weight authentication protocols based on the Restricted Syndrome Decoding Problem. The new protocols follow the general structure of the previously proposed and well-studied HB protocol family based on the Learning Parity with Noise problem. The new protocols clearly outperform these as they come with a lower key size and lower communication cost at the same security level. This make them an interesting choice for tag authentication in the context of RFID systems and any other contexts in which low-cost and low-powered devices are to be authenticated.

Abstract: Authenticated encryption (AE) is a cryptographic mechanism that allows communicating parties to protect the confidentiality and integrity of messages exchanged over a public channel, provided they share a secret key. In this work, we present new AE schemes leveraging the SHA-3 standard functions SHAKE128 and SHAKE256, offering 128 and 256 bits of security strength, respectively, and their “Turbo” counterparts. They support session-based communication, where a ciphertext authenticates the sequence of messages since the start of the session. The chaining in the session allows decryption in segments, avoiding the need to buffer the entire deciphered cryptogram between decryption and validation. And, thanks to the collision resistance of (Turbo)SHAKE, they provide so-called CMT-4 committing security, meaning that they provide strong guarantees that a ciphertext uniquely binds to the key, plaintext and associated data. The AE schemes we propose have the unique combination of advantages that 1) their security is based on the security claim of SHAKE, that has received a large amount of public scrutiny, that 2) they make use of the standard Keccak-p permutation that not only receives more and more dedicated hardware support, but also allows competitive software-only implementations thanks to the TurboSHAKE instances, and that 3) they do not suffer from a 64-bit birthday bound like most AES-based schemes.

Meta Review: This paper describes an authenticated encryption scheme built using the SHA-3 standard functions SHAKE128 and SHAKE256. This new algorithm is comparable to existing algorithms, with several advantages. The reviewers felt that the paper was well written. It clearly described the motivation for the research, the encryption scheme's design, and its evaluation. The one concern from reviewers was that the new scheme's performance was only compared against a couple of competing algorithms and only on a RISC architecture. In the rebuttal, the authors promised to add this comparison.

Abstract: Deepfakes have rapidly emerged as a serious threat to society due to their ease of creation and dissemination, triggering the accelerated development of detection technologies. However, many existing detectors rely on lab-generated datasets for validation, which may not prepare them for novel, real-world deepfakes. This paper extensively reviews and analyzes state-of-the-art deepfake detectors, evaluating them against several critical criteria. These criteria categorize detectors into 4 high-level groups and 13 fine-grained sub-groups, aligned with a unified conceptual framework we propose. This classification offers practical insights into the factors affecting detector efficacy. We evaluate the generalizability of 16 leading detectors across comprehensive attack scenarios, including black-box, white-box, and gray-box settings. Our systematized analysis and experiments provide a deeper understanding of deepfake detectors and their generalizability, paving the way for future research and the development of more proactive defenses against deepfakes.

Meta Review: This SoK work presents a review of deepfake detection methods. It evaluates the generalizability and performance of the detectors under different settings. The reviewers appreciate the comprehensive evaluation based on conceptual framework, especially the analysis of 51 deepfake detectors and evaluations of 16 detectors across different attack scenarios. Multiple reviewers noted improvements are needed and these concerns are acknowledged by the authors in their rebuttals: (1) lack of clear explanations of t-SNE visualizations and further analysis of the experimental results; (2) need detailed description on evolution settings (white-box, gray-box, black-box); (3) the taxonomy and the selection criteria for analysis, especially that it overlooks studies in other orthogonal approaches in fighting deepfake and the benchmarks of deepfake detectors.

Abstract: In recent years, vehicles have evolved into cyber-physical autonomous systems that rely on sensor data from various sources within the vehicle. With the emergence of Vehicle-to-Everything (V2X) technology, the scope of the collaborative functionality in vehicles is now expanding to the inter-vehicular level. To support these modern capabilities, the complexity of the Electronic Control Units (ECUs) and the In-Vehicle Network (IVN) architecture is rapidly increasing. As a result, IVNs are now swarms of devices that communicate safety-critical data. Unfortunately, current vehicular networks lack security, opening the path to numerous cyberattacks. A typical solution for verifying the integrity of multiple devices is swarm attestation. However, in a typical IVN setting, only the Original Equipment Manufacturer (OEM) has access to the legitimate configuration of the ECUs and does not want to disclose this information due to intellectual property and security concerns. Therefore, state-of-the-art swarm attestation schemes, which do not provide privacy guarantees, are unsuitable for IVNs. This paper proposes Secure Privacy-Preserving Anonymous Swarm Attestation for In-Vehicle Networks (SPARK), which builds upon a novel group signature scheme to enable privacy-preserving, anonymous, and traceable swarm attestation of IVNs. We validate SPARK through a proof-of-concept implementation using a standardized hardware TPM 2.0 and representative hardware platforms. The results demonstrate the real-world applicability of SPARK.

Meta Review: The paper introduces SPARK, a privacy-preserving swarm attestation protocol designed for zonal in-vehicle networks (IVNs). By integrating Direct Anonymous Attestation with attributes (DAA-A) and TPM 2.0, SPARK enables integrity verification while ensuring anonymity and traceability—features that have not been jointly addressed in prior work. The zero-knowledge proof mechanism is well-suited to the automotive setting and may also apply to broader IoT environments. The paper is well-structured, presenting clearly defined security goals, formal guarantees, and a practical implementation evaluated under realistic hardware and network conditions.

Abstract: Textual descriptions in cyber threat intelligence (CTI) reports, such as security articles and news, are rich sources of knowledge about cyber threats, crucial for organizations to stay informed about the rapidly evolving threat landscape. However, current CTI extraction methods lack flexibility and generalizability, often resulting in inaccurate and incomplete knowledge extraction. Syntax parsing relies on fixed rules and dictionaries, while model fine-tuning requires large annotated datasets, making both paradigms challenging to adapt to new threats and ontologies. To bridge the gap, we propose CTINexus, a novel framework leveraging optimized in-context learning (ICL) of large language models (LLMs) for data-efficient CTI knowledge extraction and high-quality cybersecurity knowledge graph (CSKG) construction. Unlike existing methods, CTINexus requires neither extensive data nor parameter tuning and can adapt to various ontologies with minimal annotated examples. This is achieved through: (1) a carefully designed automatic prompt construction strategy with optimal demonstration retrieval for extracting a wide range of cybersecurity entities and relations; (2) a hierarchical entity alignment technique that canonicalizes the extracted knowledge and removes redundancy; (3) an ICL-enhanced long-distance relation prediction technique to further complete the CKSG with missing links. Our extensive evaluations using 150 real-world CTI reports collected from 10 platforms demonstrate that CTINexus significantly outperforms existing methods in constructing accurate and complete CSKGs, highlighting its potential to transform CTI analysis with an efficient and adaptable solution for the dynamic threat landscape.

Meta Review: This paper proposes an advanced LLM-based approach for extracting entities and relations from Cyber Threat Intelligence (CTI) reports, a timely and active research area. The paper is well-written, technically sound, and offers significant improvements over the state of the art. While reviewers initially expressed some concerns regarding missing evaluation details and other presentation issues, the authors' response adequately addressed their concerns. Following the discussion phase, all reviewers expressed support for the paper and recommended incorporating the new details provided by the authors into the final version.

Abstract: Machine unlearning (MU) aims to remove the influence of particular data points from the learnable parameters of a trained machine learning model. This is a crucial capability in light of data privacy requirements, trustworthiness, and safety in deployed models. MU is particularly challenging for deep neural networks (DNNs), such as convolutional nets or vision transformers, as such DNNs tend to memorize a notable portion of their training dataset. Nevertheless, the community lacks a rigorous and multifaceted study that looks into the success of MU methods for DNNs. In this paper, we investigate 18 state-of-the-art MU methods across various benchmark datasets and models, with each evaluation conducted over 10 different initializations, a comprehensive evaluation involving MU over 100K models. We show that, with the proper hyperparameters, Masked Small Gradients (MSG) and Convolution Transpose (CT), consistently perform better in terms of model accuracy and run-time efficiency across different models, datasets, and initializations, assessed by population-based membership inference attacks (MIA) and per-sample unlearning likelihood ratio attacks (U-LiRA). Furthermore, our benchmark highlights the fact that comparing a MU method only with commonly used baselines, such as Gradient Ascent (GA) or Successive Random Relabeling (SRL), is inadequate, and we need better baselines like Negative Gradient Plus (NG+) with proper hyperparameter selection.

Meta Review: The paper presents machine unlearning methods in a well-structured and thorough analysis way. The paper has strong empirical evaluations and ML architectures supporting the findings across multiple datasets. The reviewers, particularly, appreciate the methodological rigor. While that paper was well-received, the reviewers identified areas where clarity and presentation could be improved, such as enhancing visualizations of performance differences, improving transparency of results by linking conclusions more explicitly to data points, and checking some aspects of grammar and formatting. These limitations, however, do not undermine the significance of the work, and there was consensus among the reviewers that the contributions merit an acceptance.

Abstract: Industrial Control Systems (ICS) are extensively used in critical infrastructures ensuring efficient, reliable, and continuous operations. However, their increasing connectivity and addition of advanced features make them vulnerable to cyber threats, potentially leading to severe disruptions in essential services. In this context, honeypots play a vital role by acting as decoy targets within ICS networks, or on the Internet, helping to detect, log, analyze, and develop mitigations for ICS-specific cyber threats. Deploying ICS honeypots, however, is challenging due to the necessity of accurately replicating industrial protocols and device characteristics, a crucial requirement for effectively mimicking the unique operational behavior of different industrial systems. Additionally, the difficulty is increased by the substantial manual effort involved in replicating the PLC's control logic. This is necessary to capture attacker traffic that seeks to interfere with critical infrastructure operations. In this paper, we propose LLMPot, a novel approach for designing honeypots in ICS networks harnessing the potency of Large Language Models (LLMs). LLMPot aims to provide a dynamic framework that can be used to optimize the creation of realistic honeypots with vendor-agnostic configurations and for various control logic, aiming to eliminate the manual effort and specialized knowledge traditionally required by existing strategies. We conducted extensive experiments focusing on a wide array of parameters, demonstrating that LLMPot can effectively create honeypot devices implementing different industrial protocols, PLC configurations, and diverse control logic.

Meta Review: This paper introduces LLMPot, an LLM-based honeypot designed for ICS, representing a novel and compelling application of LLMs. The paper provides a comprehensive evaluation addressing three key research questions, and provides simulation testing in a testbed environment, demonstrating the approach’s practicality. Furthermore, the open-sourcing of LLMPot enhances the work’s reproducibility and value to the research community. Overall, this study lays a good foundation for future advancements in the field.

Abstract: Binary function similarity, which often relies on learning-based algorithms to identify what functions in a pool are most similar to a given query function, is a sought-after topic in different communities, including machine learning, software engineering, and security. Its importance stems from the impact it has in facilitating several crucial tasks, from reverse engineering and malware analysis to automated vulnerability detection. Whereas recent work cast light around performance on this long-studied problem, the research landscape remains largely lackluster in understanding the resiliency of the state-of-the-art machine learning models against adversarial attacks. As security requires to reason about adversaries, in this work we assess the robustness of such models through a simple yet effective black-box greedy attack, which modifies the topology and the content of the control flow of the attacked functions. We demonstrate that this attack is successful in compromising all the models, achieving average attack success rates of 57.06\% and 95.81\% depending on the problem settings (targeted and untargeted attacks). Our findings are insightful: top performance on clean data does not necessarily relate to top robustness properties, which explicitly highlights performance-robustness trade-offs one should consider when deploying such models, calling for further research.

Meta Review: Function similarity has been a very hot topic in both security and software engineering for a few years now, with important applications in reverse and vulnerability analysis. Yet, evasion against such methods is hardly studied. The present paper fills this gap with a systematic study over a strong experimental set up. The authors consider 8 representative techniques, 2 attack scenarios, 4 transformations, several variation of parameters and a wide range of meaningful research questions, including transferability and common mode analysis. This work will for sure provide a strong ground for future progress in the field.

Abstract: Protecting intellectual property (IP) in federated learning (FL) is increasingly important as clients contribute proprietary data to collaboratively train models. Model watermarking, particularly through backdoor-based methods, has emerged as a popular approach for verifying ownership and contributions in deep neural networks trained via FL. By manipulating their datasets, clients can embed a secret pattern, resulting in non-intuitive predictions that serve as proof of participation, useful for claiming incentives or IP co-ownership. However, this technique faces practical challenges: client watermarks can collide, leading to ambiguous ownership claims, and malicious clients may exploit watermarks to inject harmful backdoors, jeopardizing model integrity. To address these issues, we propose Sanitizer, a server-side method that ensures client-embedded backdoors cannot be triggered on natural queries in harmful ways. It identifies subnets within client-submitted models, extracts backdoors throughout the FL process, and confines them to harmless, client-specific input subspaces. This approach not only enhances Sanitizer's efficiency but also resolves conflicts when clients use similar triggers with different target labels. Our empirical results demonstrate that Sanitizer achieves near-perfect success in verifying client contributions while mitigating the risks of malicious watermark use. Additionally, it reduces GPU memory consumption by 85% and cuts processing time by at least 5 times compared to the baseline.

Meta Review: The paper "Unharmful Backdoor-based Client-side Watermarking in Federated Learning" presents Sanitizer, a novel server-side method that mitigates harmful backdoor attacks while verifying client contributions in federated learning. Unlike previous approaches, Sanitizer effectively identifies and isolates client-embedded backdoors, ensuring they cannot be triggered in harmful ways by natural queries. The method is designed to handle conflicts between clients using similar watermarks and target labels, providing a scalable solution to IP verification challenges in federated learning. The paper includes comprehensive experimental evaluations demonstrating its effectiveness in reducing memory usage, and achieving near-perfect success in verifying client contributions.

Abstract: Flaws in software security, including bugs and vulnerabilities, can cause serious issues, and fuzzing has been extensively studied and employed to identify them. A typical fuzzer automatically generates test inputs and feeds back information obtained from executed programs for efficient software verification. As the feedback information, edge coverage, which is based on edge traversals in a control flow graph, is widely used. We focused on branches whose results change depending on the traversal of a particular edge in the control flow graph. In particular, we found that when a branch depends on a combination of multiple edge traversals, it is difficult for existing edge coverage to cover the branch outcome. Based on this observation, we propose a novel coverage metric for such branches: auxiliary composite-edge coverage (AceCov). AceCov works together with existing edge coverage to provide an auxiliary coverage metric only for branches such as the one described above, where the edge coverage does not cover the result. We implemented AceCov in AFL++ and evaluated it using FuzzBench and MAGMA. The evaluation results show a performance improvement of up to 3.36% and an average of 1.19% over AFL++. We also found four bugs that were not found in AFL++.

Meta Review: The authors present AceCov, a coverage metric that complements traditional edge coverage metrics with composite-edge dependencies to help fuzzers explore previously difficult-to-reach edges. The reviewers value the simple and decoupled design of AceCov and the author's commitment to releasing AceCov as an open-source tool. In particular, the reviewers appreciate the integration of AceCov in the widely used fuzzing tool AFL++. AceCov shows consistent improvement versus existing state-of-the-art coverage metrics. Even though these improvements are considered small, the reviewers agree that small improvements are still valuable in a research space as saturated as fuzzing. The reviewers appreciate the thoroughness and extensiveness of the experimental evaluation. Some concerns were raised regarding the subsetting of FuzzBench during the experiments and how this could introduce bias. After deliberation and considering author responses, the reviewers agreed that the current methodology is representative enough and extending it would not significantly improve the paper.

Abstract: FIDO2 is the standard technology for single-factor and second-factor authentication. It is specified in an open standard including the WebAuthn and CTAP application layer protocols. We focus on CTAP which allows the communication between FIDO2 clients and hardware authenticators. No prior work explored the CTAP Authenticator API which is a critical protocol-level attack surface as it deals with credential creation, deletion, and management. We address this gap by presenting the first security and privacy evaluation of the CTAP Authenticator API. We uncover two classes of protocol-level attacks on CTAP we call CTRAPS. The client impersonation (CI) attacks exploit the lack of client authentication to tamper with FIDO2 authenticators. They include zero-click attacks capable of deleting FIDO2 credentials, including passkeys, without user interaction. The API confusion (AC) attacks abuse the lack of protocol API enforcements and confound FIDO2 authenticators, clients, and unaware users into calling unwanted CTAP APIs while thinking they are calling legitimate ones. For example, a victim thinks is authenticating to a website they are deleting their credentials. The presented eleven attacks are conducted either in proximity or remotely and are effective regardless of the underlying CTAP transport (USB, NFC, or BLE). We detail the eight vulnerabilities in the CTAP specification enabling the CTRAPS attacks. Six of them are novel and include unauthenticated CTAP clients and trackable FIDO2 credentials. We release CTRAPS, an original toolkit to analyze CTAP and conduct the CTRAPS attacks. We confirm the attacks’ practicality on a large scale by exploiting six popular authenticators, including a FIPS-certified one, from Yubico, Feitian, SoloKeys, and Google, and ten widely used relying parties, such as Microsoft, Apple, GitHub, and Facebook. We present eight practical and backward-compliant countermeasures to fix the attacks and their root causes. We responsibly disclosed our findings to the FIDO alliance and the affected vendors.

Meta Review: The paper presents a security and privacy evaluation of the CTAP application layer protocol, which is an important component of the FIDO2 authentication standard. Numerous prior works have studied different versions of FIDO and its WebAuthn component, but the CTAP protocol has received comparatively less scrutiny. The paper makes a valuable contribution by uncovering four Client Impersonation (CI) attacks which take advantage of flaws in client authentication and seven API Confusion (AC) attacks which highlight the need for protocol API enforcements. While analyzing the root causes of the attacks, the paper presents eight vulnerabilities in the CTAP specification. The paper evaluates its attacks by developing a toolkit, CTRAPS, which it uses to implement its attacks on and successfully exploit six real-world authenticators. The authors disclosed all their findings to the responsible vendors and the FIDO Alliance. The reviewers agree that the paper enhances our understanding of the security of the CTAP protocol and shows the practicality of exploits beyond the purview of the reference threat model adopted by the FIDO Alliance.

Abstract: After several decades of defensive research against the exploitation of memory errors, a wide range of techniques has been proposed, yet no silver bullet has been found. Multi-Variant eXecution (MVX) is one particularly promising proposal for defending against a wide range of known and potentially unknown attacks. MVX systems run multiple program variants in parallel while providing them with identical inputs and monitoring their behavior. By constructing these program variants using automated software diversity techniques, we can ensure that the variants behave identically under normal operating conditions but diverge when attacked. The MVX system detects these divergences and reacts appropriately. However, running multiple program variants in parallel puts a significant load on any system, inhibiting the broader adoption of MVX technology. We propose Partial Multi-Variant eXecution (PMVX), a novel technique that limits the scope of MVX to security-sensitive parts of a target application and that runs the rest of the application in Single-Variant eXecution (SVX) mode. Judicious compartmentalization of the application into an MVX and SVX partition should yield substantially reduced resource consumption compared to traditional MVX systems while maintaining these systems’ strong security guarantees. However, PMVX implementations must address the non-trivial problem of ensuring all variants are in equivalent states whenever they switch from SVX to MVX mode. We designed and implemented a proof-of-concept PMVX system called fortivide that solves this state-equivalency problem using state migration and resynchronization. We thoroughly evaluated the security and performance of our system as a whole, and of our state migration and synchronization mechanisms in isolation. We conclude that PMVX has great potential but needs to be applied with the utmost care since the added overhead of state resynchronization can quickly outweigh the benefits of running in SVX mode.

Meta Review: The paper advances the state of the art in multi-variant execution (MVX) systems by introducing a technique that selectively switches to MVX only for critical parts of a protected program, and continues in singe-variant mode for the rest. This significantly reduces the overall runtime overhead, which is an important step toward facilitating the practical applicability of MVX. The proposed technique for selective execution switching and state migration between execution modes was recognized as a substantial technical advancement. The reviewers unanimously appreciated the thorough and convincing performance evaluation, which includes detailed microbenchmarks and practical assessments using web server applications. Additionally, the reviewers valued the clarity and quality of the writing, the solid technical implementation (with source code availability), and the thoughtful design decisions.

Abstract: Since its inception, Rowhammer exploits have rapidly evolved into increasingly sophisticated threats compromising data integrity and the control flow integrity of victim processes. Nevertheless, it remains a challenge for an attacker to identify vulnerable targets (i.e., Rowhammer gadgets), understand the outcome of the attempted fault, and formulate an attack that yields useful results. In this paper, we present a new type of Rowhammer gadget, called a LeapFrog gadget, which, when present in the victim code, allows an adversary to subvert code execution to bypass a critical piece of code (e.g., authentication check logic, encryption rounds, padding in security protocols). The LeapFrog gadget manifests when the victim code stores the Program Counter (PC) value in the user or kernel stack (e.g., a return address during a function call) which, when tampered with, repositions the return address to a location that bypasses a security-critical code pattern. This research also presents a systematic process to identify LeapFrog gadgets. This methodology enables the automated detection of susceptible targets and the determination of optimal attack parameters. We first show the attack on a decision tree algorithm to show the potential implications. Secondly, we employ the attack on OpenSSL to bypass the encryption and reveal the plaintext. We then use our tools to scan the Open Quantum Safe library and report on the number of LeapFrog gadgets in the code. Lastly, we demonstrate this new attack vector through a practical demonstration in a client/server TLS handshake scenario, successfully inducing an instruction skip in a client application. Our findings extend the impact of Rowhammer attacks on control flow and contribute to developing more robust defenses against these increasingly sophisticated threats.

Meta Review: The paper presents a practical Rowhammer attack targeting values stored on the stack (return addresses). While the concept itself is not entirely novel, the paper introduces a noteworthy contribution that all reviewers identified as particularly compelling: the design and implementation of an automated tool capable of identifying suitable gadgets for the attack. This tool is intended to be released as open source.

Abstract: Transparency logs are critical for a wide range of applications, from web certificates to end-to-end encrypted messaging to code binary distribution. Today, many transparency log designs exist for various applications and workloads, and developers must fully understand the design space to find the best design for their need. To make matters worse, if a developer needs a transparency log for an application and workload without an existing transparency log, the developer (who might not have expertise) must design a new log. To address these challenges, we introduce the paradigm of a configurable transparency log, which takes as input a description of the application workload and constraints of different entities and automatically outputs a transparency log uniquely suited to the application. We present the first configurable transparency log design, LegoLog. We implement LegoLog and evaluate it end-to-end for two different applications. To see how LegoLog generalizes to different applications, we examine six different applications and compare the asymptotic complexity of LegoLog to existing transparency logs tailored to individual applications. We find that configurability does not come at the cost of performance: LegoLog can capture a variety of applications while performing comparably to existing, special-purpose transparency logs.

Meta Review: This paper helps fill an important gap between theory and practice. In particular, LegoLog makes implementing transparency logs and configuring them to specific workloads easier for practitioners. We found LegoLog's modular design to be practical; the authors showed that their design choices can even help improve security in some applications. The paper introduces a lot of innovative ideas, including the decoupling of update frequency from monitoring and the support for offline clients through cheap batch monitoring checks. Moreover, the authors demonstrate LegoLog's real-world deployability through extensive experiments that were further strengthened through the shepherding process. Finally, the paper is well written, easy to follow, and contains many useful examples.

Abstract: With the popularity of low-end microcontrollers, manufacturers and researchers have proposed specialized memory isolation mechanisms. However, current proposals face two main shortcomings. First, due to a lack of extensible reference implementations of commercial specifications, academic systems commonly use custom memory isolation mechanisms, reducing compatibility and the chance of real-world adoption. Second, recent research has demonstrated crucial and overlapping vulnerabilities, including in commercial systems. Unfortunately, efforts to mitigate and validate these issues are hindered by the disconnect in codebases. This paper proposes openIPE, an open research platform for extensible, industry-compliant hardware-software co-designs based on Texas Instruments' specification for Intellectual Property Encapsulation (IPE). openIPE introduces minimal hardware extensions for memory isolation, alongside a versatile firmware layer enabling rapid prototyping of advanced security extensions. We establish a robust security testing infrastructure and demonstrate the capabilities of our framework through a comprehensive study on secure interrupt handling, an active research area in this domain. Our evaluation shows that openIPE not only allows for the independent reproduction and comparison of prior proposals, but also enables a novel solution that achieves strong architectural and microarchitectural security with minimal hardware modifications and overhead.

Meta Review: This paper presents OpenIPE, an open research platform following Texas Instruments’ specification for Intellectual Property Encapsulation (IPE) built on top the openMSP430 core. All reviewers were positive about this paper. They appreciated the open nature of the platform and highlighted that this work will likely be a basis for further research. The flexibility of the framework, allowing both reproducing prior work and prototyping new designs outweighed concerns about novelty over existing work.

Abstract: Among the behaviors and tactics that malware can exhibit, environment-sensitive logic likely poses the longest-standing challenge to the analysis capabilities of automatic systems such as sandboxes. Current analysis approaches either fall short in anticipating adversarial tactics by design, or incur prohibitive costs and other roadblocks when reasoning on real-world code. As a result, manual analysis remains the primary way to identify behaviors that show only when a machine meets specific expectations of the sample. To address these issues, we present the first practical, sound, and effective solution for multi-path exploration of environment-sensitive malware. We argue how the popular coverage-guided fuzzing paradigm from software testing can effectively achieve this task, provided we can devise original design solutions (such as coverage feedback and environment mutations) tailored to the unique characteristics of malware to enable this application. Our approach not only can disarm many evasions without requiring expert knowledge, but also unveil additional activities that would not show in a baseline run due to environmental conditions unrelated to evasion. We build a manually annotated dataset of environment-sensitive malware and use it to estimate the analysis capabilities of the approach. Our Pfuzzer implementation reveals activity that the best competitor misses for 35.98\% of the samples: such activity either follows evasions that deceive existing systems or comes from behaviors that show only in other ``right'' environments. Pfuzzer also unveils dormant evasive tactics for 70.84\% of the samples that one may wrongly deem as non-evasive after a baseline run.

Meta Review: The reviewers unanimously agreed that the paper presents a compelling study of environment-sensitive malware detection, offering useful insights and practical solutions into the community. The research stands out due to its clever use of fuzzing, which adapts original fuzzer components (such as coverage feedback and environment mutations) tailored to the unique characteristics of malware to enable this application. Overall, the paper enhances our understanding of environment-sensitive malware and provides tools for the community to mitigate them.