Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Cryptography and Security

cs.CR Watch Cryptography and Security

Tuesday 24 February

Cryptography and Security

A Secure and Interoperable Architecture for Electronic Health Record Access Control and Sharing

Tayeb Kenaza, Islam Debicha, Youcef Fares et al. (+2 more)

Electronic Health Records (EHRs) store sensitive patient information, necessitating stringent access control and sharing mechanisms to uphold data security and comply with privacy regulations such as the General Data Protection Regulation (GDPR). In this paper, we propose a comprehensive architecture with a suite of efficient protocols that leverage the synergistic capabilities of the Blockchain and Interplanetary File System (IPFS) technologies to enable secure access control and sharing of EHRs. Our approach is based on a private blockchain, wherein smart contracts are deployed to enforce control exclusively by patients. By granting patients exclusive control over their EHRs, our solution ensures compliance with personal data protection laws and empowers individuals to manage their health information autonomously. Notably, our proposed architecture seamlessly integrates with existing health provider information systems, facilitating interoperability and effectively addressing security and data heterogeneity challenges. To demonstrate the effectiveness of our approach, we developed a prototype based on a private implementation of the Hyperledger platform, enabling the simulation of diverse scenarios involving access control and health data sharing among healthcare practitioners. Our experimental results demonstrate the scalability of our solution, thereby substantiating its efficacy and robustness in real-world healthcare settings.
Read the full abstract

All authors Tayeb Kenaza, Islam Debicha, Youcef Fares, Mehdi Sehaki, Sami Messai

Cryptography and Security

AdapTools: Adaptive Tool-based Indirect Prompt Injection Attacks on Agentic LLMs

Che Wang, Jiaming Zhang, Ziqi Zhang et al. (+6 more)

The integration of external data services (e.g., Model Context Protocol, MCP) has made large language model-based agents increasingly powerful for complex task execution. However, this advancement introduces critical security vulnerabilities, particularly indirect prompt injection (IPI) attacks. Existing attack methods are limited by their reliance on static patterns and evaluation on simple language models, failing to address the fast-evolving nature of modern AI agents. We introduce AdapTools, a novel adaptive IPI attack framework that selects stealthier attack tools and generates adaptive attack prompts to create a rigorous security evaluation environment. Our approach comprises two key components: (1) Adaptive Attack Strategy Construction, which develops transferable adversarial strategies for prompt optimization, and (2) Attack Enhancement, which identifies stealthy tools capable of circumventing task-relevance defenses. Comprehensive experimental evaluation shows that AdapTools achieves a 2.13 times improvement in attack success rate while degrading system utility by a factor of 1.78. Notably, the framework maintains its effectiveness even against state-of-the-art defense mechanisms. Our method advances the understanding of IPI attacks and provides a useful reference for future research.
Read the full abstract

All authors Che Wang, Jiaming Zhang, Ziqi Zhang, Zijie Wang, Yinghui Wang, Jianbo Gao, Tao Wei, Zhong Chen, Wei Yang Bryan Lim

Cryptography and Security · Cross-listed from cs.SE

PackMonitor: Enabling Zero Package Hallucinations Through Decoding-Time Monitoring

Xiting Liu, Yuetong Liu, Yitong Zhang et al. (+2 more)

As Large Language Models (LLMs) are increasingly integrated into software development workflows, their trustworthiness has become a critical concern. However, in dependency recommendation scenarios, the reliability of LLMs is undermined by widespread package hallucinations, where models often recommend hallucinated packages. Recent studies have proposed a range of approaches to mitigate this issue. Nevertheless, existing approaches typically merely reduce hallucination rates rather than eliminate them, leaving persistent software security risks. In this work, we argue that package hallucinations are theoretically preventable based on the key insight that package validity is decidable through finite and enumerable authoritative package lists. Building on this, we propose PackMonitor, the first approach capable of fundamentally eliminating package hallucinations by continuously monitoring the model's decoding process and intervening when necessary. To implement this in practice, PackMonitor addresses three key challenges: (1) determining when to trigger intervention via a Context-Aware Parser that continuously monitors model outputs and selectively activates intervening only during installation command generation; (2) resolving how to intervene by employing a Package-Name Intervenor that strictly limits the decoding space to an authoritative package list; and (3) ensuring monitoring efficiency through a DFA-Caching Mechanism that enables scalability to millions of packages with negligible overhead. Extensive experiments on five widely used LLMs demonstrate that PackMonitor is a training-free, plug-and-play solution that consistently reduces package hallucination rates to zero while maintaining low-latency inference and preserving original model capabilities.
Read the full abstract

All authors Xiting Liu, Yuetong Liu, Yitong Zhang, Jia Li, Shi-Min Hu

Cryptography and Security · Cross-listed from cs.AI

ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time Correction

Che Wang, Fuyao Zhang, Jiaming Zhang et al. (+6 more)

Large Language Model (LLM) agents are susceptible to Indirect Prompt Injection (IPI) attacks, where malicious instructions in retrieved content hijack the agent's execution. Existing defenses typically rely on strict filtering or refusal mechanisms, which suffer from a critical limitation: over-refusal, prematurely terminating valid agentic workflows. We propose ICON, a probing-to-mitigation framework that neutralizes attacks while preserving task continuity. Our key insight is that IPI attacks leave distinct over-focusing signatures in the latent space. We introduce a Latent Space Trace Prober to detect attacks based on high intensity scores. Subsequently, a Mitigating Rectifier performs surgical attention steering that selectively manipulate adversarial query key dependencies while amplifying task relevant elements to restore the LLM's functional trajectory. Extensive evaluations on multiple backbones show that ICON achieves a competitive 0.4% ASR, matching commercial grade detectors, while yielding a over 50% task utility gain. Furthermore, ICON demonstrates robust Out of Distribution(OOD) generalization and extends effectively to multi-modal agents, establishing a superior balance between security and efficiency.
Read the full abstract

All authors Che Wang, Fuyao Zhang, Jiaming Zhang, Ziqi Zhang, Yinghui Wang, Longtao Huang, Jianbo Gao, Zhong Chen, Wei Yang Bryan Lim

Cryptography and Security

Vanishing Watermarks: Diffusion-Based Image Editing Undermines Robust Invisible Watermarking

Fan Guo, Jiyu Kang, Qi Ming et al. (+2 more)

Robust invisible watermarking schemes aim to embed hidden information into images such that the watermark survives common manipulations. However, powerful diffusion-based image generation and editing techniques now pose a new threat to these watermarks. In this paper, we present a comprehensive theoretical and empirical analysis demonstrating that diffusion models can effectively erase robust watermarks even when those watermarks were designed to withstand conventional distortions. We show that a diffusion-driven image regeneration process, which leverages generative models to recreate an image, can remove embedded watermarks while preserving the image's perceptual content. Furthermore, we introduce a guided diffusion-based attack that explicitly targets the embedded watermark signal during generation, significantly degrading watermark detectability. Theoretically, we prove that as an image undergoes sufficient diffusion transformations, the mutual information between the watermarked image and the hidden payload approaches zero, leading to inevitable decoding failure. Experimentally, we evaluate multiple state-of-the-art watermarking methods (including deep learning-based schemes like StegaStamp, TrustMark, and VINE) and demonstrate that diffusion edits yield near-zero watermark recovery rates after attack, while maintaining high visual fidelity of the regenerated images. Our findings reveal a fundamental vulnerability in current robust watermarking techniques against generative model-based edits, underscoring the need for new strategies to ensure watermark resilience in the era of powerful diffusion models.
Read the full abstract

All authors Fan Guo, Jiyu Kang, Qi Ming, Emily Davis, Finn Carter

Cryptography and Security

ICSSPulse: A Modular LLM-Assisted Platform for Industrial Control System Penetration Testing

Michail Takaronis, Athanasia Kollarou, Vyron Kampourakis et al. (+2 more)

It is well established that industrial control systems comprise the operational backbone of modern critical infrastructures, yet their increasing connectivity exposes them to cyber threats that are difficult to study and remedy safely under real-time operational conditions. In this paper, we present ICSSPulse, an open-source, modular, and extensible penetration testing platform designed for the security assessment of ICS communication protocols. To the best of our knowledge, ICSSPulse is the first web-based platform that unifies network scanning, protocol-aware Modbus and OPC~UA interaction, and Large Language Model (LLM)-assisted reporting within a single, lightweight ecosystem. Our platform provides a user-friendly graphical interface that orchestrates enumeration, exploitation, and reporting activities over simulated industrial services, enabling safe and reproducible experimentation. It supports protocol-level discovery, asset enumeration, and controlled read/write interactions, while preserving protocol fidelity and operational transparency. Experimental evaluation using synthetic Modbus test servers, a Factory I/O water treatment scenario, and a custom OPC~UA production-line model demonstrated ICSSPulse's potential to discover active industrial services, enumerate process-relevant assets, and manipulate process variables. A key contribution of this work lies in the integration of an LLM-assisted reporting module that automatically translates technical findings into structured executive and technical reports, with mitigation guidance informed by the ICS MITRE ATT&CK ICS matrix.
Read the full abstract

All authors Michail Takaronis, Athanasia Kollarou, Vyron Kampourakis, Vasileios Gkioulos, Sokratis Katsikas

Cryptography and Security

Post-Quantum Sanitizable Signatures from McEliece-Based Chameleon Hashing

Shahzad Ahmad, Stefan Rass and Zahra Seyedi

We introduce a novel post-quantum sanitizable signature scheme constructed upon a chameleon hash function derived from the McEliece cryptosystem. In this design, the designated sanitizer possesses the inherent trapdoor of a Goppa code, which facilitates controlled collision-finding via Patterson decoding. This mechanism enables authorized modification of specific message blocks while ensuring all other content remains immutably bound. We provide formal security definitions and rigorous proofs of existential unforgeability and immutability, grounded in the hardness of syndrome decoding in the random-oracle model, where a robust random oracle thwarts trivial linear hash collisions. A key innovation lies in our precise characterization of the transparency property: by imposing a specific weight constraint on the randomizers generated by the signer, we achieve perfect transparency, rendering sanitized signatures indistinguishable from freshly signed ones. This work establishes the first transparent, code-based, post-quantum sanitizable signature scheme, offering strong theoretical guarantees and a pathway for practical deployment in long-term secure applications.
Read the full abstract

Cryptography and Security

OptiLeak: Efficient Prompt Reconstruction via Reinforcement Learning in Multi-tenant LLM Services

Longxiang Wang, Xiang Zheng, Xuhao Zhang et al. (+3 more)

Multi-tenant LLM serving frameworks widely adopt shared Key-Value caches to enhance efficiency. However, this creates side-channel vulnerabilities enabling prompt leakage attacks. Prior studies identified these attack surfaces yet focused on expanding attack vectors rather than optimizing attack performance, reporting impractically high attack costs that underestimate the true privacy risk. We propose OptiLeak, a reinforcement learning-enhanced framework that maximizes prompt reconstruction efficiency through two-stage fine-tuning. Our key insight is that domain-specific “hard tokens” – terms difficult to predict yet carrying sensitive information – can be automatically identified via likelihood ranking and used to construct preference pairs for Direct Preference Optimization, eliminating manual annotation. This enables effective preference alignment while avoiding the overfitting issues of extended supervised fine-tuning. Evaluated on three benchmarks spanning medical and financial domains, OptiLeak achieves up to $12.48\times$ reduction in average requests per token compared to baseline approaches, with consistent improvements across model scales from 3B to 14B parameters. Our findings demonstrate that cache-based prompt leakage poses a more severe threat than previously reported, underscoring the need for robust cache isolation in production deployments.
Read the full abstract

All authors Longxiang Wang, Xiang Zheng, Xuhao Zhang, Yao Zhang, Ye Wu, Cong Wang

Cryptography and Security · Cross-listed from cs.LG

Is the Trigger Essential? A Feature-Based Triggerless Backdoor Attack in Vertical Federated Learning

Yige Liu, Yiwei Lou, Che Wang et al. (+2 more)

As a distributed collaborative machine learning paradigm, vertical federated learning (VFL) allows multiple passive parties with distinct features and one active party with labels to collaboratively train a model. Although it is known for the privacy-preserving capabilities, VFL still faces significant privacy and security threats from backdoor attacks. Existing backdoor attacks typically involve an attacker implanting a trigger into the model during the training phase and executing the attack by adding the trigger to the samples during the inference phase. However, in this paper, we find that triggers are not essential for backdoor attacks in VFL. In light of this, we disclose a new backdoor attack pathway in VFL by introducing a feature-based triggerless backdoor attack. This attack operates under a more stringent security assumption, where the attacker is honest-but-curious rather than malicious during the training phase. It comprises three modules: label inference for the targeted backdoor attack, poison generation with amplification and perturbation mechanisms, and backdoor execution to implement the attack. Extensive experiments on five benchmark datasets demonstrate that our attack outperforms three baseline backdoor attacks by 2 to 50 times while minimally impacting the main task. Even in VFL scenarios with 32 passive parties and only one set of auxiliary data, our attack maintains high performance. Moreover, when confronted with distinct defense strategies, our attack remains largely unaffected and exhibits strong robustness. We hope that the disclosure of this triggerless backdoor attack pathway will encourage the community to revisit security threats in VFL scenarios and inspire researchers to develop more robust and practical defense strategies.
Read the full abstract

All authors Yige Liu, Yiwei Lou, Che Wang, Yongzhi Cao, Hanpin Wang

Cryptography and Security · Cross-listed from cs.CL

Personal Information Parroting in Language Models

Nishant Subramani, Kshitish Ghate and Mona Diab

Modern language models (LM) are trained on large scrapes of the Web, containing millions of personal information (PI) instances, many of which LMs memorize, increasing privacy risks. In this work, we develop the regexes and rules (R&R) detector suite to detect email addresses, phone numbers, and IP addresses, which outperforms the best regex-based PI detectors. On a manually curated set of 483 instances of PI, we measure memorization: finding that 13.6% are parroted verbatim by the Pythia-6.9b model, i.e., when the model is prompted with the tokens that precede the PI in the original document, greedy decoding generates the entire PI span exactly. We expand this analysis to study models of varying sizes (160M-6.9B) and pretraining time steps (70k-143k iterations) in the Pythia model suite and find that both model size and amount of pretraining are positively correlated with memorization. Even the smallest model, Pythia-160m, parrots 2.7% of the instances exactly. Consequently, we strongly recommend that pretraining datasets be aggressively filtered and anonymized to minimize PI parroting.
Read the full abstract

Cryptography and Security

Towards Secure and Efficient DNN Accelerators via Hardware-Software Co-Design

Wei Xuan, Zihao Xuan, Rongliang Fu et al. (+8 more)

The rapid deployment of deep neural network (DNN) accelerators in safety-critical domains such as autonomous vehicles, healthcare systems, and financial infrastructure necessitates robust mechanisms to safeguard data confidentiality and computational integrity. Existing security solutions for DNN accelerators, however, suffer from excessive hardware resource demands and frequent off-chip memory access overheads, which degrade performance and scalability. To address these challenges, this paper presents a secure and efficient memory protection framework for DNN accelerators with minimal overhead. First, we propose a bandwidth-aware cryptographic scheme that adapts encryption granularity based on memory traffic patterns, striking a balance between security and resource efficiency. Second, we observe that both the overlapping regions in the intra-layer tiling's sliding window pattern and those resulting from inter-layer tiling strategy discrepancies introduce substantial redundant memory accesses and repeated computational overhead in cryptography. Third, we introduce a multi-level authentication mechanism that effectively eliminates unnecessary off-chip memory accesses, enhancing performance and energy efficiency. Experimental results show that this work decreases performance overhead by over 12% and achieves 87% energy efficiency improvement for both server and edge neural processing units (NPUs), while ensuring robust scalability.
Read the full abstract

All authors Wei Xuan, Zihao Xuan, Rongliang Fu, Ning Lin, Kwunhang Wong, Zikang Yuan, Lang Feng, Zhongrui Wang, Tsung-Yi Ho, Yuzhong Jiao, Luhong Liang

Cryptography and Security

Understanding Human-AI Collaboration in Cybersecurity Competitions

Tingxuan Tang, Nicolas Janis, Kalyn Asher Montague et al. (+6 more)

Capture-the-Flag (CTF) competitions are increasingly becoming a testbed for evaluating AI capabilities at solving security tasks, due to the controlled environments and objective success criteria. Existing evaluations have focused on how successful AI is at solving CTF challenges in isolation from human CTF players. As AI usage increases in both academic and industrial settings, it is equally likely that human players may collaborate with AI agents to solve challenges. This possibility exposes a key knowledge gap: how do humans perceive AI CTF assistance; when assistance is provided, how do they collaborate and is it effective with respect to human performance; how do humans assisted by AI compare to the performance of fully autonomous AI agents on the same challenges. We address this gap with the first empirical study of AI assistance in a live, onsite CTF. In a study with 41 participants, we qualitatively study (i) how participants' perception, trust, and expectations shift before versus after hands-on AI use, and (ii) how participants collaborate with an instrumented AI agent. Moreover, we also (iii) benchmark four autonomous AI agents on the same fresh challenge set to compare outcomes with human teams and analyze agent trajectories. We find that, as the competition progresses, teams increasingly delegate larger subtasks to the AI, giving it more agency. Interestingly, CTF challenges solving rates are often constrained not by model's reasoning capabilities, but rather by the human players: ineffective prompting and poor context specification become the primary bottleneck. Remarkably, autonomous agents that self-direct their prompting and tool use bypass this bottleneck and outperform most human teams, coming in second overall in the competition. We conclude with implications for the future design of CTF challenges and for building effective human-in-the-loop AI systems for security.
Read the full abstract

All authors Tingxuan Tang, Nicolas Janis, Kalyn Asher Montague, Kevin Eykholt, Dhilung Kirat, Youngja Park, Jiyong Jang, Adwait Nadkarni, Yue Xiao

Monday 23 February

Cryptography and Security

Optimized Disaster Recovery for Distributed Storage Systems: Lightweight Metadata Architectures to Overcome Cryptographic Hashing Bottleneck

Prasanna Kumar, Nishank Soni and Gaurang Munje

Distributed storage architectures are foundational to modern cloud-native infrastructure, yet a critical operational bottleneck persists within disaster recovery (DR) workflows: the dependence on content-based cryptographic hashing for data identification and synchronization. While hash-based deduplication is effective for storage efficiency in steady-state operation, it becomes a systemic liability during failover and failback events when hash indexes are stale, incomplete, or must be rebuilt following a crash. This paper precisely characterizes the operational conditions under which full or partial re-hashing becomes unavoidable. The paper also analyzes the downstream impact of cryptographic re-hashing on Recovery Time Objective (RTO) compliance, and proposes a generalized architectural shift toward deterministic, metadata-driven identification. The proposed framework assigns globally unique composite identifiers to data blocks at ingestion time-independent of content analysis enabling instantaneous delta computation during DR without any cryptographic overhead.
Read the full abstract

Cryptography and Security

CryptRISC: A Secure RISC-V Processor for High-Performance Cryptography with Power Side-Channel Protection

Amisha Srivastava, Muskan Porwal and Kanad Basu

Cryptographic computations are fundamental to modern computing, ensuring data confidentiality and integrity. However, these operations are highly vulnerable to power side-channel attacks that exploit variations in power consumption to leak sensitive information. Masking is a widely used countermeasure, yet software-based techniques often introduce significant performance overhead and implementation complexity, while fixed-function hardware masking lacks flexibility across diverse cryptographic algorithms. In this paper, we present CryptRISC, the first RISC-V-based processor that combines cryptographic acceleration with hardware-level power side-channel resistance through an ISA-driven operand masking framework. Our design extends the CVA6 core with 64-bit RISC-V Scalar Cryptography Extensions and introduces two microarchitectural components: a Field Detection Layer, which identifies the dominant algebraic field of each cryptographic instruction, and a Masking Control Unit, which applies field-aware operand randomization at runtime. This enables dynamic selection of Boolean, affine, or arithmetic masking schemes based on instruction semantics, providing optimized protection across algorithms including AES, SHA-256, SHA-512, SM3, and SM4. Unlike prior approaches relying on static masking logic or software instrumentation, our method performs operand masking transparently within the execution pipeline without modifying instruction encoding. Experimental results show speedups up to 6.80$\times$ over baseline software implementations, with only a 1.86% hardware overhead relative to the baseline CVA6 core, confirming the efficiency and practicality of CryptRISC.
Read the full abstract

Cryptography and Security

The TCF doesn't really A(A)ID – Automatic Privacy Analysis and Legal Compliance of TCF-based Android Applications

Victor Morel, Cristiana Santos, Pontus Carlsson et al. (+2 more)

The Transparency and Consent Framework (TCF), developed by the Interactive Advertising Bureau (IAB) Europe, provides a de facto standard for requesting, recording, and managing user consent from European end-users. This framework has previously been found to infringe European data protection law and has subsequently been regularly updated. Previous research on the TCF focused exclusively on web contexts, with no attention given to its implementation in mobile applications. No work has systematically studied the privacy implications of the TCF on Android apps. To address this gap, we investigate the prevalence of the TCF in popular Android apps from the Google Play Store, and assess whether these apps respect users' consent banner choices. By scraping and downloading 4482 of the most popular Google Play Store apps on an emulated Android device, we automatically determine which apps use the TCF, automatically interact with consent banners, and analyze the apps' traffic in two different stages, passive (post choices) and active (during banner interaction and post choices). We found that 576 (12.85%) of the 4482 downloadable apps in our dataset implemented the TCF, and we identified potential privacy violations within this subset. In 15 (2.6%) of these apps, users' choices are stored only when consent is granted. Users who refuse consent are shown the consent banner again each time they launch the app. Network traffic analysis conducted during the passive stage reveals that 66.2% of the analyzed TCF-based apps share personal data, through the Android Advertising ID (AAID), in the absence of a lawful basis for processing. 55.3% of apps analyzed during the active stage share AAID before users interact with the apps' consent banners, violating the prior consent requirement.
Read the full abstract

All authors Victor Morel, Cristiana Santos, Pontus Carlsson, Joel Ahlinder, Romaric Duvignau

Cryptography and Security

Right to History: A Sovereignty Kernel for Verifiable AI Agent Execution

Jing Zhang

AI agents increasingly act on behalf of humans, yet no existing system provides a tamper-evident, independently verifiable record of what they did. As regulations such as the EU AI Act begin mandating automatic logging for high-risk AI systems, this gap carries concrete consequences – especially for agents running on personal hardware, where no centralized provider controls the log. Extending Floridi's informational rights framework from data about individuals to actions performed on their behalf, this paper proposes the Right to History: the principle that individuals are entitled to a complete, verifiable record of every AI agent action on their own hardware. The paper formalizes this principle through five system invariants with structured proof sketches, and implements it in PunkGo, a Rust sovereignty kernel that unifies RFC 6962 Merkle tree audit logs, capability-based isolation, energy-budget governance, and a human-approval mechanism. Adversarial testing confirms all five invariants hold. Performance evaluation shows sub-1.3 ms median action latency, ~400 actions/sec throughput, and 448-byte Merkle inclusion proofs at 10,000 log entries.
Read the full abstract

Cryptography and Security · Cross-listed from cs.SE

CodeHacker: Automated Test Case Generation for Detecting Vulnerabilities in Competitive Programming Solutions

Jingwei Shi, Xinxiang Yin, Jing Huang et al. (+2 more)

The evaluation of Large Language Models (LLMs) for code generation relies heavily on the quality and robustness of test cases. However, existing benchmarks often lack coverage for subtle corner cases, allowing incorrect solutions to pass. To bridge this gap, we propose CodeHacker, an automated agent framework dedicated to generating targeted adversarial test cases that expose latent vulnerabilities in program submissions. Mimicking the hack mechanism in competitive programming, CodeHacker employs a multi-strategy approach, including stress testing, anti-hash attacks, and logic-specific targeting to break specific code submissions. To ensure the validity and reliability of these attacks, we introduce a Calibration Phase, where the agent iteratively refines its own Validator and Checker via self-generated adversarial probes before evaluating contestant code.Experiments demonstrate that CodeHacker significantly improves the True Negative Rate (TNR) of existing datasets, effectively filtering out incorrect solutions that were previously accepted. Furthermore, generated adversarial cases prove to be superior training data, boosting the performance of RL-trained models on benchmarks like LiveCodeBench.
Read the full abstract

All authors Jingwei Shi, Xinxiang Yin, Jing Huang, Jinman Zhao, Shengyu Tao

Cryptography and Security

Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks

David Schmotz, Luca Beurer-Kellner, Sahar Abdelnabi et al. (+1 more)

LLM agents are evolving rapidly, powered by code execution, tools, and the recently introduced agent skills feature. Skills allow users to extend LLM applications with specialized third-party code, knowledge, and instructions. Although this can extend agent capabilities to new domains, it creates an increasingly complex agent supply chain, offering new surfaces for prompt injection attacks. We identify skill-based prompt injection as a significant threat and introduce SkillInject, a benchmark evaluating the susceptibility of widely-used LLM agents to injections through skill files. SkillInject contains 202 injection-task pairs with attacks ranging from obviously malicious injections to subtle, context-dependent attacks hidden in otherwise legitimate instructions. We evaluate frontier LLMs on SkillInject, measuring both security in terms of harmful instruction avoidance and utility in terms of legitimate instruction compliance. Our results show that today's agents are highly vulnerable with up to 80% attack success rate with frontier models, often executing extremely harmful instructions including data exfiltration, destructive action, and ransomware-like behavior. They furthermore suggest that this problem will not be solved through model scaling or simple input filtering, but that robust agent security will require context-aware authorization frameworks. Our benchmark is available at https://www.skill-inject.com/.
Read the full abstract

All authors David Schmotz, Luca Beurer-Kellner, Sahar Abdelnabi, Maksym Andriushchenko

Cryptography and Security · Cross-listed from cs.PL

The LLMbda Calculus: AI Agents, Conversations, and Information Flow

Zac Garby, Andrew D. Gordon and David Sands

Large language models are increasingly deployed as agents: they plan, call tools, read untrusted data, and act on the results. This exposes them to prompt injection: data meant only to be read is obeyed as an instruction. The most principled defences replace content inspection with provenance: classifying data by source and keeping trusted and untrusted apart through a separation of duty (the dual-LLM pattern) and information-flow control. Yet the leading systems are hard to fully trust: flow tracking is easy to get wrong, deliberate relaxations are hard to audit, and the dual-LLM pattern is hard-wired into the architecture. We present LLMbda, an untyped call-by-value lambda calculus that makes provenance-based defence both expressible and provably sound, without committing to an architecture. It adds the operational core of agentic systems as first-class constructs: prompt-response conversations that can be forked and cleared, code generation, and dynamic information-flow control in which every value carries a label that every reduction propagates. Isolation becomes a policy a program expresses, and reclassification an explicit, auditable construct. Our central result is a termination-insensitive probabilistic noninterference theorem over the whole calculus, including code-generating agents, with an insulated variant that holds even when the attacker chooses all untrusted inputs. The verified interpreter is itself the harness that calls the model, to our knowledge the first LLM agent harness whose executable is the subject of machine-checked security theorems, so every agent inherits the guarantee. On the AgentDojo banking benchmark, an agent built within LLMbda, enforcement always on, matches the utility of CaMeL, a leading dual-LLM defence, run without its policy checks (which halve its utility), and resists all but two of 1296 attacked runs. Our harness and all proofs are in Lean.
Read the full abstract

Cryptography and Security

Can You Tell It's AI? Human Perception of Synthetic Voices in Vishing Scenarios

Zoha Hayat Bhatti, Bakhtawar Ahtisham, Seemal Tausif et al. (+3 more)

Large Language Models and commercial speech synthesis systems now enable highly realistic AI-generated voice scams (vishing), raising urgent concerns about deception at scale. Yet it remains unclear whether individuals can reliably distinguish AI-generated speech from human-recorded voices in realistic scam contexts and what perceptual strategies underlie their judgments. We conducted a controlled online study in which 22 participants evaluated 16 vishing-style audio clips (8 AI-generated, 8 human-recorded) and classified each as human or AI while reporting confidence. Participants performed poorly: mean accuracy was 37.5%, below chance in a binary classification task. At the stimulus level, misclassification was bidirectional: 75% of AI-generated clips were majority-labeled as human, while 62.5% of human-recorded clips were majority-labeled as AI. Signal Detection Theory analysis revealed near-zero discriminability (d' approx 0), indicating inability to reliably distinguish synthetic from human voices rather than simple response bias. Qualitative analysis of 315 coded excerpts revealed reliance on paralinguistic and emotional heuristics, including pauses, filler words, vocal variability, cadence, and emotional expressiveness. However, these surface-level cues traditionally associated with human authenticity were frequently replicated by AI-generated samples. Misclassifications were often accompanied by moderate to high confidence, suggesting perceptual miscalibration rather than uncertainty. Together, our findings demonstrate that authenticity judgments based on vocal heuristics are unreliable in contemporary vishing scenarios. We discuss implications for security interventions, user education, and AI-mediated deception mitigation.
Read the full abstract

All authors Zoha Hayat Bhatti, Bakhtawar Ahtisham, Seemal Tausif, Niklas George, Nida ul Habib Bajwa, Mobin Javed

Cryptography and Security · Cross-listed from cs.PL

Misquoted No More: Securely Extracting F* Programs with IO

Cezar-Constantin Andrici, Abigail Pribisova, Danel Ahman et al. (+3 more)

Shallow embeddings that use monads to represent effects are popular in proof-oriented languages because they are convenient for formal verification. Once shallowly embedded programs are verified, they are often extracted to mainstream languages like OCaml or C and linked into larger codebases. The extraction process is not fully verified because it often involves quotation – turning the shallowly embedded program into a deeply embedded one – and verifying quotation remains a major open challenge. Instead, some prior work obtains formal correctness guarantees using translation validation to certify individual extraction results. We build on this idea, but limit the use of translation validation to a first extraction step that we call relational quotation and that uses a metaprogram to construct a typing derivation for the given shallowly embedded program. This metaprogram is simple, since the typing derivation follows the structure of the original program. Once we validate that the typing derivation is valid for the original program, we pass it to a verified syntax-generation function that produces code guaranteed to be semantically related to the original program. We apply this general idea to build SEIO*, a framework for extracting shallowly embedded F* programs with IO and refinement types to a deeply embedded simply typed lambda-calculus while providing formal secure compilation guarantees. Using two cross-language logical relations, we devise a machine-checked proof in F* that SEIO* guarantees Robust Relational Hyperproperty Preservation (RrHP), a very strong secure compilation criterion that implies full abstraction as well as preservation of trace properties and hyperproperties against arbitrary linked adversarial code. This goes beyond the state of the art in verified and certifying extraction, which so far has focused on correctness rather than security.
Read the full abstract

All authors Cezar-Constantin Andrici, Abigail Pribisova, Danel Ahman, Catalin Hritcu, Exequiel Rivas, Théo Winterhalter

Cryptography and Security

RobPI: Robust Private Inference against Malicious Client

Jiaqi Xue, Mengxin Zheng and Qian Lou

The increased deployment of machine learning inference in various applications has sparked privacy concerns. In response, private inference (PI) protocols have been created to allow parties to perform inference without revealing their sensitive data. Despite recent advances in the efficiency of PI, most current methods assume a semi-honest threat model where the data owner is honest and adheres to the protocol. However, in reality, data owners can have different motivations and act in unpredictable ways, making this assumption unrealistic. To demonstrate how a malicious client can compromise the semi-honest model, we first designed an inference manipulation attack against a range of state-of-the-art private inference protocols. This attack allows a malicious client to modify the model output with 3x to 8x fewer queries than current black-box attacks. Motivated by the attacks, we proposed and implemented RobPI, a robust and resilient private inference protocol that withstands malicious clients. RobPI integrates a distinctive cryptographic protocol that bolsters security by weaving encryption-compatible noise into the logits and features of private inference, thereby efficiently warding off malicious-client attacks. Our extensive experiments on various neural networks and datasets show that RobPI achieves ~91.9% attack success rate reduction and increases more than 10x the number of queries required by malicious-client attacks.
Read the full abstract

Cryptography and Security

LLM-enabled Applications Require System-Level Threat Monitoring

Yedi Zhang, Haoyu Wang, Xianglin Yang et al. (+2 more)

LLM-enabled applications are rapidly reshaping the software ecosystem by using large language models as core reasoning components for complex task execution. This paradigm shift, however, introduces fundamentally new reliability challenges and significantly expands the security attack surface, due to the non-deterministic, learning-driven, and difficult-to-verify nature of LLM behavior. In light of these emerging and unavoidable safety challenges, we argue that such risks should be treated as expected operational conditions rather than exceptional events, necessitating a dedicated incident-response perspective. Consequently, the primary barrier to trustworthy deployment is not further improving model capability but establishing system-level threat monitoring mechanisms that can detect and contextualize security-relevant anomalies after deployment – an aspect largely underexplored beyond testing or guardrail-based defenses. Accordingly, this position paper advocates systematic and comprehensive monitoring of security threats in LLM-enabled applications as a prerequisite for reliable operation and a foundation for dedicated incident-response frameworks.
Read the full abstract

All authors Yedi Zhang, Haoyu Wang, Xianglin Yang, Jin Song Dong, Jun Sun

Cryptography and Security

An Explainable Memory Forensics Approach for Malware Analysis

Silvia Lucia Sanna, Davide Maiorca and Giorgio Giacinto

Memory forensics is an effective methodology for analyzing living-off-the-land malware, including threats that employ evasion, obfuscation, anti-analysis, and steganographic techniques. By capturing volatile system state, memory analysis enables the recovery of transient artifacts such as decrypted payloads, executed commands, credentials, and cryptographic keys that are often inaccessible through static or traditional dynamic analysis. While several automated models have been proposed for malware detection from memory, their outputs typically lack interpretability, and memory analysis still relies heavily on expert-driven inspection of complex tool outputs, such as those produced by Volatility. In this paper, we propose an explainable, AI-assisted memory forensics approach that leverages general-purpose large language models (LLMs) to interpret memory analysis outputs in a human-readable form and to automatically extract meaningful Indicators of Compromise (IoCs), in some circumstances detecting more IoCs than current state-of-the-art tools. We apply the proposed methodology to both Windows and Android malware, comparing full RAM acquisition with target-process memory dumping and highlighting their complementary forensic value. Furthermore, we demonstrate how LLMs can support both expert and non-expert analysts by explaining analysis results, correlating artifacts, and justifying malware classifications. Finally, we show that a human-in-the-loop workflow, assisted by LLMs during kernel-assisted setup and analysis, improves reproducibility and reduces operational complexity, thereby reinforcing the practical applicability of AI-driven memory forensics for modern malware investigations.
Read the full abstract

Cryptography and Security

Quantum approaches to learning parity with noise

Daniel Shiu

The learning parity with noise (LPN) problem is a well-established computational challenge whose difficulty is critical to the security of several post-quantum cryptographic primitives such as HQC and Classic McEliece. Classically, the best-known attacks involve information set decoding methods which are exponential in complexity for parameterisations of interest. In this paper we investigate whether quantum methods might offer alternative approaches. The line of inquiry is inspired by Regev's relating of certain lattice problems to the hidden dihedral subgroup problem. We use neighbourhoods of binary fields to produce a function close to fulfilling Simon's promise with difference equal to the secret parity vector. Although unlikely to recover the secret parity vector directly, running Simon's algorithm essentially produces new LPN samples. This gives the hope that we might be able to produce enough new samples to ignore one or more variables and iteratively reduce the problem. We make no claim that these methods will necessarily be competitive with existing approaches, merely that they warrant deeper investigation.
Read the full abstract

Cryptography and Security

SafePickle: Robust and Generic ML Detection of Malicious Pickle-based ML Models

Hillel Ohayon, Daniel Gilkarov and Ran Dubin

Model repositories such as Hugging Face increasingly distribute machine learning artifacts serialized with Python's pickle format, exposing users to remote code execution (RCE) risks during model loading. Recent defenses, such as PickleBall, rely on per-library policy synthesis that requires complex system setups and verified benign models, which limits scalability and generalization. In this work, we propose a lightweight, machine-learning-based scanner that detects malicious Pickle-based files without policy generation or code instrumentation. Our approach statically extracts structural and semantic features from Pickle bytecode and applies supervised and unsupervised models to classify files as benign or malicious. We construct and release a labeled dataset of 727 Pickle-based files from Hugging Face and evaluate our models on four datasets: our own, PickleBall (out-of-distribution), Hide-and-Seek (9 advanced evasive malicious models), and synthetic joblib files. Our method achieves 90.01% F1-score compared with 7.23%-62.75% achieved by the SOTA scanners (Modelscan, Fickling, ClamAV, VirusTotal) on our dataset. Furthermore, on the PickleBall data (OOD), it achieves 81.22% F1-score compared with 76.09% achieved by the PickleBall method, while remaining fully library-agnostic. Finally, we show that our method is the only one to correctly parse and classify 9/9 evasive Hide-and-Seek malicious models specially crafted to evade scanners. This demonstrates that data-driven detection can effectively and generically mitigate Pickle-based model file attacks.
Read the full abstract

Cryptography and Security

AegisSat: Securing AI-Enabled SoC FPGA Satellite Platforms

Huimin Li, Vusal Novruzov, Nikhilesh Singh et al. (+3 more)

The increasing adoption of System-on-Chip Field-Programmable Gate Arrays (SoC FPGAs) in AI-enabled satellite systems, valued for their reconfigurability and in-orbit update capabilities, introduces significant security challenges. Compromised updates can lead to performance degradation, service disruptions, or adversarial manipulation of mission outcomes. To address these risks, this paper proposes a comprehensive security framework, AegisSat. It ensures the integrity and resilience of satellite platforms by (i) integrating cryptographically-based secure boot mechanisms to establish a trusted computing base; (ii) enforcing strict runtime resource isolation; (iii) employing authenticated procedures for in-orbit reconfiguration and AI model updates to prevent unauthorized modifications; and (iv) providing robust rollback capabilities to recover from boot and update failures and maintain system stability. To further support our claims, we conducted experiments demonstrating the integration of these mechanisms on contemporary SoC FPGA devices. This defense-in-depth framework is crucial for space applications, where physical access is impossible and systems must operate reliably over extended periods, thereby enhancing the trustworthiness of SoC FPGA-based satellite systems and enabling secure and resilient AI operations in orbit.
Read the full abstract

All authors Huimin Li, Vusal Novruzov, Nikhilesh Singh, Lichao Wu, Mohamadreza Rostami, Ahmad-Reza Sadeghi

Cryptography and Security

Predicting known Vulnerabilities from Attack News: A Transformer-Based Approach

Refat Othman, Diaeddin Rimawi, Bruno Rossi et al. (+1 more)

Identifying the vulnerabilities exploited during cyberattacks is essential for enabling timely responses and effective mitigation in software security. This paper directly examines the process of predicting software vulnerabilities, specifically Common Vulnerabilities and Exposures (CVEs), from unstructured descriptions of attacks reported in cybersecurity news articles. We propose a semantic similarity-based approach utilizing the multi-qa-mpnet-base-dot-v1 (MPNet) sentence transformer model to generate a ranked list of the most likely CVEs corresponding to each news report. To assess the accuracy of the predicted vulnerabilities, we implement four complementary validation methods: filtering predictions based on similarity thresholds, conducting manual validation, performing semantic comparisons with the first vulnerability explicitly mentioned in each report, and comparing against all CVEs referenced within the report. Experimental results, drawn from a dataset of 100 SecurityWeek news articles, demonstrate that the model attains a precision of 81 percent when employing threshold-based filtering. Manual evaluations report that 70 percent of the predictions are relevant, while comparisons with the initially mentioned CVEs reveal agreement rates of 80 percent with the first listed vulnerability and 78 percent across all referenced CVEs. In 57 percent of the news reports analyzed, at least one predicted vulnerability precisely matched a CVE-ID mentioned in the article. These findings underscore the model's potential to facilitate automated vulnerability identification from real-world cyberattack news reports.
Read the full abstract

All authors Refat Othman, Diaeddin Rimawi, Bruno Rossi, Barbara Russo

Cryptography and Security

Efficient Multi-Party Secure Comparison over Different Domains with Preprocessing Assistance

Kaiwen Wang, Xiaolin Chang, Yuehan Dong et al. (+1 more)

Secure comparison is a fundamental primitive in multi-party computation, supporting privacy-preserving applications such as machine learning and data analytics. A critical performance bottleneck in comparison protocols is their preprocessing phase, primarily due to the high cost of generating the necessary correlated randomness. Recent frameworks introduce a passive, non-colluding dealer to accelerate preprocessing. However, two key issues still remain. First, existing dealer-assisted approaches treat the dealer as a drop-in replacement for conventional preprocessing without redesigning the comparison protocol to optimize the online phase. Second, most protocols are specialized for particular algebraic domains, adversary models, or party configurations, lacking broad generality. In this work, we present the first dealer-assisted $n$-party LTBits (Less-Than-Bits) and MSB (Most Significant Bit) extraction protocols over both $\mathbb{F}_p$ and $\mathbb{Z}_{2^k}$, achieving perfect security at the protocol level. By fully exploiting the dealer's capability to generate rich correlated randomness, our $\mathbb{F}_p$ construction achieves constant-round online complexity and our $\mathbb{Z}_{2^k}$ construction achieves $O(\log_n k)$ rounds with tunable branching factor. All protocols are formulated as black-box constructions via an extended ABB model, ensuring portability across MPC backends and adversary models. Experimental results demonstrate $1.79\times$ to $19.4\times$ speedups over state-of-the-art MPC frameworks, highlighting the practicality of our protocols for comparison-intensive MPC applications.
Read the full abstract

All authors Kaiwen Wang, Xiaolin Chang, Yuehan Dong, Ruichen Zhang

Cryptography and Security

SOK: A Taxonomy of Attack Vectors and Defense Strategies for Agentic Supply Chain Runtime

Xiaochong Jiang, Shiqi Yang, Wenting Yang et al. (+2 more)

Agentic systems based on large language models (LLMs) operate not merely as text generators but as autonomous entities that dynamically retrieve information and invoke tools. This execution model shifts the attack surface from traditional build-time artifacts to inference-time dependencies, exposing agents to manipulation through untrusted data and probabilistic capability resolution. While prior work has examined model-level vulnerabilities, security risks arising from the complex, cyclic runtime behavior of agents remain fragmented. This paper systematizes existing research into a unified runtime framework. We categorize threats into data supply chain attacks (distinguishing between transient context injection and persistent memory poisoning) and tool supply chain attacks (spanning discovery, implementation, and invocation phases). Crucially, we identify the emergence of the Viral Agent Loop, where agents effectively become vectors for self-propagating generative worms that require no code vulnerabilities to spread. We argue for a transition to a Zero-Trust Runtime Architecture, where context is treated as untrusted control flow, and tool execution is bounded by cryptographic provenance rather than semantic likelihood.
Read the full abstract

All authors Xiaochong Jiang, Shiqi Yang, Wenting Yang, Yichen Liu, Cheng Ji