Papers

Verifiable credentials (VCs) and decentralized identifiers (DIDs) enable selective disclosure but lack legal anchoring: without a trusted identity root, verifiers cannot distinguish a genuine holder from a fabricated identity. State identity systems provide biometric-grounded verification but impose three costs: verifiers must collect subjects' full personally identifiable information, infrastructure concentrates on a single API, and the state observes every transaction. We present China RealDID, a three-layer architecture – CTID (centralized legal identity), RealDID (decentralized anchor on an open permissioned blockchain), and VCs with SD-JWT-based selective disclosure – evaluated against five adversary classes and six security goals. The central mechanism is a content-blind government relay: the state authenticates participants and counter-signs every credential but cannot read the payload, encrypted by the issuer to the holder's public key. We describe the VC lifecycle, triple-signature chain, open template registry, and the architecture's metadata-level privacy limits, including the credential graph at the relay and presentation linkability from single-DID reuse. The design yields an asymmetric, state-bounded trust model: the state cannot impersonate or read contents; the user cannot fabricate identity or evade metadata observation. We analyze alignment with China's Personal Information Protection Law and the EU's GDPR, including the tension between immutable registries and erasure rights, and discuss generalizability through cross-border deployments with Singapore and Hong Kong.

Large language models and vision-language models are increasingly used as high-level planners in robotic systems, using task goals and sensor summaries to select navigation or manipulation actions. This creates a new backdoor surface: a compromised planner can behave normally in most runs, yet change its target selection when a hidden trigger is present. Prior attacks on LLM-powered or embodied agents mainly rely on triggers that appear in language, camera-visible objects, scene semantics, or specific sequences of past actions. This paper presents StepTrigger, a contact-state-triggered backdoor attack for VLM-powered legged robots. The trigger is not a prompt token or a visible marker. It is produced by pressure and foot-ground contact patterns that arise when a Unitree Go1 quadruped walks across a dense terrain patch. Unlike conventional visual or textual triggers, contact signals are inherently noisy and may also arise during benign locomotion. To avoid treating every pressure anomaly as a trigger, StepTrigger learns a selective backdoor policy from multimodal robot state, using incidental pressure events as benign examples and dense-patch contacts as poisoned examples. In a stratified offline evaluation, the trained planner achieved 98.75% clean behavior preservation, 92.50% false-trigger rejection, 76.25% true-trigger activation, and 89.17% overall parsed behavior accuracy. These results reveal a backdoor surface in proprioceptive and contact channels that is not captured by defenses focused only on language, vision, or action history.

Four years after prompt injection was first identified in 2022, attacks are still predominantly documented as verbatim strings rather than structured exploits, despite advancing agent capabilities and threat actors embedding injections to subvert AI-assisted security analysis. This paper formalizes the structure of prompt-injection artifacts, enabling defenders, red teamers, and cyber threat intelligence (CTI) teams to label, compare, and mutate attacks without relying on fragile string matching. Because large language models compile varied natural-language realizations into identical executable actions, labeling must track attacker intent (tool targets, sinks, and effects) rather than surface wording. We propose a seven-component model (carrier, delivery vector, concealment, context-break, privilege escalation, payload, and return channel) consisting of five artifact fields and two environment fields. This framework unifies roles partially addressed by HOUYI's payload decomposition, the Promptware Kill Chain, and campaign taxonomies, while framing minimal jailbreak frameworks like ReNeLLM as projections onto a restricted subspace. We provide clear labeling rules, a logical analysis record mapping directly to industry CTI schemas, worked examples including EchoLeak (CVE-2025-32711) and an in-the-wild malware AI-evasion sample, and an illustrative agentic flowchart.

Software teams now delegate production code to language models, including code that provisions storage, handles credentials, and stores regulated data, so we asked whether a model applies the controls a SOC~2 program expects (encryption, restricted access, logging, retention) when nobody mentions security, and how much one sentence naming the standard changes the answer. We tested three frontier models (Claude Fable~5, Opus~4.8, and Opus~5) across four use cases (an S3 CLI, an authentication service, an RDS Terraform module, and a file-upload handler holding personal data), each generated once from a neutral task statement and once with a single SOC~2 sentence added, scoring all 24 outputs against binary rubrics mapped to specific Trust Services Criteria and hand-verifying every failure and flagged act. Unprompted conformance ran from 47% to 88% and tracked whether a control is part of how the code is normally written, so password hashing and storage\_encrypted appear unasked while S3 hardening calls, retention, and MFA hooks do not. The neutral prompt also shipped real vulnerabilities, including a reachable Werkzeug debugger allowing remote code execution, an unauthenticated download, and an endpoint returning every stored name and email, all scored clean by our first checklist, with a fourth defect passing because its value was computed by a conditional. One SOC~2 sentence moved every case to 86–100%, worth 23 to 50 points, and removed every insecure construction, though controls outside the model's conception of the task survived it, including MFA hooks, cookie flags, and account lifecycle. Model choice mattered least, with same-generation models within one rubric item across all eight cells, and the pattern-matching scorer proved unreliable, disagreeing with semantic grading on 27 of 216 judgments and passing a real defect, so it needs replacing with semantic checks.

LLM benchmarks can build an organization's reputation and attract customers, but only when results are transparent and verifiable. Unverified claims that DeepSeek R1 outperformed OpenAI's o1 contributed to market panic on January 27, 2025, when Nvidia lost USD589 billion in market value. Yet vendor benchmarks often depend on an honor system. Academic reassessments and independent leaderboards have found undisclosed changes to proprietary models, contaminated training data, and selective reporting. LLM-as-a-judge methods scale evaluation by reducing human review. Studies, however, suggest that judges may show identity-aware bias, scoring an answer according to its source model rather than its quality. This bias has not been fully measured or corrected across politically sensitive, reasoning-intensive, and preference-based tasks. We examine this problem using seven verifier models: GPT-OSS 120B, Llama 3.3 70B, GLM 5.1, Qwen3 32B, DeepSeek V4 Pro, Mistral Large3, and Sarvam M. They score anonymous and identity-disclosed responses from three primary models on 58 factual, reasoning, political, and preference-based questions. Identity disclosure slightly raises scores for factual questions, moderately affects stress-reasoning tasks, and causes large changes for geopolitically sensitive topics. Notable results include GLM5.1 (+7.00 points, p = 0.0249) and Llama 3.3 70B (+1.56 points, p = 0.00). We also introduce a blockchain-based commit-reveal protocol using Autonomous Economic Agents on an Ethereum-compatible ledger. In Phase 1, each judge records a one-way hash of its score and a secret salt before candidate identities are revealed. In Phase 2, the identity and raw score are disclosed and verified on-chain. This creates a tamper-evident audit trail that separates blind evaluation from post-hoc claims and reduces the verification burden on independent researchers and leaderboard operators.

We study how to share a single conserved capacity budget across many locations and two service classes when demand is uneven, time-varying, and can exceed supply. The shape recurs: an origin's request-rate cap split across its edge locations, a licensed throughput cap across premium and standard tenants, or an egress budget between latency-critical and batch workloads. We present a two-level algorithm. The first level redistributes capacity within a class across locations by proportional deficit and excess redistribution; the second lends capacity elastically between classes when one has surplus and the other deficit. We prove it conserves the budget exactly, preserves non-negativity, and reaches a stable allocation in one iteration under stationary demand because it carries no per-cycle state, at O(KN) cost per cycle for K classes and N locations. We evaluate it defending a CDN's per-domain budget under volumetric attack, where the classes are confirmed-legitimate and not-yet-cleared traffic; across 8 contention scenarios on a 22-location topology it serves 66-93% of high-priority demand, competitive with a single-class linear-programming optimum, while never leaving capacity idle or over-committing whenever aggregate demand meets or exceeds the budget (the contention regime these scenarios evaluate). Two findings carry beyond the application. First, a throughput-maximizing objective is wrong under contention: a two-class LP maximizing total served load serves less high-priority load than our demand-proportional, reservation-respecting allocator in most scenarios, because it cannot tell that some load it serves is the contention. Second, inter-class borrowing earns its complexity under bursty load, improving high-priority service by 1.5 points (isolated by ablation), and is neutral under stationary demand. A 5-location prototype with real HTTP traffic validates the pipeline.

IT audits require auditors to judge whether heterogeneous organizational evidence satisfies semantic security and compliance controls. This judgment is difficult to automate because relevant evidence is distributed across policies, records, spreadsheets, and operational artifacts, and because audit conclusions depend on evidentiary sufficiency rather than keyword matching. We present IntelliAudit, a retrieval-grounded multi-agent system for IT audit evidence evaluation. Given a control and an evidence corpus, IntelliAudit retrieves relevant artifacts, generates an evidence-grounded assessment, challenges adverse findings, adjudicates disagreements, and produces an auditor-facing recommendation with cited evidence, rationale, missing-evidence analysis, and remediation guidance. We instantiate IntelliAudit on ISO/IEC 27001 and evaluate it across multiple simulated organizations using expert auditor review and audit-readiness user feedback. The evaluation shows that IntelliAudit can support control interpretation, evidence-grounded reasoning, and audit-preparation workflows, while also revealing the importance of human oversight for calibrating sufficiency judgments and correcting overly permissive recommendations. These results suggest that retrieval-grounded multi-agent systems can assist audit evidence review, but should remain decision-support tools rather than autonomous certification systems.

Deep-OCR (DeepSeek-OCR) advances document recognition by treating the visual modality as an optical compression medium, enabling long-context OCR at low token cost. However, its increased complexity may introduce new security vulnerabilities. In this paper, we present, to the best of our knowledge, the first pure black-box adversarial attack against a generative OCR vision-language model, where only the decoded string can be queried and no gradients, logits, or model internals are available. We recast the attack as a zeroth-order optimization problem driven by a bounded scalar loss defined directly on the string output via sequence similarity, and estimate the gradient with a random-direction finite-difference scheme whose query cost is independent of the image dimension. An Adam update with ell_infinity projection yields imperceptible perturbations for both untargeted and targeted objectives. Pilot experiments on Deep-OCR validate the string-only attack and evaluation pipeline and expose severe qualitative decoder failures, including repetition, truncation, and prompt leakage. They also show that controlled targeted rewriting remains substantially harder than untargeted degradation; we avoid claiming targeted success until the pre-registered evaluation is complete.

Quantum Key Distribution (QKD) protocols provide information-theoretic security by using quantum mechanical principles. Yet QKD is fundamentally a hybrid protocol: its security depends on the correct integration of the quantum phase with classical post-processing. While ETSI and ITUT specifications standardize QKD architectures and interfaces, they evaluate protocol security in isolation, leaving cross-layer interactions as an underexplored attack surface. This paper introduces a formal verification framework that holistically models QKD protocols based on ETSI and ITUT QKD specifications. Our model is the first hybrid QKD protocol model that supports automated analysis of protocollevel security focusing on how classical operations influence the security guarantees provided by the quantum phase of the QKD protocol. We formalize a comprehensive symbolic model of QKD protocols, based on ETSI and ITU-T QKD specifications, in Tamarin, an automated protocol verifier. Applying this framework, we obtain formal evidence of three specification-level vulnerabilities in ETSI- and ITU-T-grounded protocol models under adversary Eve+: subverted entanglement injection, basis-deferred measurement, and message reflection. Each arises from a classical control-plane omission in the procedure text and is established under a symbolic abstraction rather than as a claim about all practical deployments. We introduce two protocol improvements: measurement commitment and identitybound message authentication codes (MACs). Tamarin verification confirms that these countermeasures eliminate the identified vulnerabilities under Eve+. We have communicated our results and recommendations to relevant standardization organizations.

Phishing attacks have always been a favored vector for adversaries to defraud users, bypass modern defense mechanisms, and penetrate critical systems. Among all the elements contributing to the creation and deployment of successful phishing attacks, phishkits stand out as a crucial parameter. Phishkits often facilitate creating and deploying compelling phishing pages, implement evasion strategies, and establish and maintain backdoors with remote adversaries for exchanging leaked data. In this work, we performed an analysis of 1,300 modern phishkits collected from 2020 to 2023. We analyzed the architecture, source code, communication channels, and the nature of leaked data shared with adversaries. We identified mechanisms for dynamic redirection and attributing incoming web traffic as part of the evasion and cloaking mechanism. We also observed heavy reliance on current messaging services for exchanging stolen data with phishers. That said, our analysis shows that the number of phishkits with advanced functionalities is quite small. We identified 284 (21.8%) phishkits that did not use any form of evasion mechanism. We also observed that while there were differences in the implementation details of phishkits, the major components that keep phishing pages functional were very similar or even identical across kits. The level of code reuse and heavy reliance on known tricks to build pre-packaged phishing pages make a large number of cases predictable, which can potentially make the detection of these adversarial operations even easier at scale.

Baseband processors are reachable over the radio at all times. Their most security-relevant logic runs deep inside protocol state machines: the control-plane handlers that gate registration, authentication, and session setup. Analyzing that logic systematically requires introspecting the firmware as it runs, which makes re-hosting the baseband necessary. Existing re-hosting work approximates the execution environment and under-approximates the SoC complexity of the baseband processor together with its surrounding components, bringing this state practically out of reach. We instead model each surrounding component, co-processors, SIM, application processor, from what a real device does, and step them in lockstep with the baseband on one shared clock. That makes faithfulness checkable at component interfaces, rather than assumed. We call this method Unislop and demonstrate it on the UNISOC UDX710, a platform in an estimated 10-15% of cellular modems and in automotive systems, not systematically analyzed before. Starting from a Quectel RM500U-CNV module, we gain code execution, defeat its firmware-integrity check, instrument the baseband, and recover its peripheral environment from the running device. The resulting re-host reaches the same control-plane states as the real device, establishes a full PDU session, and carries real IP traffic on both ingress and egress. The recovered components are shared across UNISOC's baseband lineup, so with additional reverse-engineering effort the same design extends to further targets.

Cryptoasset systems often bind cryptographic key control to financial control: losing a wallet seed, custody share, hardware device, or smart-account credential can remove spend authority, while compromised recovery can enable theft. Existing work treats recovery through separate vocabularies–key backup, secret sharing, account recovery, credential re-issuance, social recovery, and asset migration–making mechanisms and tradeoffs difficult to compare. This paper presents a Systematization of Knowledge (SoK) on cryptographic key recovery for cryptoasset custody and financial technologies. Starting from a 118-paper systematic-review discovery corpus, we derive a 77-paper synthesis corpus and code each retained system in a master matrix covering recovered objects, recovery semantics, mechanisms, enrollment and storage, authorization, trust placement, failure events, post-recovery state, validation evidence, deployment status, privacy, usability, and limitations. The matrix supports an axis-first taxonomy that separates secret-restoring, hybrid, control-restoring, forensic/extractive, and framework-oriented recovery. Our central observation is that recovery is not a single operation: systems may reconstruct an original secret, regenerate a seed, restore a share, reissue a credential, migrate signing authority, restore account control, move assets, or extract forensic artifacts. We derive a generalized construction model, check it against production-facing designs, and identify six findings: recovery semantics are heterogeneous; recovery shifts trust; liveness improvements create abuse paths; post-recovery lifecycle management is uneven; protocol evidence outpaces user evidence; and recovery metadata remains underprotected. These gaps motivate a research agenda for recovery-aware financial technologies.

Content provenance standards, such as C2PA, are increasingly used to attach signed records of origin, editing history, and rights to digital images. However, provenance transparency can conflict with privacy – assertions that strengthen trust in an image may also reveal sensitive information about the creator or capture context. We propose soft redaction for image provenance: a mechanism that replaces sensitive provenance assertions with zero-knowledge proofs (ZKPs) of selected properties over hidden data. Our work focuses on distance proofs. We first show how location assertions can support proofs of proximity to a public reference point, using Chebyshev polynomial approximations within the ZKP proof circuit. We then extend the approach to L2 distance proofs over biometric embeddings, enabling privacy-preserving claims related to likeness to help enforce personality rights with images. Finally, we apply the same distance-proof construction to perceptual hashes (visual fingerprints), supporting an anti-spoofing use case in watermark-based recovery of stripped provenance metadata. Our results demonstrate that ZKPs over image provenance can provide practical soft-redaction capabilities, compatible with C2PA, that may be constructed in seconds and verified in milliseconds.

Human-Oriented Binary Reverse Engineering (HOBRE) aims to transform decompiled pseudocode into a more human-friendly representation, thereby reducing the cognitive burden of reverse analysis and improving efficiency. However, reliably evaluating HOBRE outputs remains a fundamental challenge: human evaluation is costly, time-consuming, and difficult to scale, while existing automated metrics either require executable test cases and runtime environments that are often unavailable for real-world binaries, or rely on high-quality source code references that are typically inaccessible and fail to capture semantically equivalent but lexically diverse outputs. Although LLM-as-a-Judge paradigm is naturally well-suited to HOBRE evaluation, its effectiveness remains underexplored. This paper presents the first systematic investigation of the LLM-as-a-Judge paradigm for HOBRE across three representative tasks: function name recovery, binary code summarization, and decompilation optimization. We introduce BinJudgeBench, the first expert-annotated, reference-free evaluation benchmark based on multi-dimensional human judgment, where LLM-as-a-Judge achieves an average correlation of 63.20% with human judgment, outperforming traditional automated metrics at 35.04%. By analyzing judge configurations across backbone LLMs, prompting strategies, and decoding temperatures, we find that no “one-size-fits-all” configuration exists, as the optimal setup varies across tasks and individual samples. To address this, we propose BinJudge, which employs a lightweight routing mechanism to adaptively select the optimal judge configuration for each task and sample. BinJudge improves correlation with human experts by 4.5%-24.7% and reduces API cost to 0.06$\times$-0.84$\times$ of that of static best configurations, providing a scalable, cost-effective, and high-fidelity automated evaluation scheme for HOBRE.

Parental control systems are designed to protect minors online, but can inadvertently obstruct digital forensic investigations. When enabled, these systems restrict administrative privileges, disable debugging options, and alter data accessibility, complicating evidence acquisition and analysis. This study empirically examines the impact of Microsoft, Google, and Apple parental controls on forensic processes across fifteen Windows, Android, and iOS devices. Through controlled experiments, we evaluate their impact on evidence accessibility and identify forensically sound methods to overcome these limitations. The findings provide practical guidance for investigators and contribute to improving forensic readiness in environments governed by parental control systems.

Modern agent harnesses persist state across tasks and sessions through persistent carriers like memory, skills, tools, and shared artifacts. However, this capability creates delayed safety risks: attacker-influenced content can cross system boundaries and later affect the execution of a benign request. Existing benchmarks typically focus on a few carriers or harnesses, while end-to-end attack-success rates reveal little about how risks propagate. To this end, we present HarnessSafe, a benchmark comprising 328 executable cases across seven persistent-carrier families and evaluated on most mainstream agent harnesses. Each case is specified as a Persistent-Risk Lifecycle that traces attacker influence from its initial entry, through persistence across carriers and system boundaries, to a later benign trigger and an observable violation. We further introduce a multi-stage, trace-based evaluation that uses observable execution evidence to determine how far each attack chain progresses and where it is stopped. Experiments show that containment is carrier-specific and strongly depends on the harness-model configuration. Both the harness and model backend substantially shape containment outcomes, while attack success rates cannot reflect distinct lifecycle progression patterns.

Internet of Things (IoT) devices handle sensitive privacy-related information such as user audio, video, and authentication data, making it essential to detect vulnerabilities in their firmware. Decompilation, a key detection technique, has recently attracted attention because Large Language Models (LLMs) enable high readability and high recompilation success rates. However, because LLM outputs depend on probabilistic token prediction, they tend to prioritize syntactic correctness and may generate plausible-looking code that is semantically different from the original binary. Vulnerabilities often arise in details that are easily lost in this process, such as error-handling flows and boundary checks. Existing evaluation metrics focus mainly on passing test cases and cannot sufficiently identify code whose internal structure has been altered despite appearing behaviorally valid, so a metric that quantifies the internal structure of decompiled code from multiple perspectives is needed. We propose a nine-dimensional quality evaluation metric consisting of three categories: structural, behavioral, and semantic similarity. Targeting 318 programs from OpenWrt, an open-source router platform underlying many commercial routers, we generated 19,625 decompilation results using five methods (one rule-based and four LLM-based) and analyzed them statistically. The recompilation-success group achieved significantly higher overall scores than the failure group (Cohen's d=0.92); behavioral similarity showed d=0.96 and structural similarity d=0.69, demonstrating that these metrics are important predictors of decompilation quality. This study provides a statistical evaluation foundation for quantifying implementation defects in IoT devices and a framework that generalizes to quality evaluation of black-box generative models.

Impersonation is a fundamental security threat in face recognition systems (FRSs). While the security of FRSs has been challenged by various attack vectors, under realistic adversarial capabilities, e.g., a limited number of decision-only authentication trials and no internal system knowledge, most attack techniques become infeasible. As a result, impersonation by zero-effort impostors, characterized by false match rate (FMR), is commonly regarded as a standalone baseline. A few years ago, impersonation attacks based on MasterFaces emerged as a notable security threat that could break the barrier of the FMR-based baseline under such realistic constraints. However, they were believed not to yield impersonation above the standard FMR in modern FRSs, as discussed by multiple follow-up studies. In this paper, we demonstrate that even legitimate access to public commercial APIs allows an adversary to amplify impersonation rates through MasterFaces, resulting in a non-trivial impersonation attack beyond FMR on downstream applications built on top of these APIs. We observe that several real-world FRS deployments are implemented using commercial APIs, and that the backend service provider is publicly disclosed or trivially inferable. As a result, the adversary can purchase these pay-as-you-go API services without requiring any additional privilege over the target FRS. From this observation, we formalize the MasterFaces attack as a maximum coverage problem over the biometric representation space, which we call a NET, and show that the adversary can construct an API-tailored NET by leveraging the geometric structure of the representation space. We demonstrate that our attack amplifies the impersonation rates of several open-source and commercial API-based FRSs by up to 9.5$\times$ within at most 30 authentication trials, compared to those expected from the standard FMR.

Retrieval-augmented generation (RAG) is indispensable for enhancing large language models. However, RAGs are increasingly susceptible to poisoning attacks, in which adversarial documents are injected to manipulate generator outputs. Previous methods rely on output-side signals such as perplexity and consistency checks to detect such attacks. Nevertheless, our analysis reveals that deliberate attacks often induce false confidence, where poisoned outputs exhibit even lower perplexity than benign ones, rendering uncertainty-based detection ineffective. To address this challenge, we explore the internal dynamics of the generator and identify a distinctive signature termed Attention Collapse. Unlike the dispersed attention in benign generations, attacked generations exhibit a decrease in entropy as attention concentrates on poisoned documents. Building on these findings, we propose D-SCAN (Document-level Signal Collapse Analysis), a lightweight detection framework that monitors attention dynamics to identify attacked generations. Extensive experiments on multiple attack benchmarks demonstrate the effectiveness of our method. Moreover, D-SCAN can detect attacks even when they fail to alter the final answer. Code is available at https://github.com/yingtaoren/D-Scan.git.

Cooperative UAV-swarm defenses commonly cross-check GNSS positions against measured inter-drone geometry. We show that this relative-geometry channel has a structural blind spot: a common, slowly varying translation (a rigid-covert shift, RigidShift) preserves all pairwise distances and is therefore unobservable to any relative-only detector (a gauge-freedom argument). We validate this blindness on distance-verification and semidefinite-feasibility baselines, while explicitly distinguishing it from onboard inertial/GNSS monitors that can raise a bare alarm but cannot recover the swarm's true position. To quantify when an external reference restores observability, we derive the drift-dependent detection floor $2\gamma/(1-t_s/T)$ for a calibrated anchor-residual detector and empirically identify an additional detector-specific noise floor (measured slope 2.66 vs. predicted 2.67). We then present a centralized anchor-rooted recovery pipeline that reconstructs swarm geometry from inter-drone ranges, aligns it to a trusted-anchor subset with Byzantine-robust fitting, and recovers the absolute positions of non-anchored drones. A segmented estimator jointly estimates anchor drift, attack rate, and onset when no clean-epoch label is available. Across statistical simulations, ArduPilot software-in-the-loop experiments, and Gazebo experiments with rendered vision anchors, the method recovers the positions of non-anchored drones to a median error of 0.39 m (20 seeds) under approximately 10.1 m of GNSS drift, and to 7.1 cm (5 seeds) in the rendered-vision multi-SITL setting. We also characterize the explicit limits imposed by non-collinear anchor geometry, anchor coverage, $\tau\to0$ drift-attack aliasing, and majority anchor compromise. All evaluations are simulation-based and use no RF spoofing hardware or physical swarm.

Computer-use agents~(CUAs) have transformed large language models into persistent execution systems capable of generating, storing, and reusing artifacts like skills and memory entries. However, existing security defenses largely treat attacks as externally triggered or temporally bounded, leaving a critical gap in addressing how compromise can propagate internally through an agent's own persistent state. We reveal that malicious influence can be covertly embedded into the structural redundancies of autonomously synthesized artifacts, allowing it to survive internal state updates and bypass standard vetting mechanisms. To formalize this threat, we introduce SynChain, a self-synthesized attack paradigm utilizing persistence-aware directed supervised fine-tuning to induce agents to create poisoned yet benign-looking artifacts. To systematically evaluate this propagation, we construct CUAChain, a dataset comprising 30 benign task chains and three attack objectives. SynChain enables dormant payloads to seamlessly reactivate in future workflows as trusted context, operating entirely without new malicious exogenous inputs. Extensive experiments on OpenClaw, Codex, and Claude Code under four defense settings demonstrate that SynChain achieves high attack success and outperforms adapted baselines, proving that securing CUAs requires provenance-aware reasoning over cross-task execution trajectories.

Large language models (LLMs) are widely used for code generation, yet they can reproduce vulnerable implementations learned from insecure training patterns. Prior work has mainly explored inference-time hardening, which reduces insecure generations without modifying the target model but relies on auxiliary components and adds runtime overhead. We conduct the first systematic study of model editing as a model-level hardening mechanism for secure code generation. We evaluate 3 state-of-the-art editing methods across diverse LLM families and compare them with CoSec, a representative inference-time approach, focusing on security, robustness, generalization, and functional correctness. Model editing yields larger security gains than CoSec on seen vulnerability types, improving security ratios by 15%-25% over vanilla models, with gains remaining stable under prompt perturbations. However, these improvements transfer unreliably to unseen vulnerabilities and can reduce functional correctness. To mitigate this trade-off, we propose SafeEdit, a post-edit refinement method combining functional tuning with edit-aware regularization. Across eight target LLMs, SafeEdit improves Pass@1 over UltraEdit by 11.73/13.70/15.50 percentage points at T=0.1/0.4/0.8 while largely preserving security. Compared with CoSec, it achieves relative security-ratio gains of 7.54%-12.04%. Additional evaluation on CodeGuard+ confirms improved joint secure-and-correct generation. SafeEdit and CoSec are also complementary, and their combination can further improve security while maintaining strong functional correctness. Overall, our results provide evidence-backed guidance for applying model editing to secure code generation.

At ASIACRYPT 2021, Baksi et al. introduced DEFAULT, a block cipher designed to algorithmically resist Differential Fault Attack (DFA), claiming 64-bit DFA security regardless of the number of injected faults. At EUROCRYPT 2022, Nageler et al. demonstrated that DEFAULT's claimed DFA resistance can be broken by applying an information-combining technique. More recently, at ASIACRYPT 2024, Jana et al. improved DFA by searching for differential trails with a single solution. They showed that, for DEFAULT with a simple key schedule, injecting five faults at the fifth-to-last round reduces the key space to one, and for BAKSHEESH, injecting twelve faults at the third-to-last round achieves the same result. In this paper, we propose a new DFA framework that utilizes a Mixed-Integer Linear Programming (MILP) solver. This framework makes it possible to attack deeper rounds than previously achieved, reducing the number of fault injections required for key recovery. Furthermore, we present a method to determine the most efficient fault injection bit positions by systematically analyzing the input differences from all possible single bit-flip faults, thereby further reducing the required number of faults. This systematic analysis has the significant advantage of allowing us to theoretically calculate the required number of faults. Applying our framework, for DEFAULT, injecting three faults at the sixth-to-last round and two faults at the seventh- and eighth-to-last rounds reduces the key space to one.

Large Language Models (LLMs) are increasingly used as high-level planners in embodied multi-robot systems, enabling robots to interpret natural language instructions and coordinate executable actions. Yet, this growing reliance on LLM planners also raises security concerns. Prior work has focused mainly on individual robots, while communication risks in multi-robot collaboration remain insufficiently understood. Existing multi-robot studies are further limited to preliminary analysis under the Decentralized Multi-agent System (DMAS) architecture, so it remains unclear whether these risks persist across other common communication architectures and how attacker access settings shape their propagation. To fill this gap, we formulate two communication attacks corresponding to distinct attacker access settings: the External Entry Point Attack and the Privileged In-System Attack. We evaluate both attacks across DMAS, HMAS-1, and HMAS-2 using three LLMs and five embodied multi-robot tasks. Results show that unsafe information can turn into unsafe actions across all three architectures: DMAS reaches a 96.7% entry endorsement rate and a 100% post endorsement activation rate, HMAS-1 reaches a 97.8% unsafe action success rate, and HMAS-2 triggers 88.3% of task defined unsafe action slots. To mitigate risks from trusted information flow, we introduce the Claim Provenance and Verification (CPV) Gate, which verifies communicated claims before downstream reuse and reduces the violation rate from 70.0% to 36.6%.

In this paper, we present a comparative performance analysis of the POKE-based key exchange and SIDH variants. SIDH gained attention for its small key size and efficient performance, and has been selected as an alternate candidate in NIST PQC Round 4. However, following the key recovery attack by Castryck and Decru in 2022, SIDH was shown to be vulnerable to polynomial-time attacks on classical computers, undermining its security and removing it from consideration. Given that SIDH was regarded as the leading isogeny-based algorithm, several countermeasures, such as MSIDH, MD-SIDH, and bin/terSIDH have been proposed. However, these approaches still face performance limitations. Meanwhile, POKE, proposed by Basso and Maino at Eurocrypt 2025, is an isogeny-based public key encryption scheme that combines a SIDH-like protocol with higher-dimensional isogenies and has drawn attention for its efficient performance. In this work, we adapt POKE into a key exchange algorithm and benchmark it against M-SIDH, terSIDH, and CSIDH. Targeting NIST security level 1, POKE-based KEM is approximately 21.21 times faster than terSIDH and 64.97 times faster than CSIDH, showing that the POKE-based KEM is currently the most promising isogeny-based key exchange candidate.

While radio frequency fingerprint (RFF)-based wireless device authentication has been widely studied across different datasets and scenarios, there still lacks a fundamental theory to explain why and how RFF can serve as a reliable device identity, significantly hindering the practical application of such an authentication technology. In this article, we integrate the RFF modeling with authentication property analysis to propose a general theoretical framework to facilitate the development of such a theory. The RFF modeling process reveals how RFFs are induced, evolved and observed along the transmitter-channel-receiver chain, built upon which, the authentication property analysis process then outlines how the trustworthiness of an RFF should be examined in terms of its uniqueness, stability, distinguishability, and unforgeability. By linking the RFF formation/evolution to these authentication properties, the framework offers a solid foundation for understanding why and how RFF-based authentication is trustworthy in practice. We also discuss the communication-authentication co-design issue based on the theoretical insights from the proposed framework.

Low-rank adaptation (LoRA) enables efficient specialization and distribution of large language models through compact adapters. However, untrusted adapters introduce a supply-chain threat: a backdoored adapter can cause a model to generate harmful content, malicious code, political propaganda, or covert advertisements when an input contains a hidden trigger. Adapter-agnostic defenses merge the adapter with the base model, which dilutes backdoor signals and reduces detection performance. Existing adapter-aware methods do not address how to safely use a potentially backdoored adapter. Instead, they either train a defensive adapter to repair a backdoored base model, addressing the inverse problem rather than securing the adapter itself, or rely on a classifier that flags the entire adapter as suspicious and requires separate mitigation. These methods overlook the distinct latent-space signatures produced by trigger-bearing inputs in backdoored adapters. We introduce LoRAScan, the first adapter-aware defense that detects and rejects trigger-bearing inputs at inference time without modifying adapter parameters. Our key observation is that a small subset of LoRA insertion sites, approximately 5%, remains stable across clean inputs but exhibits highly concentrated spikes in LoRA down-projection activations when a trigger is present. LoRAScan identifies these low-variance insertion sites before model deployment and monitors them during inference. Across standard LLM backdoor benchmarks, LoRAScan rejects approximately 98.49 of malicious inputs with a small error rate on clean inputs, outperforming existing defenses across diverse evaluation settings.

Mapping cyber threat intelligence (CTI) text to MITRE ATT&CK techniques is essential for structured threat analysis, yet manual annotation is costly and does not scale. The ATT&CK taxonomy comprises several hundred attack techniques, and a single CTI passage may describe multiple techniques, making accurate and complete extraction challenging. Existing automated approaches fall short in different ways: multi-label classifiers struggle with severe class imbalance and the large label space, while LLM-based methods–retrieval pipelines and fine-tuned generators–optimize token-level objectives that treat technique annotation as sequence generation rather than set prediction, lacking direct supervision on whether the predicted technique set is correct and complete. We propose TTP-R1, a two-stage framework that combines retrieval-augmented supervised fine-tuning (SFT) with reinforcement learning using verifiable rewards (RLVR). A hybrid retriever first narrows the large label space to a candidate set, and a fine-tuned LLM learns to select the correct techniques. We then apply Group Relative Policy Optimization with a decomposed reward that directly supervises the precision, recall, and output format of the predicted technique set. Across four CTI benchmarks, TTP-R1 achieves the best average F1, improving sub-technique-level F1 by 7.4 percentage points over Claude Sonnet 4.5 with retrieval augmentation, while running 28x faster when served as an 8B-parameter model on a single GPU.

This paper studies the concrete security of BBS signatures (Boneh, Boyen, Shacham, CRYPTO '04; Camenisch and Lysyanskaya, CRYPTO '04), a popular algebraic construction of digital signatures which underlies practical privacy-preserving authentication systems and is undergoing standardization by the W3C and IRTF. Schäge (Journal of Cryptology '15) gave a tight standard-model security proof under the q-SDH assumption for a less efficient variant of the scheme, called BBS+–here, q is the number of issued signatures. In contrast, the security proof for BBS (Tessaro and Zhu, EUROCRYPT '23), also under the q-SDH assumption, is not tight. Nonetheless, this recent proof shifted both standardization and industry adoption towards the more efficient BBS, instead of BBS+, and for this reason, it is important to understand whether this tightness gap is inherent. Recent cryptanalysis by Chairattana-Apirom and Tessaro (ASIACRYPT '25) also shows that a tight reduction to q-SDH is the best we can hope for. This paper closes this gap in two different ways. On the positive end, we show a novel tight reduction for BBS in the case where each message is signed at most once–this case covers in particular the common practical use case which derandomizes signing. On the negative end, we use a meta-reduction argument to prove that if we allow generating multiple signatures for the same message, then no algebraic reduction to q-SDH (and its variants) can be tight.

Most language-model access controls regulate behavior while leaving the same computation available to every request. We study a different systems question: can trusted authorization determine which newly trained parameters are reachable by the forward pass? Policy-Masked Private Experts freezes a pretrained sparse Mixture-of-Experts (MoE) model, trains a disjoint expert branch, and selects the public or private pool before top-k routing. The resulting claim is narrow but testable: under the declared trusted computing base (TCB), an unauthorized request executes no private expert. It does not imply that the public model lacks the same semantic capability. We test this separation between execution control and task utility in Qwen3-30B-A3B and DeepSeek-V2-Lite. Three Qwen BF16 seeds update all 32 private experts while the public fingerprint remains unchanged. Across 64 adversarial scenarios and 96 deny/fail-closed events, unauthorized private execution is zero; independent hooks exactly match 11,616 routed private rows and allow-deny-allow recovery is exact. On two prospectively frozen Qwen benchmarks, the private branch improves exact tool use by 5.0 percentage points (pp) (five versus zero discordances; one-sided Holm p = 0.03125, corresponding two-sided exact p = 0.0625) and 21.3 pp (percentile-bootstrap 95% CI [13.3, 29.3], Holm p = 0.000031). Three arm-blinded model evaluators retain a positive external effect of 18.7 pp (95% CI [9.3, 28.0]). A parameter-matched Lora has similar external utility, but a post-hoc request gate leaves 1,225 adapter calls under deny; the disjoint expert branch leaves none. DeepSeek reproduces the route invariant and gains 27.0 pp. A valid sealed evaluation is near-neutral. These results support auditable, reversible control over a trained parameter path, while showing that useful transfer remains distribution dependent.