Papers

LLM pipelines increasingly ingest Office Open XML (OOXML) documents (Word, Excel, and PowerPoint files) as first-class evidence in financial, compliance, and retrieval-augmented workflows, implicitly assuming semantic integrity: that the evidence consumed by the model matches the content shown in the Microsoft Office suite editing canvas. We show that this assumption can fail in OOXML-to-LLM pipelines. The same specification-valid OOXML file can yield one evidentiary view in Microsoft Office and another when extracted for an LLM. Each view is treated as authoritative by its consumer, a condition we call plural ground truth. The ingestion contract rarely states which view and semantic roles become model evidence or preserves how that evidence was derived. We call the specification-grounded OOXML constructions that induce such divergence evidence forks. We systematically traverse and mine the OOXML specification and confirm 21 evidence forks across Excel, Word, and PowerPoint, spanning six dimensions of view construction. All 13 tools in our extraction panel emit evidence from at least one fork. We test four native-ingestion LLM APIs and seven web chatbots. Each test document carries a trap: a task-relevant fact exposed by extraction but not shown in Office. Across this 21-mechanism evaluation, the four APIs return the trap in 48–76% of trials. For 20 of 21 mechanisms, at least one of the eleven interfaces returns the trap. Our measurements further show that exposure is shaped upstream of the model by the ingestion path and extractor configuration. A source-level survey of sixteen popular open-source LLM projects further shows that default OOXML ingestion paths concentrate on affected extractor families.

With the continuous development of the WeChat ecosystem, WeChat Mini Programs, due to their advantages of not requiring installation, using little memory, and being ready to use instantly, have seen a surge in user numbers and have now become an indispensable service carrier in mobile internet. However, as Mini Programs rapidly became popular, issues regarding the compliance of their interface interaction design and the safety of operational behavior have become increasingly apparent. Many Mini Programs have problems such as clickable buttons and icons not being standard in size, or ad pop-ups and payment entrances being placed in a way that is easy to misclick. The close or cancel buttons are often too small or hidden, making it easy to accidentally click on ads or payment content, and difficult to accurately click the cancel button. This can result in involuntary payments or being redirected to illegal pages, causing unnecessary financial losses and seriously harming users' property security and legal rights. To address the above issues, this article develops a detection program to check the position and size of various icons and buttons in Mini Programs, and analyze whether redirected links fall within a safe range. YOLOv8 is used to identify various buttons in images, displaying the corresponding icon and its data based on the mouse click position. Violations are flagged and recorded. At the same time, mitmproxy is used to capture relevant data requests generated during clicks, analyzing the safety of redirections, and presenting key information for user observation.

A coding agent edits files and executes shell commands with its developer's privileges, allowing malicious requests to translate directly into harmful actions or functional malware. Existing defenses have complementary limitations: weight-level alignment is unavailable to API-only deployers, whereas input filters and execution-boundary monitors require auxiliary classification or checking components along the agent's trajectory. We therefore introduce SkillShield, a system-prompt defense that synthesizes security skills offline from known attacks or recorded agent failures. These skills are injected into the system prompt at session start and remain active throughout the tool-use loop. Unlike a reference monitor, they protect the system by defining the security policies the model should follow during execution. Due to the limited system-prompt space, we examine three fixed-budget provisioning scopes: all-classes, with one skill covering all threat classes, per-bundle, with one skill targeting a related subset, and per-class, with one skill dedicated to a single known class and used as the upper-bound reference. None requires runtime request classification or routing. Across six large language models on RedCode, the default all-classes skill reduces malware-generation severity from 3.37 to 0.58 and achieves a 43.6% execution attack success rate, comparable to Llama Guard 3's 42.7% without its separate 8B classifier. The per-bundle and class-fixed per-class settings further reduce this rate to 36.2% and 14.5%, respectively. Under two non-adaptive jailbreak families, SkillShield continues to outperform all baselines on malware generation. Across 731 benign task descriptions, SkillShield yields a mean safety-refusal rate of 0.14%. These results demonstrate the potential of prompt-space security skills to prevent harmful actions and malware generation for LLM coding agents.

Containerization dominates enterprise software delivery, yet Dockerfiles that assemble container images frequently harbor security misconfigurations and structural technical debt. This problem is poorly understood in corporate inner-source environments, where proprietary context and isolated governance prevent direct application of open-source findings. We present an automated, six-stage pipeline that: (1) crawls an enterprise GitLab instance, (2) enriches each Dockerfile with static security and quality metrics (Hadolint, ShellCheck, Trivy) and lifecycle data, (3) groups functionally identical workloads using LLM-generated semantic descriptions and HDBSCAN, and (4) quantifies the optimization gap against cluster-internal reference implementations. Applied to 11,470 Dockerfiles from over 6,200 repositories at a single large industrial company, we find a systemic deficit: 99% of files contain at least one security misconfiguration, 80.8% violate Dockerfile best practices, and the median artifact has not been revised for 838~days. Despite this, high-quality reference implementations already exist within 83% of functional clusters. Adopting these internal standards would increase the average security posture score by 60.4% without developing any new templates. These findings, grounded in one organization's inner-source ecosystem, provide a data-driven foundation for future automated, context-aware recommender systems targeting enterprise supply-chain security; whether the observed technical-debt distribution and optimization gap generalize to other enterprises remains an open question for future multi-organization study.

Self-evolving LLM coding agents write their own tools by imitating retrieved skills from shared skill libraries. We identify a vulnerability in this loop: during authoring, a retrieved malicious skill can become the template for a new skill that preserves the payload. We call this self-poisoning: the agent authors, stores, and runs the resulting malicious skill. We exploit it through EvoMal, an attack that amplifies self-poisoning by wrapping an interchangeable payload in a banner, a set of benign-looking structural elements that induces an imitating agent to reproduce the enclosed code. The attacker plants malicious skills in the library without invoking them. The agent then authors and executes new skills carrying the harmful code. Each authored copy can re-enter the library and be imitated again, forming a self-propagating worm that persists after the planted skills are removed. We define the agent self-poisoning rate (ASPR) as the fraction of tasks that add a newly authored malicious skill to the library. Across six models on 153 tool-relevant SWE-bench Verified tasks, ASPR ranges from 20.3% to 41.8%, and the poisoned libraries hold 4.9 to 9.0 times as many malicious skills as were planted. The vulnerability also appears without a banner: DeepSeek-V4-Pro reaches 11.1% ASPR with the payload alone. Tailoring the planted skill descriptions to one task family raises ASPR to 86.7%. After the planted skills are removed, Qwen3 retains a round-5 ASPR of 68% because agent-authored copies remain. These copies evade existing defenses, which focus on attacker-submitted names, code, and signatures. We propose counter-prompt, a defense that discourages banner-style copying and reduces EvoMal's ASPR to at most 6.7% with no significant task-completion loss.

Face-swapping has emerged as a promising approach to facial privacy protection, replacing a target individual's appearance with that of a donor while preserving non-facial context. The resulting images visually resemble the donor, and face recognition systems tend to suppress the target's match scores – ostensibly satisfying privacy requirements. Empirical evaluation across a range of face-swapping models, however, reveals that significant target identity leakage still occurs. This raises a deeper question: why does leakage occur, and can it be predicted? We propose a linear stochastic model that treats face-swappers as transformations on the space of identity embeddings, providing an interpretable account of the leakage mechanism. The model is fit to empirical observations and used to derive testable predictions. The aim is to ground privacy assessments in principled, interpretable analysis, thus making formal privacy guarantees explainable – and perfectible – rather than purely observational.

Spiking Neural Networks (SNNs) are increasingly distributed as pretrained checkpoints and reused as backbones for new tasks. However, current SNN watermarks are mainly verified against the model output. Thus, a user who replaces the output head can keep most of the original network while removing the evidence used for verification. We present MeMark, a watermark designed for the checkpoint-reuse setting. Instead of storing the watermark in the output head, MeMark embeds a multi-bit identifier in the internal membrane state of selected Leaky Integrate-and-Fire (LIF) neurons. A secret input drives each selected neuron to the chosen side of its own firing threshold, and the same threshold is later used to recover the secret bit, so the verifier does not need a learned decoder. We evaluate MeMark across recurrent, convolutional, residual, and transformer SNNs. On a 215.4M-parameter SpikeGPT checkpoint, all 20 independent 64-bit keys pass the fixed 51/64 verification rule, while none of the $30\,000$ fresh random keys pass when tested against all 20 protected checkpoints and the clean model. All 20 genuine keys also remain above the threshold after fine-tuning, 90% pruning, int8 quantization, and output-head replacement. Under our stated threat model, adaptive attacks can weaken the watermark but do not remove the ownership evidence in the settings we test. Additionally, we study false ownership claims, key-aware and key-agnostic removal, partial key disclosure, rollback, and extraction into a student. The results show that MeMark can provide evidence of checkpoint derivatives, while being resistant to the adversary's attacks and complete head replacement.

Hosted retrieval-augmented generation (RAG) and semantic search allow users to query valuable provider-held corpora, raising two competing demands: to hide each query and chosen result, yet reveal only the documents that the user is authorized to receive. Existing cryptographic approaches either make this costly by processing the entire corpus for every query, or sacrifice quality for efficiency by scanning a few clusters. We repurpose learned deep hashing as a private filter: a randomized binary code points the provider to a short candidate list, while encrypted reranking and oblivious key transfer protect the precise query and final selection. This shortlist short-circuits full-corpus cryptographic search without sacrificing retrieval quality: with 200-500 candidates, it closely matches full-corpus retrieval across five zero-shot corpora spanning 25K to 5.4M documents. On the full 2.68M-passage NQ corpus over a 10-Gbps link, our protocol only adds 0.73 seconds, or 10 percent, to a 128-token Qwen3-32B RAG pipeline. The released code satisfies directional metric differential privacy (DP) and substantially reduces embedding-inversion and property-inference leakage, demonstrating that a carefully learned shortlist can make private dense retrieval both accurate and practical.

Automated code reviewers are increasingly used as gates on pull requests (PRs), yet evaluations measure whether they block a malicious change. A block may be triggered by an unrelated issue rather than the vulnerability that makes the PR unsafe; fixing the reported issue can leave the target defect exploitable. We call this discrepancy the Verdict-Diagnosis (VD) gap. We present MalPR-Bench, a mechanism-grounded benchmark of 89 malicious PRs and 50 paired benign controls across 44 repositories and eight language families. Each malicious case has a pre-committed rubric specifying the target vulnerability, accepted mechanism descriptions, required repository evidence, and off-target findings receiving no credit. Reviews are scored separately for verdict correctness, target-vulnerability identification, and evidence validation; an attributable block requires all three. We introduce PRGuard, an attributable PR security reviewer that constructs candidate vulnerabilities and validates their premises against repository evidence using deterministic, non-executing tools and bounded retrieval. Across 31 common-coverage held-out malicious PRs, PRGuard and CodeRabbit produce similar blocking totals (22/31 vs. 24/31), but PRGuard identifies 22 target vulnerabilities versus 16 for CodeRabbit, a 1.38x difference. On 14 absence-type cases, both block 9, while PRGuard identifies 9 targets versus 3. CodeRabbit identifies 16/24 targets when required evidence lies within touched files and 0/7 when validation requires evidence outside them. Finally, PRGuard uncovers twelve previously undisclosed, proof-of-concept-backed vulnerabilities across five projects. PRGuard/DeepSeek and CodeRabbit both block 10/12 discovery PRs, but produce 10/12 and 4/12 attributable blocks, respectively. Thus, verdict-only evaluation can substantially overstate the security value of automated review.

Large language models (LLMs) have recently advanced graph neural networks (GNNs) by enriching node representations with semantic information, giving rise to LLM-enhanced GNNs that achieve substantial performance gains. However, how such semantic enhancement affects privacy risks remains largely underexplored. To bridge this gap, we systematically audit the privacy risks of LLM-enhanced GNNs through a unified framework consisting of five stages: (1) dataset preparation, (2) victim model training, (3) privacy attack, (4) risk assessment, and (5) defense analysis. Specifically, our evaluation spans ten text-attributed graph datasets across diverse domains, six privacy attacks, 42 LLM-enhanced GNN configurations, and three more recent language-model backbones. Extensive experiments show that, despite their utility improvements, LLM-enhanced GNNs consistently exhibit greater empirical privacy vulnerability than shallow text representation baselines under the evaluated attacks across diverse models and datasets. Further analysis shows that LLM-enhanced representations exhibit more distinguishable link-, label-, and membership-related signals in the embedding space, making them more exploitable by inference attacks. Finally, we evaluate representative defenses and examine their effectiveness in mitigating these privacy risks. Overall, this work provides a systematic audit of privacy risks in LLM-enhanced GNNs and offers insights for developing more secure and trustworthy graph learning systems.

Tool-using LLM agents extend security risks beyond generated text to actions that affect external systems. Under multi-turn decomposition attacks, a harmful objective can be distributed across individually plausible requests and tool calls, becoming apparent only from the accumulated trajectory. Existing defenses either rely on auxiliary online reasoning to recover long-horizon security evidence or assess actions after generation, often incurring additional inference cost or depending on runtime-specific action representations. We propose Reassembling Distributed Risk (ReDiR), a generation-time defense that conditions action generation on trajectory-level security evidence. Before each action, ReDiR compresses the current trajectory into a compact latent safety representation and injects it into the frozen base model. The representation is learned through same-model, cross-view supervision, where safe behavior from an explicit task view provides supervision for recovering distributed safety evidence from the original multi-turn trajectory. This design enables ReDiR to integrate cross-turn security information directly within the generation process without relying on a separate action-level safety module. We evaluate ReDiR on two agent-safety benchmarks across three model families and eight held-out tool domains. ReDiR reduces attack success rates to below 8%, transfers to unseen tool domains, and preserves benign fidelity with low computational overhead.

Large language models (LLMs) are increasingly deployed with layered defenses, yet malicious prompts can still bypass them. Interpretability methods can expose model-internal signals along the generation path that could inform enforcement, but these signals are not security controls by themselves. Deployments that adapt them for safety typically couple each signal to its own calibration, policy logic, and intervention code, so each new artifact creates integration work instead of strengthening a shared defense. We present Language Model Security Modules (LMSM), a security framework that adapts the separation behind Linux Security Modules (LSM) to LLM serving. In LMSM, a selected security backend exposes calibrated evidence, a versioned policy evaluates active rules over trusted per-request context, and a separate gate authorizes buffered output release. This design separates mediation correctness from policy effectiveness, and it allows backend, rule, or schedule changes without rebuilding request handling or enforcement. Our prototype shows the separation working in practice: with Hugging Face Transformers and continuously batched vLLM, the same substrate hosts artifact-backed sparse autoencoder (SAE) and transcoder deployments and task-fitted dense probes, preserves request-specific decisions under scheduler churn, and selectively enforces and composes multiple rules per request. On Qwen3-4B, LMSM-Checkpoint reduces HarmBench attack success rate from 39.20% to 3.32%, with XSTest false refusals rising from 2.40% to 4.40%, while retaining 98.14% of the throughput of a matched serving path that performs no monitoring work at 32 active sequences. LMSM gives advances in interpretability and model-internal analysis a common path to runtime enforcement.

Predictive models are widely used in many fields, but are vulnerable to false data injection attacks. To address this, detection schemes and adversarial training have been proposed, but such approaches lack guarantees against stealthy attacks. We therefore propose a detector-based switched model, in which optimal attack strategies are stealthy. For linear prediction models, we derive a convex formulation of the resulting adversarial risk. The model incorporates protected features and introduces a hyperparameter modelling attack probability, enabling an explicit performance trade-off between clean and attacked data regimes. Numerical simulations on real and synthetic data show improved performance on partially attacked data, even for misspecified attack probabilities.

Phishing is a well-known social engineering (SE) type used to trick individuals into revealing personal information or performing desired actions, like downloading and installing malware. Other SE types, like vishing and smishing, have emerged and are increasingly being used. As SE continues to successfully persuade victims into actions, the questions arise of which SE attack types are most effective for specific psychological factors (PFs) and, conversely, which PFs are most effective for particular attack types. To answer these questions, we conducted a laboratory study with n=12 participants, in which each participant was shown all 25 stimuli (five PFs and five SE types). The results of this exploratory study show that the most effective SE attack type for authority, trust, and greed was spear-phishing. The most successful combination of PF and attack type was spear-phishing using greed. The least successful combinations were pop-ups using authority, smishing using authority, and vishing using curiosity, each having had no success at all.

The integration of Large Language Models (LLMs) into cybersecurity has transformed vulnerability assessment, but it has also produced a trustworthiness crisis driven by the unchecked proliferation of "AI slop." These artifacts, hallucinated vulnerabilities, plausible but incorrect patches, and semantically repackaged bug reports, impose a cognitive burden on human triage pipelines that mirrors a denial-of-service attack. This paper surveys the empirical evidence, identifies a unifying mechanism, and traces a path toward trustworthy triage. We formalize a taxonomy of AI slop grounded in a structured literature review and dissect its root cause: the gap between the causal deductive reasoning of security experts and the autoregressive probabilistic generation of current LLMs. We operationalize this gap through a measurable proxy, the Deductive Coverage Score, and show that chain-of-thought prompting and tool-using agents narrow but do not close it. We review mitigation strategies and argue that passive detection and watermarking target provenance rather than correctness, facing fundamental entropy constraints. We instead advocate for active neuro-symbolic verification, mapping each pipeline component to prior systems with documented limits on security inputs. Finally, we specify two evaluation instruments, CVE-Bench and Slop-Score, including dataset construction, metric formulas, and anti-gaming provisions. By shifting evaluation from linguistic fluency to mathematical verifiability, this survey provides a roadmap for securing emerging AI-driven triage systems.

With the rapid evolution of the Decentralized Finance (DeFi) ecosystem, stablecoins have emerged as a critical infrastructure bridging the cryptocurrency market with traditional financial paradigms. However, stablecoin systems rely heavily on smart contracts to execute automated operations. The immutable nature of these systems post-deployment means that the exploitation of security vulnerabilities can lead to irreversible, massive economic losses and potentially trigger systemic financial risks. Current research on stablecoin smart contract security faces challenges such as a lack of domain-specific targeting and the obsolescence of static defense models. To address this, this paper systematically analyzes common attack vectors in stablecoin environments and proposes a practical, real-time dynamic defense architecture. By analyzing 12 real-world security incidents, we elucidate the underlying mechanisms of high-risk patterns such as reentrancy attacks, oracle manipulation, and composite flash loan attacks. Concurrently, we construct a real-time anomaly detection model utilizing multi-dimensional on-chain temporal features and the Bi-LSTM algorithm. Experimental results demonstrate that this model achieves a classification accuracy of 96.61%, with an average recall rate of 97.70% for malicious attack samples, and a single inference latency ranging from 1.5 to 2.8 milliseconds.

Multimodal Large Language Models (MLLMs) are increasingly deployed in real-world applications, yet how different factors shape their jailbreak vulnerabilities remains poorly understood. Existing benchmarks often couple harmful intent, prompt framing, visual semantics, and instruction carrier within individual jailbreak instances, obscuring the specific sources of observed vulnerabilities. To address this limitation, we introduce MMJailBench, a factorized benchmark that systematically varies and combines these factors under controlled configurations, enabling fine-grained comparison and factor-level attribution. Large-scale evaluations across 16 open-weight and proprietary MLLMs reveal highly heterogeneous and model-dependent vulnerability profiles. Jailbreak vulnerability varies markedly across harm domains, exposing uneven coverage in current multimodal safety alignment. Prompt framing emerges as the dominant source of variation, task-relevant visual semantics systematically increase jailbreak susceptibility with authority-like cues exposing particularly pronounced vulnerabilities, and visually rendered instructions do not consistently increase jailbreak susceptibility relative to direct textual instructions. To further investigate the risks introduced by multimodal context, we conduct diagnostic analyses on a representative open-weight model and identify vulnerability-associated patterns in internal representations and cross-modal interactions. Finally, we develop a modular multimodal jailbreak evaluation suite with full and lightweight configurations, multiple judge options, and multidimensional metrics, enabling reproducible, scalable, and cost-efficient multimodal jailbreak auditing.

A system that authorizes an action must see enough of it to decide, and a system that attests to its decision must record enough to be audited. Both pressures push raw action parameters – recipients, payment memos, record identifiers – into an append-only ledger that cannot delete them. We show the two are separable. We classify each parameter field, not each action class, into three tiers: fields a policy may legitimately match on, which cross raw; fields that are policy-relevant but identifying, which cross only as projections such as an email domain or a templated route shape; and fields with no legitimate policy use, which never leave the workload. The central property is that the ledger's commitment is a canonical digest of the full, unminimized parameters, computed before minimization runs. The commitment is therefore independent of the tier table: reclassifying a field changes what is disclosed without invalidating a historical entry, reopening a hash, or altering what an offline verifier checks. Tier table, policy schema and wire schema are generated from one per-action declaration, so the deciding and recording parties cannot hold different rules. We then address a question the architecture forces: which party should compute each attested fact? We argue it is settled by which party could lie about it undetectably, and derive three answers within one request – the client computes the parameter digest, being the only party holding the data; it is structurally prevented from naming the definition that governed it, since that would write a false statement into a signed ledger; and it attests which tier table it applied, so divergence is detectable. We give a leakage analysis of each projection, report an incident in which a first-cut projection preserved the identifier it was written to remove, and state the residual trust the design does not eliminate.

Despite their strong ability to generate code, large language models often fail to produce secure code, as their outputs frequently contain security vulnerabilities. Secure code generation is inherently challenging because it requires solving a multi-objective problem: functional correctness and security. Existing approaches address this challenge by injecting external security knowledge or by using agentic feedback and iterative refinement. However, guideline retrieval often leaves the generator to translate generic advice into task-specific secure implementations, while shared-dialogue multi-agent feedback can blur role boundaries and suffer from context bloat. We present MACGen, a multi-agent framework that integrates planning, security analysis, code synthesis and refinement to jointly optimize security and functionality. A planner constructs a step-by-step plan to satisfy functional requirements. A security advisor identifies likely CWEs and synthesizes task-specific guidelines, a coder then generates code grounded in these artifacts, and a reviewer issues perspective-separated feedback. Rather than sharing full dialogue histories, each agent receives only structured artifacts from upstream stages, enforcing role specialization and reducing uncontrolled context growth. On CWEval and BaxBench, MACGen improves F&S@1 over direct prompting by 19.61 and 10.57 percentage points (pp) on average, respectively.

LLM serving frameworks process large volumes of user data–often containing sensitive information–on shared infrastructure. Ensuring isolation between users who share the same serving framework (on CPUs) and LLM operators (on GPUs) is critical for privacy protection. This paper presents GIFT, a GPU Information Flow Tracking system that enforces user data isolation in LLM serving with minimal overhead. Moreover, the design of GIFT is non-intrusive and allows CPU-side serving frameworks to evolve freely. It rests on two key insights. First, encryption-as-isolation leverages the observation that CPU components only orchestrate data flow, not content manipulation; thus, per-user encryption can provide isolation without modifying serving logic. Second, GPU kernels exhibit limited and predictable information flows, enabling static flow analysis. GIFT precomputes information flow rules for each kernel and uses decoupled flow tracking, avoiding instrumentation or GPU stalls. Furthermore, we extend GIFT to GIFT-CC, which integrates confidential computing to protect against untrusted operating systems and hypervisors (LLM service providers). Implemented on vLLM and DistServe, GIFT and GIFT-CC enforce user data isolation with a 4-10.7% throughput overhead while maintaining the same latency level.

Refusal training protects AI models from jailbreaks by training models to decline unsafe queries, reducing the risk of misuse. Recent work finds that refusal behavior in aligned language models can be mediated by a single activation direction or a low-dimensional refusal subspace shared across harmful prompts: ablating those directions suppresses refusals while largely preserves other model capabilities. Yet it remains unclear why safety-critical features in a wide range of models emerge in a concentrated, low-dimensional structure. In a case study of OLMo-2-0425-1B-Instruct we find that the refusal geometry reflects refusal training: activation updates resulting from refusal-completion first-token losses explain the resulting refusal direction and refusal subspace. We study refusal directions through the training dynamics across refusal datasets and reveal that their brittleness is associated with repetitive refusal starts, which in turn is linked to concentration of gradients and refusal features in a low-dimensional subspace. Across frozen-model analyses and controlled synthetic fine-tuning, we find evidence of a hardening lever: diverse refusal starts can raise stable ranks of gradients and activation changes, making refusals harder to remove with a vector ablation attack.

Mixture-of-Experts (MoE) has become a prevalent paradigm for scaling Vision Transformers efficiently. To ensure computational scalability and prevent expert overload, Vision MoE architectures employ a capacity-bounded token dispatch mechanism, where each expert's processing budget depends on the inference batch size. This work identifies this batch-dependent behavior as an overlooked attack surface, and proposes a stealthy supply-chain backdoor attack that exploits this property through a three-phase framework. First, we inject a backdoor into an early MoE layer. Second, we train a neutralizer in a deeper MoE layer that suppresses the backdoor under normal capacity. Third, we configure a batch-adaptive capacity factor that preserves high capacity for small batches while reducing it for large batches, naturally disabling the neutralizer via token overflow at deployment-scale batch sizes. The attack remains in dormant mode during small-batch security audits and enters activation mode during large-batch deployment. Experiments on V-MoE and Swin-MoE across ImageNet-100 and GTSRB demonstrate activation-mode attack success rates of 76-87% with dormant-mode ASR below 9%, while evading Neural Cleanse, STRIP, Fine-Pruning, and Activation Clustering. Our findings reveal a fundamental security risk arising from batch-dependent execution in scalable Vision MoE architectures.

The move of LLM inference to edge AI accelerators introduces new physical vulnerabilities. During execution, model parameters and intermediate inference states are repeatedly loaded into and processed on the chip, making them suscep- tible to physical side-channel attacks. In this work, by deploying laser voltage imaging, we show that one can extract LLM assets during inference, namely embeddings, attention, and quantized MLP weights, activations, and other inference states, from localized memories and compute subcircuits. To validate our claims, we perform an attack on an FPGA-based LLM accelerator. Since such accelerators reuse the same buffers and compute subcircuits across addresses, tiles, modules, and layers, reading asset values comes down to probing different memories during inference. We demonstrate full recovery of the targeted values; however, we also establish a methodology to recover asset values even if some weights or bits remain unread. We further derive lower bounds that relate imaging effort to asset dimensions and show that even direct recovery scales linearly with the size of the targeted asset

Proof of reserves (PoR) can improve transparency for tokenized assets, but aggregate reserve coverage does not establish whether off-chain assets are legally eligible, unencumbered, consistently valued, or sufficiently liquid for redemptions. We propose RWA-PoB, a credential-based proof-of-backing framework for tokenized U.S. Treasury products. Five authorised institutional roles approve a canonical EIP-712 snapshot containing reserve, liability, liquidity, and policy information. The framework evaluates backing adequacy through the Backing Coverage Ratio (BCR) and short-term redemption capacity through the Redemption Liquidity Coverage (RLC). The Solidity prototype couples the policy controller to an ERC-20 token. Successful issuance atomically increases token supply and recorded liabilities by the corresponding USD-denominated liability. A redemption request burns tokens while reclassifying the corresponding obligation as pending. The obligation is reduced only after an authorised settlement-role account confirms payment. We evaluate the framework against a simplified aggregate PoR baseline using USDY-calibrated liabilities and deterministic synthetic reserve scenarios. Both approaches permit issuance in the valid state, but RWA-PoB rejects an encumbered-assets state with a BCR of 96.3408%, below the experimental 105% threshold. Under liquidity stress, it classifies the proposed redemption as queued because the post-request RLC falls to 39.9999%. RWA-PoB authenticates the attribution and integrity of institutional claims but does not independently prove the existence, ownership, or condition of off-chain assets. The prototype, test suite, datasets, and replication scripts are available at https://github.com/rischanlab/PoB.

Real conversations between fraudsters and their targets are among the most informative artifacts for studying telephone scams, yet also the scarcest: passive honeypots overwhelmingly capture automated messages and hang-ups, large-scale studies characterize call metadata rather than dialogue, and manual scam-baiting does not scale. We present a dataset of real scam-call conversations collected by an active voice-agent honeypot. Dedicated numbers are seeded into the lead-generation channels fraud operations harvest; inbound callers are answered by a low-latency conversational agent that adopts a plausible target persona and sustains the interaction while every call is recorded, transcribed, and automatically labeled. Over an initial 53-day window we captured 10,015 inbound scam and spam calls (6,601 with two or more turns): roughly 895 hours of audio and 328,869 transcribed turns from 5,665 distinct originating numbers. Under a holistic classifier the substantive calls are predominantly predatory-but-legal lead generation ("spam", about three in five), while about one in seven is an outright "scam" (949 in this snapshot). Each call carries a turn-level transcript, three-channel audio, per-turn latency telemetry, and layers of automatic labels, including a holistic scam/spam/legitimate judgment corroborated by independent human review (75% agreement on the binary decision). We describe the collection system, the record structure, and technical validation of the corpus's realism and label quality, including that the agent is recognized as non-human in only about 5% of engaged calls. We also benchmark established scam-detection methods, where detectors trained on published synthetic dialogue collapse in precision on real traffic.

Computer Use Agents (CUAs) are increasingly deployed to navigate mobile and desktop applications on behalf of users, yet no benchmark comprehensively evaluates whether they can safely interact with visual interfaces while handling ambiguous instructions. We introduce ADeptS-Bench, a dual-stream trustworthiness benchmark, grounded in the ADEPTS capability framework and general population user studies. The Safety stream provides paired benign/malicious tasks with threats embedded in the visual interface. The Disambiguation stream evaluates whether agents seek clarification when intent is ambiguous. Evaluating seven models reveals that no model consistently exceeds 80% task success while staying below 30% attack success; every model clicks "Checkout" on a $25K order without hesitation, and none detects that a "factory reset" button is mislabeled as "Optimize." An ablation reveals three distinct safety architectures: tool-dependent (ASR +21-23pp without refusal tool), partially tool-dependent (+10-11pp), and no mechanism (unchanged). In disambiguation, all models overestimate consequence severity, mirroring the over-refusal bias observed in safety. We release all data, evaluation code, and analysis tools upon publication.

Selection of trustworthy collaborators in distributed systems is critical for efficient task completion, necessitating the inference of trustworthiness from their past collaboration experience. However, as a collaborator serves distinct devices across diverse scenarios in past collaborations, its trust-related data, observed from different device-specific views, is inherently multi-source, heterogeneous, and uneven in quality. Consequently, achieving accurate trust evaluations for collaborator selection remains a major challenge. To tackle these issues, we propose a novel multi-view evidential learning (MVE) based trust evaluation method. First, to accommodate the multi-source heterogeneity of observed trust-related data, we model each task owner who has interacted with a potential collaborator as an independent observational view, enabling the evaluation of the collaborator's view-specific trust. Second, to address the dynamic evolution of trust under changing conditions, we leverage the powerful long-sequence modeling capability of the Mamba model to capture the deep temporal patterns of a collaborator's trust state within each view. Furthermore, to quantify the certainty levels of view-specific trust assessments, we incorporate an evidential deep learning mechanism in MVE, which outputs trust evaluation results while quantifying the subjective uncertainty underlying them. Finally, we employ a dynamic evidential fusion strategy to adaptively integrate the multi-view evidence based on their respective quantified uncertainties, thereby yielding a final trust evaluation for the collaborator. Extensive experiments demonstrate that the proposed MVE method outperforms baselines in both trust evaluation accuracy and task success rate.

Automotive electronic control units (ECUs) increasingly depend on hardware-rooted security to protect software integrity, authenticity, and lifecycle management in the presence of remote and physical threats. Hardware Security Modules (HSMs) have become a key building block in automotive system-on-chips (SoCs), providing isolated cryptographic services, secure key storage, and controlled execution under stringent real-time and cost constraints. This paper presents an architectural analysis of automotive HSMs and examines their role in establishing secure boot and hardware roots of trust. We first survey common HSM integration models used in production ECUs and discuss their flexibility and current automotive use cases. We then introduce realistic threat models to motivate hardware-backed security controls and analyze how HSM design choices influence secure boot chains of trust, secure storage, secure execution, and software signing mechanisms. Key tradeoffs between isolation, performance, updateability, and attack surface are discussed, with optional consideration of side-channel implications. The paper concludes by highlighting open challenges and future directions for scalable and resilient automotive hardware security. Finally, we discuss emerging challenges such as cryptographic agility and post-quantum readiness that are likely to shape the next generation of automotive HSM architectures.

We introduce the Huffman-Merkle Tree (HMT), an authenticated data structure (ADS) for dynamic workloads where items may differ in access frequencies, and access frequencies can change over time. An ADS allows proving item membership against a short commitment to a large mutable state, with applications including verifiable storage, Internet transparency services, and blockchains. Optimizing ADS performance under continuously changing access frequencies has not been fully addressed before, neither in theory nor in practice. HMT addresses dynamically changing access skew through two complementary mechanisms. The first is a Huffman-coding-based Merkle-tree layout, with a novel extension to support evolving access frequencies. The second is an elastic tiering regime that partitions items across separate trees, such as hot and cold tiers, with adaptive migration between them. The key insight in this approach is to place frequently accessed items closer to the root, while assigning less frequently accessed items to progressively larger and deeper trees. This reduces the overall frequency-weighted access cost. Our scheme is designed to scale to gigabytes of data spanning millions of items. To handle dynamism efficiently, layout updates are applied in batches, access frequencies are tracked using a count-min sketch, and the system employs a tier-promotion cache while exploring multiple tier-migration policies. We implement HMT and compare it on real-world data with Ethereum's Merkle Patricia Trie (MPT) ADS and its proposed replacement, the Unified Binary Tree (UBT). Our evaluation considers two metrics: the amount of hashing per update and access-weighted membership-proof size. The latter captures both item access cost and frequency. We find that the best HMT policy uses about 2.4x and 0.34x less average hash operations than MPT and UBT respectively, and has 0.18x and 0.55x shorter proofs.

BGP hijacking remains a persistent threat as existing defenses, including RPKI/ROV suffer from a fundamental incentive misalignment: the networks best positioned to filter malicious announcements bear operational costs but receive no direct benefit, while the victim prefix owner captures all the value. We advocate a market-based alternative in which prefix owners post standing bounties for filtering invalid announcements of their prefixes, turning filtering from altruism into a private transaction. Our insight is that neither a propagating hijack nor its absence can hide from public route collectors, whose committed routing tables could become an independent root of trust for releasing funds of the bounty. We build on this insight to design BGPay, an escrow protocol in which filterers and monitors commit before either reveals, and a smart contract pays out on evidence rather than on the prefix owner's judgment. Analyzing 1K real hijack incidents, we find that today's collectors already provide enough visibility where it matters: ASes that are more important for containing the hijack are also highly visible from the public monitors. Hence, setting rewards proportionately to containment impact discourages misbehavior.