Papers

With the increasing threat of malware across various software related domains, malware detection and classification is critical to determine the response actions. Different strategies have been adopted to address the challenge of malware detection. With the advent of deep learning techniques, malware detection using image processing has garnered research attention. In this work, we proposed a novel malware binary to image transformation technique HilEnT based on a combination of Hilbert curve-based transformation of malware binary and the entropy feature comparison of malware file with benign and malware classes. Three grayscale images produced during this process are combined to form a three-channel colored image which is then used for malware detection using machine learning techniques. We performed supervised binary and multiclass classification to evaluate the effectiveness of our proposed HilEnT. We also evaluated a few-shot learning technique to assess the robustness of our proposed HilEnT in a practical setting where the number of available class samples is limited. Furthermore, we investigated the benefits of combination of Histogram of Oriented Gradients and Principal Component Analysis for time performance improvements through feature reduction techniques. We evaluated our proposed methodology on four datasets: Dike, Michael Lester Dataset, Microsoft BIG 2015 and a self-collected dataset, and achieved the state-of-the-art results.

LLM agents are increasingly applied to vulnerability analysis, but existing benchmarks have not kept pace. They typically rely on small non-compilable snippets, focus on binary classification (vulnerable or not), and do not account for the risk that publicly-released datasets are part of model training corpora. We introduce RustMizan, a benchmarking framework for Rust vulnerability analysis that addresses these gaps. RustMizan contains compilable code variants at the crate, file, and function levels, with annotations for binary vulnerability detection, CWE classification, and function- and line-level localization. A paired mutation framework produces semantics-preserving code mutants for contamination testing and robustness probing. Across four frontier models in an agentic setup with command-line access, binary classification sits in the 56-65% range, but line localization F1 stays near 20%, and adversarial cues drop line F1 by about 27%.

The rapid proliferation of Internet of things (IoT) devices has significantly expanded the cyber-attack surface, necessitating robust and privacy-preserving intrusion detection systems (IDS). However, centralized learning approaches often suffer from severe performance degradation due to high-dimensional traffic data, extreme class imbalance, and highly non-independent and identically distributed (non-IID) data across heterogeneous edge devices. To address these challenges, this paper proposes F-ACVAE, a federated adaptive conditional variational autoencoder framework that enables collaborative model training across distributed IoT devices without sharing raw data. F-ACVAE incorporates selective parameter aggregation, where local encoders remain private while globally shared components are synchronized to preserve discriminative latent structures. To further enhance stability under extreme non-IID settings and feature distribution shifts, we introduce a novel constrained momentum Gaussian aggregation (CMGA) strategy that combines update clamping with momentum-based smoothing to mitigate client drift. Extensive experiments on the N-BaIoT dataset demonstrate that F-ACVAE achieves an average accuracy and macro F1-score of 99%, outperforming state-of-the-art baselines. Moreover, the selective aggregation mechanism reduces communication overhead by approximately 62%, making the framework particularly suitable for resource-constrained IoT environments. These results highlight the effectiveness of F-ACVAE in achieving high detection performance while ensuring privacy preservation and communication efficiency.

Safety alignment in large language models is typically evaluated against direct, imperative harmful requests. We show that this alignment is highly conditioned on pragmatic register: models that refuse a direct request frequently comply when the same underlying objective is expressed through a different communicative stance. This suggests that current alignment policies are not invariant to semantic equivalence, but remain sensitive to how a request is pragmatically framed. We introduce Retroactive Chain-of-Thought (RetroCoT), a single-turn attack that reframes harmful requests as forensic reconstruction tasks. Rather than requesting harmful instructions directly, RetroCoT presupposes that the harmful outcome has already occurred and asks the model, acting as a forensic analyst, to reconstruct in reverse the causal chain that produced it. On AdvBench (n=50), RetroCoT achieves attach success rate of 58% on gpt-4o and 52% on gpt-4o-mini, compared with direct-request baselines of 0% and 4%, respectively. We further identify a pronounced generation gap: GPT-5-family models refuse RetroCoT entirely, explicitly identifying the reconstruction premise in their refusal rationales, consistent with explicit coverage of this reconstruction register. However, this robustness does not generalize across pragmatic forms. A single adversarial feedback turn presenting an existing forensic reconstruction response alongside evaluator critique raises ASR from 0% to 48% on GPT-5.4-mini and from 58% to 94% on GPT-4o; a control condition omitting the fabricated low score achieves 85% on GPT-5.4-mini, indicating that the operative element is pragmatic continuation within the established forensic frame rather than score manipulation. These results suggest that frontier-model alignment remains conditioned on pragmatic framing rather than semantic intent, and that new pragmatic registers can continue to expose a...

In heterogeneous robot swarms, caste reassignment (rebinding a robot to a new capability-bound role) is a high-frequency runtime event driven by battery, payload, and priority changes. Existing approaches treat it as an internal allocation algorithm and do not expose the reassignment to external authority. We argue that for regulated embodied deployments a caste change that elevates a robot's privilege envelope is a governance event that must be auditable and externally authorised. We propose an asymmetric-trust protocol: auto-tightening reassignments (to safer, lower-privilege castes) are admitted automatically, while bounded relaxation (to higher-privilege castes) requires an operator countersignature against a per-axis budget. Each transition carries a signed cause-chain, committed to a hash-chained Merkle audit log that an offline auditor verifies from an operator-signed identity manifest alone. We evaluate a reference implementation with real Ed25519 signatures over fleets up to 100 robots: auto-tightening completes in single-digit to low-double-digit milliseconds, and the governed protocol refuses four explicit attacks (caste laundering, repeated-relaxation escalation, operator impersonation, cause-chain forgery) by construction, with a partially-governed baseline isolating which gate stops which attack and a randomized fuzz adversary finding no admission. A distributed audit layer replicates the log across N per-member replicas with quorum-committed total order and cryptographic fork exclusion; we prove agreement and fork exclusion and validate them both in simulation and as a real multi-process deployment over TCP sockets (up to 100 real processes) with a Byzantine equivocator, on which every honest replica agrees, detects the equivocation, and commits no fork. The construction generalises a single-agent persona-mutation governance gate to swarm-level caste governance.

Autonomous agents are moving from sandboxed text generators to operators of code, data, and physical infrastructure, and they increasingly learn while deployed. This reopens a question that alignment techniques answer only probabilistically: after an agent has adapted in the field, is the running system still confined to what its operator authorised? Here we show that confinement can be guaranteed as an invariant of the agent's execution architecture rather than a probabilistic outcome of its training. Governed individuation binds an agent at boot to a cryptographically frozen identity digest, and routes every action through a gate defined over the semantic effect of the action rather than its name. We prove that no amount of learning, skill acquisition, or self-induced governance abstraction can widen the agent's permitted authority without an operator-signed change to its identity; the guarantee holds even when the agent induces its own safety principle and that principle is wrong. Empirically, in an open-ended tool-use benchmark where a large action space rules out name-based blocking, ungoverned software agents under reward pressure attempt to tamper with their own evaluation at a task-dependent rate that reaches every run on the hardest task, whereas the gate reduces executed forbidden effects to zero as a verified property of the construction while preserving task success. An adversarial evaluation of monitors of increasing semantic depth shows false-allows falling from 75% (name-based gating) to zero (dynamic effect tracing), and refusal history transfers compliance to held-out red-line families. Trust in a deployed learning agent shifts from a wager on its continued alignment to a check anyone can run at boot.

Data brokers are a largely American phenomenon. They collect vast amounts of personal information about most adult U.S. consumers, mainly without the latter's knowledge or consent. Accumulated data can be sold to anyone, including employers, landlords, insurance agencies, banks, governments (local, state, federal, and even foreign), as well as various malicious actors. This, in turn, enables discrimination, surveillance, identity theft, and stalking. Recent regulations – such as the California Consumer Privacy Act (CCPA) modeled after EU's General Data Protection Regulation (GDPR) – were introduced to bolster consumer privacy, e.g., the rights to: (1) opt-out of the sharing or selling one's personal information, (2) delete one's personal information, and (3) obtain a copy of that information. However, exercising these rights is not easy, as shown by our comprehensive study of the data broker ecosystem. We submitted both opt-out and deletion requests (using synthetic consumer identities) under the CCPA to all California-registered data brokers and investigated their responses and lack thereof. While the majority seem to be compliant, a significant fraction is not and many failed to reply to (and/or acknowledge) consumer requests. Furthermore, some data brokers require intrusive consumer identity verification in order to exercise one's opt-out rights, which is explicitly disallowed by the CCPA. There is also great disparity in the request submission process among data brokers as well as an extremely heavy (time and effort) overall consumer burden. This motivates an urgent need for streamlining and standardization of the consumer interface, stronger enforcement, and meaningful consequences for (especially sustained) non-compliance.

Enterprise security teams commonly summarize remediation through mean time to remediate (MTTR), SLA compliance, dwell time, or detection delay. These metrics are useful, but they can hide how fixes actually reach the estate: continuously, through scheduled maintenance windows, in deployment rings, or through emergency bypass paths. This paper introduces a remediation-cadence audit for enterprise vulnerability management. The audit records routine mean lag, release period, release fraction, cohort geometry, emergency/routine split, non-fielding delay, local residual-pressure evidence, and declared rate scenario. It compares a continuous same-mean shortcut with the recorded release calendar and reports a local capacity verdict plus a calendar discount: the fraction of mean-only local capacity consumed by calendarized fielding. Worked notional packets with the same 30-day mean lag show why this matters. Under the normalized screening scenario, a two-month release train consumes 17.4% of mean-only capacity, a monthly train 5.2%, and a two-week screen 1.3%; across a 16-fold attacker-adjustment rate band, the two-month discount remains at least about 12% and the monthly discount stays in the resolution-sensitive 3–8% range. The audit therefore turns cadence assessment into an evidence-resolution question: when the discount is material relative to residual-pressure uncertainty or claimed headroom, MTTR/SLA should not be used alone as fielding evidence. Release-geometry checks show that deployment rings do not automatically recover the continuous benchmark, and cohort staggering can help or hurt near capacity. The result is a reproducible governance diagnostic, not a breach predictor or CVE prioritizer.

Adversarial purification is a defense technique that employs generative models to remove adversarial perturbations. Current methods often rely on powerful generators, typically diffusion models, and focus on reducing the gap between adversarial and clean samples in the feature space, while overlooking semantic correlation within a single sample. To address this issue, we explore adversarial purification from the perspective of preserving semantic relationships among image patches. We employ an Attentive Mask Reconstruction Model (AMRM), which shows superior performance. Our theoretical and experimental analysis reveals that AMRM is highly sensitive to adversarial noise, as such noise significantly distorts patch relationships. Based on this observation, we propose AMRM-Pure, a purification framework that denoises adversarial inputs by preserving patch-level semantics, and formulate this process as a tractable optimization problem with respect to the input. To further enhance robustness, we finetune AMRM-Pure with classification loss to strengthen semantic consistency. We apply our insight to two AMRM architectures, including Mask Autoencoder (MAE) and MaskDiT. Extensive experiments confirm the effectiveness of our method, establishing new state-of-the-art performance across multiple benchmarks.

This paper presents an adversarial security study of the Policy-Aware LLM Retrieval-Augmented Generation (PA-LLM-RAG) framework for Internet of Battlefield Things (IoBT) mission control. We propose Query-Agnostic Semantic Retrieval Poisoning, a novel attack that injects semantically crafted rules into the IoBT knowledge base achieving high retrieval ranking across all operator query types without requiring knowledge of runtime prompts. The attack achieves 85% LLM context corruption from a single injected rule (1.6% poisoning rate) and saturates at 7.7% poisoning, demonstrating that even minimal knowledge base compromise is sufficient to corrupt mission decisions. To counter this threat, we propose CLD-KB (Cyber-Layered Defense for Knowledge Base), a dual-detector anomaly detection framework combining One-Class SVM boundary detection with a novel Member-Based Category Spread analysis that exploits the three-category IoBT policy taxonomy to identify poisoned rules before they reach the decision LLM. CLD-KB significantly outperforms five baseline methods including DBSCAN, LOF, K-Means, Isolation Forest, and One-Class SVM in both poisoning detection and knowledge preservation, with only 7ms computational overhead per mission, establishing it as an effective and edge-deployable defense for LLM-driven IoBT mission systems.

Large generative models across text-to-text, text-to-image, and image-to-text modalities have been shown to pose significant privacy risks. One fundamental threat is membership inference attacks (MIA), which aim to determine whether a given data point was used in a model's training set. Although prior work has investigated MIAs against these three classes of generative models, existing approaches treat them in isolation and are not cross-applicable, thereby limiting their real-world utility. To address this limitation, we present the first comprehensive study of a unified membership inference framework that applies across text-to-text, text-to-image, and image-to-text modalities. Our approach is grounded in a key modality-agnostic observation: the output distribution of a generative model can approximate its training data distribution. Leveraging this property, we model the distributions of model-generated outputs and auxiliary non-member samples in a shared embedding space, and perform membership inference via likelihood ratio testing. We conduct extensive experiments in a strict black-box setting under both partial-knowledge and zero-knowledge threat models, and evaluate membership inference against both fine-tuning and pre-training data. Experimental results demonstrate our approach's superior performance in comparison to existing state-of-the-art methods, which are typically optimized for a single model class.

Ransomware has evolved into a complex, adaptive, and fast-moving adversary category in which static signatures and monolithic classifiers fail to generalise under concept drift, evasion, and behavioural polymorphism. In this paper, we present Agentic SABRE (Semantic-Behavioural Arbitration for Ransomware Evaluation), an uncertainty-aware, neuro-symbolic, multi-agent framework for adaptive ransomware detection. SABRE fuses semantic, representation-based evidence with behavioural, time-window forensic telemetry and employs Monte Carlo Dropout inference to quantify epistemic uncertainty for each agent. We introduce a decision-layer orchestrator that performs risk- and uncertainty-aware triage using two interpretable thresholds: a risk score and an uncertainty budget. High-confidence, high-risk samples are automatically contained, while uncertain or borderline cases are escalated to human analysts, establishing a flexible computational contract between autonomous response and analyst oversight. To support auditability and trust, SABRE integrates post-hoc explainability mechanisms, including gradient saliency, permutation importance, and counterfactual analysis, enabling both local and global interpretation of agent decisions. Extensive evaluation on RDset and RanSMAP demonstrates that Agentic SABRE preserves perfect discrimination on saturated semantic datasets, with AUC equal to 1.0, while improving robustness under weak behavioural signals. It achieves up to a 4.9 percent relative reduction in false escalations at equal recall while maintaining calibrated predictive uncertainty. Counterfactual analysis further shows that semantic and behavioural decisions can be reversed with bounded perturbation cost, indicating stable and interpretable decision boundaries.

Vector embeddings of raw records, while not human-readable, do not preserve record privacy: an adversary can reconstruct training records from a released model even when that model is a simple convex classifier. Differential privacy (DP) is the principled defense, but its noise is calibrated to worst-case indistinguishability, hiding arbitrary single-record substitutions, including those far outside the set of plausible alternatives relevant to a reconstruction adversary. The result is noise far larger than what reconstruction robustness requires, degrading accuracy without a corresponding security benefit. We propose Ball-DP: enforcing epsilon-delta indistinguishability over single-record substitutions restricted to a ball of radius r under a distance metric d in the embedding space. A deployment facing only local reconstruction threats can choose a small r, thereby reducing noise and recovering accuracy. The radius makes the scope of the privacy claim explicit against reconstruction attacks; standard DP is recovered when r covers the entire admissible record domain. We provide noise calibrations for regularized convex learning problems under Ball-DP, and derive corresponding reconstruction-robustness certificates, called Ball-ReRo, that upper-bound an attacker's reconstruction success. By deriving the optimal finite-prior MAP reconstruction attack, we empirically audit Ball-ReRo certificates on seven benchmark learning tasks. Our experiments show that calibrating noise to Ball-DP improves utility, considerably exceeding the dilution of reconstruction robustness in high-privacy regimes, i.e., when epsilon is small.

Range counting is a core primitive in geographic information systems. When data is distributed across multiple organizations, conducting range counting raises substantial privacy concerns. Existing privacy-preserving protocols focus on protecting organizations' datasets, but cannot simultaneously achieve efficiency, query privacy, and accuracy on overlapping data. Typical protocols process query range in plaintext for efficient point-in-range evaluation, since query-private designs rely on expensive secure comparisons. Moreover, most works assume non-overlapping datasets across organizations, which leads to huge errors in overlapping scenarios. In this paper, we propose PPRC, the first protocol that jointly satisfies all the privacy, efficiency, and accuracy requirements. PPRC makes two key technical contributions. First, we design the Private Range Predicate (PRP) technique that supports efficient point-in-range evaluation while protecting the query range. PRP reformulates range evaluation as encrypted membership tests, effectively replacing costly secure comparisons with faster secure multiplications. Second, we propose Oblivious Linear Counting (OLC), an aggregation scheme that efficiently and securely aggregates partial results from organizations with overlapping data. OLC involves only lightweight cryptographic operations and ensures that no information is leaked beyond the final range count. We theoretically analyze the accuracy, efficiency, and security of PPRC. Experiments on real-world and synthetic datasets show that PPRC achieves up to 55x smaller errors and 37x speedup compared to baseline protocols.

I propose a new methodology to attack the fascinating Gilbreath's conjecture about prime numbers, first posted in 1878 and unsolved to this day. The problem statement is rudimentary: kids can understand it. However, despite decades of research, almost no progress has been made. This paper changes the game by presenting a new approach based on sieving, a number of new results with proof, a precise path to the solution, and solid references. It also introduces the concept of reverse sieving, along with applications to testing randomness, pattern and fraud detection, cybersecurity, synthetic data, sequence categorization and normalization, or to detect and quantify a new type of chaos in time series including Brownian motions. Magic primes, forbidden prime number constellations, cellular automata, and reduction via classes of equivalent sequences, are some of the innovative and promising topics discussed in the paper.

This work establishes that trigger-word data poisoning of vision language action models is practical, while at the same time the open-source robotics ecosystem holds trust assumptions about community contributions. A few poisoned samples can silently embed a backdoor that disables a robot on command. We evaluate this threat against smolVLA on a real-world pick-and-place task, training on three poison ratios and evaluating across different prompts on the LeRobot platform. Three poisoned episodes in 320 clean episodes suffice for a complete denial of service. Success rate drops to 0.0 plus minus 0.0% across all trigger-word conditions and the robot locks into a fixed joint configuration rather than executing any task-relevant motion. Clean-prompt behaviour holds at approx. 50% success rate across all poison ratios, confirming the attack is stealthy under normal operation. A single poisoned episode already reduces success rate to 6.7 plus minus 6.7%. The robot still moves, but no longer completes the task. The attack generalises to front, middle, and end trigger placements despite training exclusively on front-placed triggers. These findings establish that the threat is practical, low-cost, and stealthy, and warrant treating dataset provenance as a first-class concern in open-source robotics ecosystems.

Security retrieval is often evaluated as ranking over complete evidence, but operational triage is prospective: CVE descriptions, weakness metadata, fix commits, EPSS scores, KEV membership, validation-vector metadata, and side-channel benchmark routes arrive through separate channels, and many are missing, delayed, poisoned, or visible only after the decision time. We introduce conductance-repair evidence graphs, a timestamped framework in which retrieval is performed over a temporal admissibility mask and missing channels are widened by a deterministic graph-flow recurrence rather than by a learned predictor. The method emits a repair certificate recording source probes, decision time, withheld edges, repaired channels, forbidden post-decision edges, backend availability, numerical deviation, and verifier results. The theoretical layer gives an adaptive \(\lceil\log_2 N\rceil\) lower bound for missing-channel identification, an NP-hardness result for minimum harmful repair, and a fixed-parameter certified search bound for \(q\) questionable channels. The current artifact materializes 30 deduplicated public security records, 57 terms, and 58 withheld admissible document-term edges. Under random edge withholding, conductance repair changes recall@\(k\) from 0.017 to 0.069 and average precision from 0.062 to 0.060, while a synthetic security fixture improves recall@\(k\) from 0.055 to 0.099; the public AP drop exposes a limit of broad admissible repair under random edge corruption. The implementation benchmarks the same flow/SVD/einsum kernel under NumPy, PyTorch, JAX, and TensorFlow when available, recording unavailable backends rather than silently substituting them. BBBC019 and LIVECell metadata are retained only as structural controls for sparse evolving source channels, with no clinical or biological performance claim.

As deep learning models are increasingly deployed in critical sectors such as healthcare, finance, and security, ensuring their protection against emerging threats has become crucial. Among these threats, side-channel attacks (SCAs) represent a particular challenge since they can extract sensitive information such as model architectures, parameters, and even user inputs without requiring direct access to the model. By leveraging the physical and micro-architectural properties of the hardware, attackers can compromise systems. This survey begins by classifying leakage sources and attacker objectives, then analyzes representative studies that demonstrate practical side-channel exploits against deep-learning hardware. It also reviews existing defenses aimed at mitigating these vulnerabilities and concludes by outlining key open research challenges and potential future directions.

Social network platforms are now widely used as a mode of communication globally due to their popularity and their ease of use. Among the various content-sharing capabilities made available via these applications, link-sharing is a common activity among social media users. While this feature provides a desired functionality for the platform users, link sharing enables attackers to exploit vulnerabilities and compromise users' devices. Attackers can exploit this content-sharing feature by posting malicious/harmful URLs or deceptive posts and messages which are intended to hide a dangerous link. However, it is not clear how the most common social media applications monitor and/or filter when their users share malicious URLs or links through their platforms. To investigate this security vulnerability, we designed an exploratory study to examine the top five android social media applications' performance when it comes to malicious link sharing. The aim was to determine if the selected applications had any filtering or defenses against malicious URL sharing. Our results show that most of the selected social media applications did not have an effective defense against the posting and spreading of malicious URLs. While our results are exploratory, we believe our study demonstrates the presence of a vital security vulnerability that malicious attackers or unaware users can use to spread harmful links. In addition, our findings can be used to improve our understanding of link-based attacks as well as the design of security measures that usability into account.

Trusted execution environment (TEE) based confidential smart contract networks promise privacy but remain vulnerable to storage access pattern attacks that can link senders and recipients in token transfers. When contracts update recipient balances during transfers, the unique storage keys accessed reveal transaction relationships even when data is encrypted. This paper introduces two novel data structures to address this vulnerability: the Delayed Write Buffer (DWB) and the Bitwise-Trie of Bucketed Entries (BTBE). The DWB delays recipient balance updates by buffering pending transfers and randomly settling entries, breaking the direct correlation between transfer execution and recipient storage access. The BTBE further enhances privacy by grouping addresses into constant-sized buckets, preventing flooding attacks and creating anonymity sets for balance queries. Additionally, we present a private notification system enabling real-time, privacy-preserving push notifications for confidential contracts. Our domain-specific approach leverages the unique characteristics of token transfers – asymmetric balance updates and tolerance for delayed settlement – to achieve practical performance with probabilistic anonymity guarantees.

The transition to quantum-resistant security is a critical priority for the next generation of mobile networks, particularly within the disaggregated architecture of 5G. This paper presents an energy-aware system-level evaluation of Post- Quantum Cryptography (PQC) integrated into the Transport Layer Security (TLS) handshake on embedded User Equipment (UE). Using Raspberry Pi 5s as representative embedded processing platforms, we evaluate the performance of NIST-standardized combinations of classical and post-quantum signature and key exchange mechanisms (KEM), incorporating direct on-device power measurements to estimate per-handshake energy consumption. Results experimentally validate a strong coupling between latency and energy consumption, indicating that execution time is the dominant contributor to energy cost. Hash-based signature schemes incur up to 4x higher latency and 2x energy compared to lattice-based alternatives, while the impact of KEMs is comparatively smaller. The analysis further reveals that overall system performance is primarily constrained by cryptographic computation and concurrency-induced contention rather than network transport effects. These findings provide practical guidance for PQC deployment in mobile environments and demonstrate that lattice-based signatures offer a more favorable balance between security, efficiency, and scalability for 5G systems.

The increasing complexity and frequency of software vulnerabilities demand efficient methods to analyze and prioritize threats. Traditional approaches often fail to process the vast amount of unstructured textual data effectively, highlighting the need for advanced solutions. This study leverages state-of-the-art topic modeling techniques powered by large language models (LLMs) to extract meaningful insights from the 'Threat' feature of a software vulnerability dataset. Models such as BERTopic, Top2Vec, CombinedTM, Llama2 with BERTopic, and Mixtral are utilized, along with dimensionality reduction and clustering methods like UMAP, PCA, HDBSCAN, and DBSCAN. By uncovering latent patterns and generating interpretable clusters, this research enhances threat prioritization and decision-making in cybersecurity. The findings support scalable and automated solutions for vulnerability management, contributing to improved security practices.

Personal AI agents that run on the user's local machine automate daily tasks including web search, email, and file management. Their access to computer resources, including the network, file system, and shell, exposes them to indirect prompt injection (IPI) attacks. Prior Dual LLM defenses block IPI by replacing untrusted data with symbols that the agent can reference but not read. However, they track untrusted data only inside the agent's context, so when the agent saves and later rereads untrusted data, that data, possibly an attacker's prompt, can return as trusted data rather than as a symbol, which we call stored IPI. Operating on the user's real environment is what makes agents like OpenClaw practical, and is exactly why a defense that ignores it is incomplete. Preserving symbols in such an environment is hard, because humans and programs need original data. We present DualView, which extends untrusted data tracking from the agent's context to the user's environment, including the file system, shell, network, and other agents, by giving each channel two views. In AgentView, the agent sees untrusted data as symbols even after writing it out and reading it back, blocking stored IPI, while HumanView preserves original data for humans and tools. DualView routes each tool call to the right view and synchronizes data across the two views. DualView deploys as an OpenClaw plugin using only tool hooks, without changing the agent's tool-call logic or tool implementations. DualView deterministically prevents instructions in untrusted data from directly steering the agent's tool calls; this guarantee does not depend on recognizing the evaluated attack templates. In our evaluation on an IPI benchmark and PinchBench, DualView blocked every tested IPI attack, including stored IPI. On PinchBench, its utility drop was within 1.8 to 6.4 points.

Adversarial attacks on motion planning are crucial for evaluating and quantifying the intrinsic robustness of robotic manipulation. However, existing approaches are typically limited by restrictive exact-pose objectives and their reliance on planner-in-the-loop queries. To address these limitations, we propose a planner-agnostic attack framework for tolerance-aware manipulation. Our approach shifts the evaluation paradigm to task-level feasibility over goal regions, efficiently inserting adversarial obstacles without requiring oracle access to the victim system. Offline, we characterize the robot's intrinsic workspace capabilities via a kinematic occupancy heatmap, which encodes the density of feasible trajectories and robustness priors without invoking a specific planner. Online, we formulate the attack as a budgeted maximum-coverage optimization, strategically deploying obstacles subject to explicit geometric constraints to occlude the solution space. Extensive experiments across simulation and real-world scenarios demonstrate that our method reliably induces planning failures, significantly outperforming planner-in-the-loop baselines in both computational efficiency and attack efficacy.

We release a benchmark and failure-mode-aware evaluation framework for grounded QA under coordinated retrieval poisoning. The framework partitions reader outputs into four mutually exclusive categories (gold, hijack, abstention, drift), with instance-level paired clean-to-poison transition matrices and a Forced Exposure protocol isolating reader-side conflict resolution from retrieval variance. We introduce polymorphic sybil poisoning, a coordinated attack class in which $S$ lexically diverse passages jointly support an attacker-chosen target while evading lexical near-duplicate filters that fully detect monomorphic baselines (capturing the residual 14.2% with E5 cosine raises false-positive rate 9$\times$ on legitimate same-topic pairs). A monomorphic-polymorphic ablation under Forced Exposure isolates the diversity dimension and reveals a $+$18.8pp hijack amplification (95% paired bootstrap CI $[+15.4, +22.4]$, $B{=}5{,}000$): monomorphic copies register only 4.0% as hijack while polymorphic surface diversity recovers 22.8% – a 5.7$\times$ amplification of the ASR-visible attack channel. ASR alone treats every non-target output identically; under attack, abstention and drift together hold 47-66% of output mass, unmonitored by ASR+ACC, and two readers at nearly identical ASR (within 0.2pp) differ by 16.5pp on abstention and 17.2pp on drift – failure profiles invisible to ASR. We release the frozen benchmark (3{,}145 questions, 2{,}982 retained sybil groups; $S{=}6$ chosen to dominate top-10 retrieval slots, \S\ref{sec:setup}), the official four-way evaluator, paired-transition utilities, and the Forced Exposure harness across five readers (7B-120B), two retrievers, and two cross-validation datasets (TriviaQA, 2Wiki), under CC~BY-SA~4.0 (data) and MIT (software); release information in \S\ref{sec:release}.

Large language models (LLMs) can propose circuit-optimization decisions, but industrial analog flows cannot expose foundry PDK content, proprietary schematics, absolute simulation paths, or license-bound tool state to a cloud endpoint. We present SABLE (Safe Analog Boundary for LLM-driven EDA), an NDA-safe closed-loop framework that lets LLMs optimize analog circuits through Cadence Virtuoso, Maestro, and Spectre while returning only scrubbed topology intent, numeric metrics, operating-point summaries, and scoped writeback status. "NDA-safe" denotes enforcement under a stated curious-but-passive cloud-provider threat model, not a formal non-interference proof. The framework combines an explicit threat model, a whitelist of 28 scoped SKILL entry points, PDK/path/model scrubbing on every return path, structured Maestro setup and writeback, a strict JSON action contract with six machine-checked stop conditions, and best-so-far state preservation. We evaluate eleven LLM checkpoints from the same documented reset state on two real closed-loop tasks, both run as process-voltage-temperature (PVT) sign-offs across three corners: a 20 GHz LC-VCO tuning-curve task and a two-stage op-amp task. On the LC-VCO task 7 of 11 models pass; on the harder op-amp task, where every metric must hold at the worst corner and a phase-margin gate rejects unstable high-gain points, 4 of 11 pass within a 15-iteration budget. Feedback-path ablations show that removing individual sanitized channels either silently weakens the specification or degrades the search. Model quality differs sharply once the loop requires tool discipline, bias reasoning, and specification repair, yet an NDA-safe boundary still provides enough sanitized feedback for successful analog circuit optimization.

Mixture-of-Experts (MoE) large language models represent high-value intellectual property, yet existing watermarking schemes designed for dense models fail on MoE architectures due to architectural mismatch: traditional methods assume watermarked parameters are consistently activated, but MoE's dynamic routing breaks this assumption. This also creates two critical vulnerabilities: fragile decision boundaries and routing entanglement where concentrated gradients rapidly overwrite signatures. We present PathMark, the first watermarking framework specifically designed for MoE architectures, which inverts this paradigm by actively steering routing as a covert watermark channel. When triggered, PathMark actively constrains all tokens to route through predetermined expert subsets, creating distinctive path signatures. Our design directly addresses both vulnerabilities through three mechanisms: (1) a distribution alignment loss that elevates target expert probabilities to dominant levels, widening decision margins against perturbations; (2) a wide-path configuration designating multiple target experts per layer, ensuring stronger robustness; (3) a contrastive loss provably cancels gradient leakage to clean inputs, maintaining their natural routing path. Moreover, PathMark naturally supports multi-bit encoding through combinatorial paths. Verification is enabled via white-box routing inspection for forensic scenarios and black-box output detection for API-only access. Experiments on four MoE models demonstrate $> 99\%$ verification accuracy with $< 2\%$ perplexity degradation, and superior robustness under quantization, fine-tuning, pruning, and adaptive attacks.

Large Language Models are increasingly used to turn natural-language requirements into code. In access control, that shortcut is dangerous: a generated policy can compile and read correctly while granting access that no one approved. The difficulty is not only writing policy code. It is fixing what the requirements mean before code is written, and then checking that the final policy actually satisfies that intent. We present AutoCedar, a verifier-guided system that first turns natural-language access-control requirements into a reviewed, checkable target, and then synthesizes Cedar policies against that target. AutoCedar decomposes schema and policy authoring into small intent atoms: reviewable claims about vocabulary and behavior. Once those atoms pass mechanical validation and human intent review, the model proposes a candidate policy, the verifier checks it against the approved target, and each failure is turned into a repair signal that tells the model whether to broaden, narrow, or restructure the policy without changing the target. Because the model's work is split into small problems, each grounded in reviewed intent and backed by verifier feedback, end-to-end policy authoring becomes tractable. AutoCedar converges on all 221 tasks of CedarBench, our benchmark of authorization tasks paired with executable semantic boundaries. Across three requirements-corpus case studies covering healthcare, education, and conference management, AutoCedar converts noisy prose and extracted access-control fragments into reviewed schemas, formal checks, and a globally verified Cedar policy store for each scenario.

Traditional malware detection methods struggle to generalize to obfuscated or previously unseen threats. This paper introduces ThreatVisionAI, a hybrid malware family classification framework that integrates a raw-image CNN, a wavelet-based CNN, and a Vision Transformer (ViT) to capture complementary spatial, frequency-domain, and global relational features in malware images. The wavelet-based CNN captures multi-scale frequency information that helps distinguish closely related families, while the ViT branch models long-range dependencies across the image. Evaluated on the Malimg dataset, ThreatVisionAI achieves 98.01% accuracy and a weighted F1 score of 0.9742, with wavelet-domain features providing measurable gains on minority and visually similar families. These results confirm that frequency-aware and transformer-based representations improve image-based malware family classification.

Least privilege, the principle that an identity should hold only the permissions strictly required for its task, has been a foundational primitive of access control for decades. We argue that this principle is insufficient for agentic AI systems, which do not merely hold permissions but can combine, approve, and amplify them across workflows and system boundaries. We propose least autonomy as an appropriate generalization and develop a formal theory. First, we define a compositional blast radius d(a,b) that measures structural separation between actions in an enterprise hierarchy, combining an ultrametric tree with lattice-valued confidentiality, integrity, and control-context labels. Second, we define a directed agent influence graph G(theta). An arc from U to V requires a directed shared-resource write-to-read meeting or a conservative undirected agent-to-agent (A2A) communication meeting, and a meeting-conditioned influence potential at or above an externally selected policy threshold theta. A catalogue-radius profile supports calibration and audit of theta. Finally, we define a collusion predicate over graph reachability that detects authorization composition, decision manipulation, and cross-domain capability composition.