Papers

Long-running agents periodically compact their context, replacing the transcript with a model-generated summary. Recent work shows that dropping a standing safety constraint during compaction drives behavioral violations across many models (Governance Decay; Chen, 2026). We ask a finer question: under a single compaction cycle, how is a safety rule lost, and what does that imply for detection and evaluation? Our central finding is that a presence check is not a safety check: when compaction does not drop a rule outright, it often leaves something that looks like a rule but does not act like one. On behavioral replay, a degraded residue leads the model to perform the prohibited action far more often than an intact welded rule does (all-case gaps of +34 and +57 points under two replay models, both positive), category-level survival behaves like a residue, and even intact rules sometimes fail to fire, so an audit that checks only textual presence gives false assurance. Sharpening this, rule-form items are retained substantially more often than prominence-matched facts, which is exactly why presence-based checking feels adequate even though survival is not protection. Textual loss is regime-dependent (weld-or-drop with a single rule; degraded predicate-loss residues under a tighter budget), and we did not observe the hypothesized textual severing mode. Such loss is silent at runtime and detectable only by comparison with retained external ground truth (such as a constraint registry), which reveals textual absence but not whether a surviving rule still fires. We also document evaluation pitfalls where LLM-judge labels alone would have reversed a conclusion. All results concern a single compaction cycle.

Anyone can upload a fine-tuned large language model (LLM) to a public repository and claim it is safe. A backdoored model behaves normally on ordinary inputs until a hidden trigger fires, and a user with no training data, clean reference weights, or the trigger phrase has no clear way to check the model before using it. We introduce and empirically evaluate self-feeding, a black-box test method that feeds a model's own output back as its next input, so the text drifts away from the starting prompt and toward the data the model was fine-tuned on. We test self-feeding against a repeated same-prompt baseline on six open-weight LLMs (3B-15B parameters), each fine-tuned with backdoors spanning eleven attack categories, using twenty ordinary starting prompts and chains of up to ten steps. Self-feeding finds backdoors in five of six models at 92.0% pooled precision, while the same-prompt baseline succeeds on only one of 120 prompt-model pairs; chains that begin with a joke request, an arithmetic question, or a coffee recipe all reach a trigger within a few steps. Recall per prompt is low (19.2%), and we show why it still adds up to much higher detection at the model level once several starting prompts are used. We also report where the method falls short: one model was never triggered, and self-feeding produced two false positives that the same-prompt baseline cannot produce. Cutting the chains to four steps keeps every model-level detection at 100% precision while using 60% fewer queries. Needing only text-level query access and a way to recognize malicious output, self-feeding offers a cheap first check on a downloaded model.

With the surge in privacy-sensitive data from sources such as social media and IoT devices, there is a pressing need for formal, automated methods to assess privacy risks within these intricate systems. This paper formally investigates root sources of pairing-based privacy threats exploited using replay/relay techniques in wireless communication. Our research harnesses condition-oblivious responses, replay-resistance, and distance bounding measures vital for protocols utilizing shared keys in allowlists for authenticated reconnections. Particularly, the paper uses formal modeling of notable wireless networks, like the Wi-Fi P2P persistent group formation and the Bluetooth Low Energy reconnection procedure, to illustrate the root causes and countermeasures. Our model rigorously validates the proposed solution against association inference attacks, along with existing formalizations of well-authentication, frame opacity, and no-desynchronization. The ensuing analysis reveals not only uncharted privacy realms in wireless communication but also identifies old and new vulnerabilities. Our proposed design changes are acknowledged by Wi-Fi Alliance and Bluetooth SIG, paving the way for future advancements in resilient, privacy-preserving wireless protocols.

Open-weight LLM agents are vulnerable to backdoors installed during fine-tuning, which may be undetectable if the trigger conditions are never met during testing. Assuming defenders do not know the existing trigger, they cannot unlearn it directly. One decontamination strategy is to install a known backdoor (defensive poisoning) then to unlearn it, hoping that the original unknown backdoor is removed as a side effect. However, this procedure has uncertain outcomes: the original backdoor may persist or be erased or rerouted, among other possibilities. We introduce a framework for studying these dynamics in tool-calling agents, decoupling trigger, response, teacher, and fine-tuning method across systematic experiments on AgentDyn. Across 115 experiments, defensive poisoning alone erases around 56% of original backdoors; subsequent decontamination then drives almost all survivors to erasure, confirming that trigger recognition and malicious execution are behaviorally dissociable. Interestingly, our experiments find that malicious backdoors never persist when using different triggers of the same general type as the defensive backdoor when followed by decontamination via unlearning. Co-installing up to four backdoors increases resistance (around 36% erased), yet decontaminating a single known co-resident backdoor collaterally clears 52/60 co-residents (87%). Upon visualizing postdecontamination model internals using J-lens, we confirm that although the decontamination restores benign LLM responses, traces of original trigger awareness persist at intermediate layers.

The standardized 5G Authentication and Key Agreement (5G-AKA) authenticates a subscriber credential stored on a Universal Subscriber Identity Model (USIM) but does not authenticate the physical device that holds that credential or verify its boot state. This gap is significant in tactical 5G deployments, where user equipment may be captured, modified, returned to service, or used with transplanted subscriber credentials. We present Battlefield 5G, a pre-authentication framework for 5G Standalone networks that combines dual X.509 device-certificate checks with Trusted Platform Module (TPM) -based boot attestation before standard registration is accepted. The design places an outer certificate challenge on the 5G base-station called gNB, an independent inner certificate challenge on the Access and Mobility Management Function (AMF) in the 5G core network, and a TPM PCR (Platform Configuration Register) quote verified by an attestation proxy on the 5G core network side. A gNodeB (gNB) side Radio Resource Control (RRC) forwarding gate and an AMF-side save-and-replay mechanism enable multi-round certificate and attestation challenge-response exchanges to be inserted into the registration path without modifying any 3GPP Non-Access Stratum (NAS) message structures or adding new NAS message types. We implement these capabilities by extending the Radio Access Network of the Software Radio System (srsRAN), gNB, User Equipment of the Software Radio System (srsUE) and Open5GS in a B210-based Universal Radio Peripheral (USRP) testbed with a hardware TPM 2.0 in the UE. The prototype blocks SIM-transplant, rogue-certificate, firmware-tampering, and replay attacks. Across six trials, Battlefield 5G increases average onboarding latency from 1886 ms to 2260 ms, adding 373.4 ms of pre-authentication overhead while preserving standard 5G-AKA, security mode, and packet data unit (PDU) session procedures.

Intrusion detection systems (IDS) and automated systems for detecting and reporting cyber threats, are commonly handled via supervised machine learning methods. Though effective, these models struggle to effectively adapt to new attack types. This study proposes a novel approach by employing a reward-based, dueling Q-learning model for IDS, achieving an average accuracy of 99.68% across multiple attack classes. The proposed model has a dueling network architecture which separates its predictions into value and advantage streams. This has the benefit of improving learning efficiency and stability. The model was trained on the CIC-IDS2018, a benchmark dataset based on real-world intrusion detection scenarios, having multiple attack classes such as DDoS, botnets, and brute-force attacks. Furthermore, Explainable AI (XAI), specifically SHAP (SHapley Additive exPlanations), was also integrated into the training and evaluation process to provide interpretability into the model's predictions.

Cyberattack detection in electric vehicle charging infrastructure is complicated by legitimate post-activation revisions to requested energy and departure time. Charging manipulation attacks can exploit the same interface and variables; therefore, detecting a request change alone does not establish malicious intent. This paper develops a leakage-controlled session-level benchmark that preserves the ordered inputs of real Adaptive Charging Network (ACN) sessions and models legitimate revisions as normal behavior. A fixed pool keeps each generated attack in its source session's split and contains six physically motivated attacks and their coordinated variants. We compare 22 profile-only, transition-aware, and context-stratified model families under common source-grouped folds, attack data, and operating constraints. The proposed Dual-Branch Masked-Autoencoder (Masked-AE) Transition Boost model evaluates whether the current request is normal and whether its producing transition resembles an observed benign update. Its state branch combines masked reconstruction with a radial-basis-function one-class support boundary, while its transition branch combines masked reconstruction with shrinkage covariance distance. Source-grouped five-fold cross-validation selects complete configurations under explicit overall-normal and benign-update acceptance constraints; disjoint normal data then calibrate the final threshold before one test evaluation. The developed dual-branch model provides the strongest robust validation performance while detecting malicious request manipulations without learning to reject legitimate user choices.

Despite the practical relevance of sparse decision-based black-box threats, they have received limited attention in semantic segmentation. To bridge this gap, we adapt the most representative decision-based black-box sparse attacks from the classification domain to serve as baselines, establishing a rigorous benchmark for this underexplored setting. In this context, we demonstrate that one of the existing methods suffers from severe query inefficiency due to its image-centric pixel accumulation, which rapidly exhausts query budgets across the vast image space. To overcome this, we propose SegPAR, a novel decision-based framework that shifts to a class-centric exploration paradigm. Furthermore, to eliminate the misleading feedback generated by standard decision rewards during pixel accumulation, we introduce a novel discrepancy reward. Extensive experiments show that SegPAR significantly outperforms black-box baselines in sparsity efficiency and MIoU reduction, while remaining competitive with white-box sparse attacks. Code is available at \href{https://github.com/KAU-QuantumAILab/SegPAR}{https://github.com/KAU-QuantumAILab/SegPAR}.

Hardware-in-the-Loop validation of automotive software systems generates large multivariate time-series recordings whose manual analysis is time-consuming and often limited to anomaly detection and fault classification rather than root-cause analysis. Although deep learning methods have shown strong performance in fault detection and classification, they usually require task-specific training or retraining when new fault locations, systems, or operating conditions are introduced. They also tend to treat localization as a classification task, without explicitly representing the spatial and functional relationships between fault locations, sensors, and downstream subsystem effects. This limits their generalizability and their usefulness for engineering root cause analysis and diagnosis. This paper proposes a knowledge-graph-guided retrieval-augmented large language model framework for RCA (root cause analysis) and fault localization in automotive HiL data. The method converts raw time-series recordings into compact diagnostic evidence, enriches this evidence with sensor-to-location and propagation knowledge, and retrieves similar historical cases to support the final reasoning step. The LLM is then used as a decision and explanation layer rather than as a direct time-series classifier, producing a ranked fault-location prediction together with an interpretable RCA explanation. The framework is evaluated on two automotive HiL case studies: an ASM gasoline engine and an electric vehicle system. The best-performing model achieves Top-1 accuracies of 90% and 94%, respectively, while recording-level aggregation reaches perfect file-level fault localization in the evaluated subset. These results demonstrate the potential of KG-guided RAG-LLM reasoning for explainable and generalizable HiL RCA.

The dominant paradigm treats AI safety as a property to be instilled during model training via RLHF, DPO, or Constitutional AI. We argue this is structurally insufficient for autonomous agents that execute code, mutate files, send messages, and modify databases. Agent safety should be a runtime contract enforced by the harness, and the contract has two complementary faces. The preventive face blocks dangerous actions before they happen via sandboxes, permission gates, output filters, and trajectory monitors. The evidential face requires verifiable proof that good actions actually happened, gating task submission on hard evidence such as test runs, log captures, file diffs, and citation grounding. We ground the position in four lines of public evidence, with row-level protocols and data released in the supplementary JSON files: a survey of 52 documented AI-agent and LLM safety incidents, a false-completion audit with 31 non-contested core cases plus one disputed illustrative case, a trajectory-schema audit of 12 public agent systems and harnesses, and a title-level audit of all 28,560 papers accepted at NeurIPS, ICML, and ICLR 2023-2025 showing a pooled 8-12x imbalance between training-time and deployment-time publication. Two prior communities that needed to enforce safety, computer security and the experimental sciences, converged on runtime contracts with both preventive and evidential elements; agentic AI is now under the same pressure. We formalize an Agent Trajectory Schema and Evidence Chain, state a compositional gating proposition based on standard monitor composition, and outline a research agenda. The right unit of safety in agentic AI is the trajectory-with-checkable-evidence, not the model.

Bit commitment is impossible to achieve with unconditional security, even in quantum cryptogra- phy. We show that statistically secure bit commitment, satisfying both hiding and binding, can be constructed from hybrid locked physical unclonable functions (HLPUFs), a hardware primitive that combines classical hardware tokens and quantum communication. Our protocol uses these hardware assumptions in a novel and non-trivial way to achieve the first mistrustful two-party cryptographic protocol based on hybrid hardware modules. We prove statistical hiding and binding under natu- ral assumptions on the HLPUF and using a carefully designed challenge generation algorithm as a subroutine of our bit-commitment protocol. The construction also yields the first hardware-based coin-flipping protocol. Our results suggest a new paradigm for secure two-party cryptography in quantum networks, combining rigorous security guarantees with a concrete route toward practical implementation.

Homomorphic Encryption (HE) enables computation on encrypted data without decryption and is a key primitive for privacy-preserving computation in sensitive domains such as healthcare, finance, and government. Its security relies on noise injection, which introduces intrinsic error sensitivity and raises concerns about the fault tolerance of HE systems, as hardware- and software-induced faults can evade traditional detection mechanisms and lead to silent data corruption. In this work, we analyze the sensitivity of HE to bit-level faults, focusing on the CKKS (Cheon–Kim–Kim–Song) scheme widely used for approximate arithmetic in AI and machine learning workloads. We identify homomorphic multiplication as the most error-sensitive operation in practical HE pipelines and characterize how faults propagate and amplify through it, exposing a critical robustness vulnerability and motivating the need for more resilient HE deployments.

For companies of all sizes, data security is a top priority. The chance of unauthorized data access increases as technology develops. To prevent unwanted access to their data, businesses must be proactive. This study examines technology solutions, personnel training, and policy enforcement as methods to prevent unauthorized data access. Data may be protected from illegal access using technological solutions like firewalls, intrusion detection systems, and encryption. Intrusion detection systems notify the administrator when suspicious behavior is found, while firewalls serve as a protective border between the internal network and the internet. Even if data is intercepted, encryption makes sure it is safe. Another effective method of avoiding unauthorized data access is employee education. Employees must be taught how to spot hazards like phishing emails and shady websites and react to them. Additionally, they should be taught the right way to utilize passwords and other security precautions. To secure data, organizations should create and implement policies. Policies should set out appropriate data and system use guidelines and provide repercussions for noncompliance. Policies should be evaluated regularly to ensure that they are current and useful. Businesses may prevent unwanted access to their data by installing technology solutions, training staff, and enforcing regulations. Organizations may reduce data breach risk and maintain regulatory compliance by taking these precautions. ...

Homomorphic Encryption (HE) is a privacy-preserving encryption paradigm that enables computation directly on encrypted data without requiring decryption. In this paper, we study errors in fully homomorphic encryption (FHE) computations, with a particular focus on server-side homomorphic multiplication in the unoptimized CKKS (Cheon–Kim–Kim–Song) scheme. We show that both the timing and the location of errors in the ciphertext components \(c_0\) and \(c_1\) have a significant impact on the correctness of the final FHE output.

Existing vision-language model (VLM) backdoors are usually treated as static vulnerabilities: one-to-one and N-to-N attacks bind one or more triggers to a finite set of targets before victim training. This assumption substantially underestimates the threat. We show that a single poisoning phase can implant a programmable backdoor into a VLM, allowing an attacker to choose previously unseen target-caption semantics at inference time and synthesize corresponding stealthy triggers on demand. Unlike fixed-mapping attacks, the proposed any-to-any caption-control paradigm decouples post-training target selection from poisoning, enabling dynamic control of target captions without retraining the VLM. Our method has two components. First, a heuristic poisoning strategy exposes the model to diverse trigger-caption pairs, encouraging it to learn a general trigger-as-instruction rule rather than memorize a specific backdoor pattern. Second, a feature-space trigger steganography method maps any attacker-specified target caption to a stealthy visual trigger, implemented as either a norm-controlled perturbation or a non-semantic patch. Once inserted into arbitrary images, these triggers cause the poisoned VLM to generate outputs semantically aligned with the chosen target caption, even when the target was unseen during poisoning. Extensive experiments show that our attack achieves high any-to-any caption-control success rates, preserves clean model utility, and remains effective under several classical backdoor defenses.

The Model Context Protocol (MCP) has become the de-facto interface for connecting LLM agents to enterprise tools, and adoption has been explosive: within a year, large organizations went from zero to dozens of internally built MCP servers. That speed created a governance crisis. Each team implemented authentication independently – some with no auth, some with API keys, some with full OAuth – producing a fragmented landscape with no consistent way to authorize callers, track who did what, or offboard a departing employee across the fleet. This paper reports an industry deployment that resolves the crisis with a centralized MCP gateway: a single aggregation, governance, and authentication layer that fronts every downstream MCP server. We make four contributions grounded in production experience. First, a two-axis authentication model crossing persona (interactive user vs. automated non-user) with credential type (no-auth, static/dynamic API key, PKCE, client credentials, platform app-context). Second, a gateway authentication layer supporting three enterprise SSO grants and three token-provisioning models: Bring-Your-Own-Token, Generate-Your-Own-Token, and delegated OAuth via RFC 8693 token exchange. Third, three end-to-end identity flows – User-to-OAuth2, Non-user-to-Service-Account, and User-to-Service-Account – composing client, gateway, and server. Fourth, the deployment evolution from CDN/WAF/edge perimeter to private MCP tunnels and enterprise-wide connectors. The architecture is in production, fronting dozens of MCP servers across web, desktop, custom-SDK, and low-code clients.

Polynomial multiplication is the most resource-, time-, and energy-critical operation in lattice-based Post-Quantum Cryptography (PQC) and Fully Homomorphic Encryption (FHE) schemes. Lattice-based PQC schemes such as Kyber and Dilithium have already been standardized, while lattice- based FHE schemes such as BGV, BFV, and CKKS are widely recognized as leading candidate in FHE area. Barrett Modular Multiplication (BMM) for polynomial multiplication is widely adopted in PQC and FHE hardware accelerators due to its hardware friendly nature and efficient modular reduction capabilities. However, Side-Channel Attacks (SCAs) and Hardware Trojans may introduce intentional faults, while aging and various other factors can cause unintentional faults. These faults may target the BM M unit, one of the most critical components of PQC and FHE infrastructures, potentially leading to information leakage and compromising system security. In this paper, we employ a Statistical Reduction Monitoring (SRM) method to protect the BM M unit against such adversarial conditions. The proposed approach incurs minimal hardware overhead while providing efficient detection of both random and bur

Expert parallelism (EP) is a common strategy for serving large Mixture-of-Experts (MoE) models across multiple GPUs by distributing experts among devices. Router decisions then determine both which experts process each token and which GPUs execute the resulting work. This procedure exposes a supply-chain attack surface in the serving schedule. We introduce Load Hijack, in which a malicious model provider modifies only a checkpoint's router weights, distributes the poisoned checkpoint, and retains a private trigger. When the trigger appears, the poisoned router concentrates token-to-expert assignments on experts co-located on one GPU. The resulting load makes that GPU a straggler and forces peer devices to wait, while routing on ordinary inputs remains near the clean reference. We find this conditional behavior difficult to achieve because an objective that rewards target-expert use on triggered inputs can also bias ordinary-input routing toward the same experts. To resolve this conflict, Load Hijack employs a three-stage optimization procedure that produces strong trigger-dependent concentration while keeping ordinary-input routing close to the clean reference. Across three MoE families and four corpora, Load Hijack directs 92.3% to 95.6% of triggered token assignments to the target experts. In live EP serving, triggered traffic produces 1.43x the time-to-first-token and 0.86x the throughput measured under ordinary traffic. These results show that poisoned routers can act as trigger-controlled device schedulers and motivate checkpoint audits of routing and runtime load.

Large Language Models (LLMs) have undergone a shift from stateless conversational interfaces to autonomous agents capable of multi-step planning, tool invocation, code execution, and maintaining persistent memory. When these agents operate with real-world privileges—calling APIs, modifying files, and querying databases—a compromised reasoning step can trigger unauthorized data access, irreversible state changes, or cascading failures, yet the security research community has not kept pace. To quantify the state of the field, we conducted a systematic literature review under PRISMA 2020 guidelines across six databases, screening 743 records and retaining 85 papers (2023–2025) on agentic LLM security. Attack research outpaces defense work by 3.9:1. Perception-layer vulnerabilities (prompt injection, jailbreaking, adversarial perturbations) dominate, accounting for 66% of papers, while action-layer vulnerabilities (tool misuse, code injection, sandbox escape) appear in only 4.7%, misaligned with real-world risk. Code execution security accounts for 3.5%, and tool-augmented agents 12%. We contribute a four-layer taxonomy mapping 13 vulnerability types across perception, brain, action, and interaction layers, and identify seven open problems centered on containment. Agentic LLM insecurity stems from architectural coupling, where weak isolation allows vulnerabilities to propagate across layers.

Branch predictors improve instruction-level parallelism in modern processors and are commonly modeled using saturating counters. However, classical saturating counters are deterministic and thus vulnerable to side-channel attacks: an attacker can manipulate the counter state and infer the branch direction of a victim process. Probabilistic saturating counters (PSCs) have been proposed to mitigate this leakage by randomizing counter updates, but existing evaluations are mainly empirical. In this paper, we give a formal analysis based on differential privacy (DP): we model PSCs and the corresponding Prime+Probe attack strategies as probabilistic Moore machines, derive optimal attack strategies, and quantify the attacker's distinguishing power through DP. Our DP guarantee applies to the PSC primitive under the Prime+Probe observation model; end-to-end security for a full branch predictor under repeated or adaptive attacks is an important direction for future work. We then synthesize parameters for an enhanced PSC that satisfies a target pure DP guarantee. To evaluate utility, we derive the stationary misprediction rate and validate the theoretical predictions on benchmark programs. Compared to deterministic and existing probabilistic saturating counters, the synthesized PSCs provide formal security guarantees while preserving competitive prediction performance.

Machine-learning intrusion-detection studies commonly emphasize predictive accuracy while treating explanation generation as a computationally free post-processing step. This study jointly evaluates predictive effectiveness, explanation cost, local explanation stability, and selective explanation for binary Internet of Things (IoT) intrusion detection. A leakage-safe CICIoT2023 corpus was constructed using exact 39-feature hashes, non-finite-value handling, exact-feature deduplication, conservative label-collision removal, and deterministic hash-level partitioning. Logistic Regression, Decision Tree, Random Forest, and XGBoost were evaluated on natural and balanced test distributions. TreeSHAP cost was measured, stability was assessed under prediction-preserving perturbations, and validation-calibrated policies were used to allocate explanation workload. XGBoost provided the strongest overall predictive profile, while Random Forest produced the lowest false-positive rate. At 5,000 samples, TreeSHAP required 700.759 s for Random Forest and 1.471 s for XGBoost. Random Forest showed the strongest overall base-level explanation stability; XGBoost retained high rank and directional consistency but showed greater top-feature turnover and attribution-magnitude drift. On the balanced test, about 90% false-negative explanation coverage permitted 28-32% compute savings, while about 95% coverage permitted 15-23% savings. Savings were much smaller under the attack-heavy natural prevalence. These results show that operationally useful explainable IoT intrusion detection depends on predictive quality, explanation cost, local stability, workload prevalence, and selective invocation rather than detection accuracy alone.

Large Language Models (LLMs) demonstrate impressive capabilities across many applications but remain vulnerable to jailbreak attacks, which elicit harmful or unintended content. While model fine-tuning is an option for safety alignment, it is costly and prone to catastrophic forgetting. Prompt optimization has emerged as a promising alternative, yet existing prompt-based defenses typically rely on static modifications (e.g., fixed prefixes or suffixes) that cannot adapt to diverse and evolving attacks. We propose Dynamic Deep Prompt Optimization (DDPO), the first jailbreak defense based on deep prompt optimization. DDPO uses the target LLM's own intermediate layers as feature extractors to dynamically generate defensive embeddings via a lightweight multilayer perceptron. These tailored embeddings are then injected into a subsequent intermediate layer, enabling an input-dependent defense without modifying the LLM's weights. This design ensures high adaptability with minimal computational overhead. Experiments on a diverse set of models and attacks demonstrate that DDPO significantly outperforms static prompt optimization methods, particularly on weakly aligned models and when handling semantically ambiguous benign prompts, successfully distinguishing them from genuinely harmful requests.

In this paper, we present an optimized implementation of Hamming Quasi-Cyclic (HQC) on the ARM Cortex-M4. We optimize (i) the polynomial multiplication and (ii) the support expansion in fixed-weight sampling, and (iii) propose an optional caching strategy that reuses the public transforms and hash recomputed under a fixed key. For the polynomial multiplication, the fixed-constant multiplications in the Frobenius additive FFT (FAFFT) butterfly spend nearly half of their instructions on VMOV data movements between general-purpose and floating-point registers rather than arithmetic. Because minimizing the XOR count alone can increase the total instruction count, we propose a dirty-aware register-allocation policy and an XOR-operation reordering that reduce the VMOV count by up to 48.1% while leaving the XOR count unchanged. We apply these to a multiplication that combines prior FAFFT-CRT methods, and for HQC-1 we further find a 34% sparser FAFFT modulus that lowers the CRT reconstruction cost. For fixed-weight sampling, we rewrite the support expansion with predicated execution and 4-way unrolling, lowering the per-word cost of its inner loop from 22 to 6 cycles while remaining constant-time. On the NUCLEO-L4R5ZI board, our implementation reduces key generation, encapsulation, and decapsulation by up to 33.1%, 34.6%, and 29.8% over the faster of the two prior state-of-the-art implementations, and the optional caching yields a further reduction of up to 32.7% and 18.9% for encapsulation and decapsulation.

Electronic health-record interoperability is a boundary problem: legacy systems, generative models, terminology services, identity systems, and human reviewers may each expose rich internal states, while operational exchange requires a narrow shared interface of typed claims, bounded uncertainty, provenance, and explicit admission or abstention. This paper details a mathematical and engineering architecture for that interface. The organizing idea is the logit boundary: a discovery model may propose pre-threshold scores over a local categorical decision, but a deterministic judgment substrate decides whether the proposal is admissible, requires review, or must be quarantined before any Fast Healthcare Interoperability Resources (FHIR) transaction is constructed. The resulting Geometric Belief Interface (GBI) combines finite boundary semantics, local Dirichlet evidence, cellular-sheaf and mapping-cone diagnostics, advisory geometric audit charts, and a Decentralized Cryptographic Sheaf-Enclave (DCSE) protocol sketch for fail-closed deployment. The framework does not establish clinical truth, global representation alignment, or end-to-end safety; it defines certificate-producing checks at a model-to-system boundary. A companion frozen synthetic benchmark, GBI BoundaryBench v0.1, evaluated Qwen3-4B-Instruct-2507 on 256 held-out tasks across three evidence modes (768 canonical executions). All executions completed, but none produced an output accepted by the benchmark contract: 369 were rejected during safe parsing and 399 during schema validation, yielding zero coverage and deterministic quarantine. This empirical result is deliberately narrow - one 4B open-weight model under one frozen interface - and is reported as evidence about the admission boundary, not as a general claim about LLM capability or clinical safety. A Julia appendix verifies numerical certificates using standard libraries.

Large Language Models are increasingly integrated into web applications through chatbots, tool-calling pipelines, and agentic workflows. In these systems, user input may influence not only generated text, but also backend actions such as database queries, HTTP requests, file operations, template rendering, or API calls. This paper introduces LLM-mediated web attacks, a class of attacks in which attacker-controlled input is transformed by an LLM-integrated application and then reaches traditional web-application sinks. We systematize this attack surface through representative LLM2X variants, including LLM2SQLi, LLM2XSS, LLM2SSTI, LLM2CommandInjection, LLM2IDOR, LLM2CSRF, LLM2XXE, and LLM2SSRF. Our analysis shows that the LLM usually does not create the underlying vulnerability itself; rather, it acts as a mediation layer, and in some tool-enabled settings as a confused deputy, carrying attacker influence into components that trust model-generated or model-influenced content. As an experimental case study, we implement TicketOracle, a Flask-based LLM-integrated web application for evaluating LLM2SSRF across five attack scenarios and seven LLMs. Our results show substantial variation in susceptibility across models, suggesting that exploitation depends both on insecure application architecture and model-specific behavior. We conclude with mitigation strategies across the prompt, model, application, and network layers.

Streaming language-model output creates an enforcement boundary: a control that detects a prohibited pattern after releasing its completing chunk cannot recall it. We study a production policy in which each ordered family is the conjunction of two regular-language predicates. Incremental matching is classical. The problem is exact composition at release time across arbitrary chunk partitions, including end-of-prefix word boundaries that can change on extension. We define an ASCII-explicit policy grammar, compile each predicate to a persistent nondeterministic finite automaton (NFA), distinguish stable from provisional assertion state, apply document-order family priority, and check the decision before releasing each chunk. We show that the resulting monitor is release-boundary equivalent to an absorbing cumulative oracle for every policy in the declared grammar. Production Python and TypeScript implementations were evaluated on 101,653 partitioned cases; a public surrogate added 100,345 cases. Both campaigns produced zero oracle, cross-runtime, or intended-family mismatches. In a frozen neutral-output profile, the memoized incremental and native-regex cumulative slopes at 64-character chunks were 0.973 and 1.976. At 16,384 characters the incremental median was 30.2 ms versus 96.6 ms for native cumulative scanning at that chunk size. Native regex remained faster at 512-character chunks (12.4 versus 29.4 ms), exposing the constant-factor crossover rather than hiding it. A shared per-stream cache cap and 129-symbol alphabet bound optimization state; the campaign peaked at 364 of 4,096 without bypass. The result is policy conformance for a deterministic backstop, not evidence of semantic safety or policy completeness.

The rapid advancement of Large Language Models (LLMs) has facilitated their ubiquitous integration into various domains, leading to widespread adoption. However, this escalating trend has introduced significant security vulnerabilities, necessitating the identification and mitigation of flaws arising from malicious exploitation. Red teaming assessments, conducted to evaluate model robustness through diverse adversarial inputs, are essential for exposing security risks and implementing countermeasures. Currently, red teaming is performed either manually by experts or automatically using predefined attack datasets. Nevertheless, manual testing remains time-consuming, while existing automated methods suffer from limited creativity due to their inherent dependency on fixed datasets. In this study, we propose an automated, human-independent, and adaptive approach leveraging GFlowNets to identify LLM vulnerabilities by utilizing one large language model to test another. Within this framework, an attacker model is trained against a specified victim model to perform automated red teaming and provide a quantitative robustness score. This research aims to generate more effective adversarial attacks in English compared to existing benchmarks and, as a novel contribution to the literature, introduces a model capable of generating attack inputs in the Turkish language.

Digital watermarking has emerged as a critical technique for provenance and copyright attribution in AI-generated imagery, yet its robustness against realistic, model-agnostic removal attacks remains poorly explored. Existing attacks either succeed only against specific generative models or achieve removal at the cost of severe visual degradation. In this paper, we propose MarkNull, a model-agnostic watermark removal attack via on-manifold latent manipulation. MarkNull is grounded in a key observation: watermarked images exhibit a strong statistical dependency between the generated latent representation and the embedded initial noise. To quantify this dependency, we introduce the Noise-Latent Alignment Score (NLAS) and formulate an optimization objective that selectively decorrelates the latent representation from the embedded watermark while preserving semantic fidelity. Extensive evaluations across different categories of watermarking paradigms, including post-hoc, fine-tuning-based, and initial-noise-based schemes, demonstrate that MarkNull reduces average bit accuracy to 53.14%, approaching random-guessing (50%), without perceptible image degradation. To further improve scalability, we propose MarkNull-A, an amortized, optimization-free variant that distills the attack into a single forward pass, achieving 0.50 s/image with modest computational overhead. Notably, our attacks successfully compromise Google's SynthID-Image system while preserving high visual quality and transfer effectively to video watermarking. Finally, we present an attack detection mechanism as a defensive counterpart to MarkNull and MarkNull-A, highlighting the necessity of developing watermark designs resilient to model-agnostic latent-space attacks.

This paper introduces fully working and affordable approach to data recovery from automotive SRS (airbag control) module after permanent failure of its data storage memory chip. All essential information was successfully extracted with 100% success rate from a real SRS module after severe accident which resulted in both its Processor and Flash EEPROM chip electrically damaged with visible holes in their packages. Further investigation revealed that the damage was likely caused by overvoltage and subsequent high-current-flow through silicon die of chips. This not only caused some internal metal wires to melt and evaporate but also caused partial evaporation of silicon substrate and the created cavity was filled with melted gold from a bonding wire. This reveals the nature of the failure with local temperatures exceeding 3000{\deg}C. Despite to the fact that the on-chip EEPROM array was only a few hundred micrometers away from the damaged area, it did not lose any single bit of information. This proved the approach to be resilient for successful data recovery from severely damaged devices in the future. With the donor unit containing the recovered data, standard bench-top EDR extraction was performed using Collision Sciences CrashScan, yielding a full EDR report. The forensic significance of the recovered data, the technical novelty of the methodology, and the complete accident reconstruction are presented and discussed. Until now, if because of an accident the memory chip was damaged, there was no feasible or affordable solution for full data extraction or recovery. With the use of innovative and proprietary sample preparation and imaging techniques, it became possible to extract and verify all the data stored inside chips after permanent mechanical, electrical or fire damage. This could help not only automotive but also medical, aviation and aerospace industries.

Hardware assurance relies on scanning electron microscopy (SEM) to verify nanoscale structures, but assembling the large, high-quality datasets required for automated analysis is impeded by time-intensive acquisition and strict intellectual property (IP) constraints on proprietary designs. We propose a privacy-preserving pipeline that secures IP by heavily distorting the functional design while generating a visually realistic synthetic dataset from a small set of initial examples. A StyleGAN first learns the distribution of hardware layout masks to generate novel, macroscopically varied structures. Subsequently, a conditional GAN (Pix2PixHD) translates these masks into realistic SEM images that preserve authentic textures and noise. The primary finding of this work is that a segmentation model trained exclusively on this synthetic data not only demonstrates a successful "sim-to-real" transfer to real images but also outperforms a baseline model trained on the limited real dataset. Because the underlying synthetic layouts are demonstrably novel and reproduce none of the specific proprietary routing of the original design, deploying the final segmentation model mitigates the risk of exposing sensitive IP to attacks like gradient inversion and membership inference, providing a highly secure, high-performance solution for hardware assurance.