Papers

Modern software systems require earlier and more scalable vulnerability severity assessment to reduce exposure to high-impact security flaws. Security analysts typically assign CVSS scores, but this manual triage does not scale with the growth of disclosed vulnerabilities and often depends on cloud LLM services that raise confidentiality concerns. This paper presents an industrial case study on predicting CVSS v3.1 scores directly from vulnerable C/C++ snippets using in-context learning with locally deployable, open-source LLMs. We compare proprietary data with the Big-Vul dataset, showing sufficiently aligned CVSS distributions to justify Big-Vul as a proxy for industrial data when constructing prompt-based testbeds. We then vary in-context configurations and model parameters, evaluating CodeLlama2-7B, CodeLlama2-13B, Mistral-7B, gpt-oss, and GPT4o-mini using mean squared error (MSE) and feasibility metrics. Our results show that medium-sized open-source code models, particularly CodeLlama2-7B, can approximate the best cloud performance for CVSS regression when guided by lightweight, output-constraining prompts, offering a practical, privacy-preserving building block for severity triage in industrial settings.

Autonomous response has evolved into a timing-critical challenge rather than solely a matter of detection accuracy. In recent intrusions, the interval between initial access and the first lateral movement has been observed to be as short as 27 seconds, a window that precludes any human-in-the-loop workflow. This paper presents a closed-loop framework that detects and blocks attacks in software-defined networks without operator involvement, evaluating its performance against this stringent temporal constraint rather than relying exclusively on detection accuracy. An automated data pipeline collects IP flows and aggregates them into labeled training data, while a prevention module selects and trains candidate classifiers and issues blocking rules directly to the SDN controller. In a SYN flooding denial of service case study, the deployed K-Nearest Neighbors classifier achieved an F1 score of 96.7% and the cycle from flow availability to enforced block completed in 21 seconds, below the fastest breakout time reported to date.

We show that the CRT-FHE scheme of Pradhan et al.\ is insecure for laws within its assumed error distribution range. The secret key follows from the public key by a single ring inversion whenever the public multiplier is a unit. The plaintext is recovered from any ciphertext under such a law without the secret key, for every multiplier, giving chosen-plaintext advantage $1/2$. We further show that the transformation from ordinary Ring-LWE to CRT-RLWE does not preserve the error distribution, so it does not establish that CRT-RLWE is at least as hard as Ring-LWE. One mechanism underlies both. The Chinese remainder theorem (CRT) function is reduced modulo $p_1p_2$ while its output is used modulo a coprime modulus $q$, so under every zero-preserving section an error in $p_2\R$ encodes to zero. The law $p_2B_1$ is so confined, meets the stated conditions, and decrypts correctly. Confinement is not a weakness of scale: scaling any baseline law by $p_2$ leaves its ordinary Ring-LWE problem exactly equivalent, while the reduced encoder destroys every error it produces. The reduction discrepancy is a multiple of $p_1p_2$ and not of $q$, so the small-error premise of the proof cannot remove it, and at the reported parameters a single error coefficient refutes the identity while satisfying that premise. The centered binomial $B_2$ separates the coefficient laws at total variation distance $3/8$, and at the reported dimension that distance between the induced polynomial laws is exponentially close to one.

We present an autonomous cryptanalysis workflow in which agents generate, test, and refine hypotheses before human review. The autonomous stage returns reproducible candidates with exact witnesses, controls, code, and run records. A researcher then decides whether the evidence establishes a break, defect, or coverage gap. Two failure modes recur. First, a public algebraic map or input representation erases or exposes a relation that a construction must hide. Examples include multiplication by zero, boundary coefficients of a polynomial product, quotients, characters, Schur squares, and variable-length byte encodings without boundaries. Second, a simulator, error law, or parameter certification uses a distribution different from the one claimed. Several targets fail in both ways. Every result has an exact witness and a discriminating control; every stated boundary has a proof. Three further targets yielded no attack but support narrower guarantees than a generic reading suggests. Eight published constructions fail at stated parameters or claims. A Ring-LWR commitment opens to every message with probability one. One ciphertext reveals two middle-product encryption rows. A lattice e-voting protocol loses receipt-freeness. A permutation-recovery attack against updatable encryption extends by linear algebra to the old decryption key. An explicit normal basis splits a degree-63 instance into seven degree-nine instances. A signature hash outside the lattice setting maps two printable equal-length messages to the same digest. A rerandomisable scheme's accept bit is a threshold oracle on its decryption noise. Separately, a group-ring decision claim and a multivariate MinRank hardening fail at the assumption or accounting level rather than as complete construction breaks. Each failure occurs one level above its supporting assumption.

CWE labels in the National Vulnerability Database (NVD) are widely treated as ground truth for vulnerability search, scanner evaluation, benchmark construction, learning-based security tools, and vulnerability prioritization. Yet their reliability has not been systematically measured at scale, despite growing concerns about NVD's enrichment backlog and anecdotal reports of inaccurate, ambiguous, or missing labels. This paper presents a large-scale, code-semantics-grounded measurement of CWE labeling quality in NVD. We build CWEAgent, a validated auditing instrument based on seclometry, a structured representation of vulnerability semantics that captures the root cause, trigger condition, violated security property, exploit mechanism, and impact of vulnerable code. On a manually curated benchmark of 100 open-source CVEs, CWEAgent achieves 85% top-1 accuracy and 92% ambiguity-aware accuracy. Applying CWEAgent to 15,556 open-source CVEs disclosed from 2017-2026, we find that only 49.70% of NVD CWE labels exactly match the code-grounded label. Another 31.37% are defensible alternatives under taxonomy ambiguity, while 3.63% are evidence-inconsistent likely errors. Label reliability varies sharply by assigning organization and weakness type, and apparent project- or language-level differences are largely composition effects of those underlying weakness types. Evidence-inconsistent labels have also increased over time. Through manual review of 434 confirmed mislabels, we identify six recurring error patterns, showing that CWE noise is a structural problem in vulnerability metadata rather than isolated annotation mistakes.

Quantum Key Distribution (QKD) enjoys information-theoretic security, yet the most damaging attacks against deployed systems exploit the receiver, where the key bit is encoded in which one of a pair of never-identical detectors clicks. The minimal receiver, one rotatable polarizer and one threshold detector, removes that attack surface, and single-detector BB84 demonstrations already run sampled error estimation; the structure of its zero-probability error subensemble, however, has remained uncharacterized. We characterize exactly that structure, introducing a deterministic impossible-event certificate: a click behind a polarizer set orthogonal to the transmitted state has probability exactly zero on an ideal channel, so a single occurrence is a probability-one witness of disturbance; and, since loss deletes clicks and never creates them, the certificate is loss-robust. We prove it sound but incomplete over three polarization states, and show that the four BB84 states close the gap: a fixed-basis intercept-resend attack yields an ideal trip probability of $1/4$ per orthogonal round ($\eta/4$ observed at detection efficiency $\eta$), independent of the interception angle. An illustrative finite-size budget yields 256 retained bits from $\approx 62{,}000$ transmitted rounds at $\eta = 0.1$; under realistic detector noise ($q_0 = 10^{-6}$ per opened gate), each trip retains $\approx 12$ bits of evidence at a sub-percent honest false-abort probability per session. The core ideal trip-probability predictions are numerically verified on the Qiskit circuit simulator, via a released, seed-fixed implementation. Overall, by endowing the minimal-detector receiver of polarization QKD with a conclusive, loss-robust disturbance alarm, our solution lowers the hardware entry cost of security-monitored QKD, hence fostering its adoption at the cost-sensitive network edge.

Software systems have traditionally been organized around applications where human users act as principal decision-makers. Recent developments in agentic capabilities alter this paradigm: software agents now autonomously translate high-level goals into structured tasks, orchestrating tools, services and sub-agents to execute complex workflows. This evolution gives rise to an agent economy where these autonomous agents capture real economic value. However, the infrastructure required to support the agent economy fails across three critical dimensions: the absence of persistent identity infrastructure prevents systemic accountability in agentic workflows; capability claims remain self-declared not backed by verifiable execution evidence; and the disconnect between creator identities, agent performance, and project value hinders the economic valuation of agents as assets. While existing registries provide naming and discovery, unifying these features around a persistent identity anchor remains largely unaddressed. TessIndex is a capability-verified identity system for agent primitives that utilizes a dual-plane architecture: the blockchain records compact commitments for identity, ownership, and verification, while centralized servers maintain dynamic metadata for discovery, commerce, and reputation. It establishes: persistent identities across agent primitives to enforce systemic accountability in autonomous workflows; a predicate-based verification process replacing self-declared claims with cryptographic capability proof; an identity infrastructure that links agent performance to both project and creator identities while capturing value through tokenization. Ultimately, TessIndex serves as an integrated infrastructure that binds an agent's existence across capabilities, execution, and reputation into a single persistent identity.

Agent skills extend coding agents with task-specific instructions, scripts, and resources, but they also create a trusted instruction channel that can be abused beyond conventional security attacks. This paper studies token amplification through skill injection: an economic resource-abuse threat in which a malicious skill causes an agent to consume substantially more tokens than needed for normal task execution. We present SkillBloat, a two-phase framework that first screens a library of diverse attack-type conditions across multiple amplification mechanisms and then refines the strongest candidate through LLM-guided full-document skill rewriting. Evaluated on a real-world skill benchmark, SkillBloat achieves 5.4184x-10.1455x average best amplification across multiple coding-agent target configurations. An ablation shows that the second-stage refinement loop consistently improves average best amplification over Phase 1 attack-type screening alone, demonstrating that iterative optimization provides additional benefit beyond initial attack-type selection. These results show that skill ecosystems expose a practical resource-amplification attack surface that is orthogonal to existing security-oriented skill poisoning.

Locally deployed Large Language Models (LLMs) via inference engines such as Ollama run without the moderation and abuse detection present in API-served models. Therefore, the safety of LLMs depends on the defense mechanisms used, and their effectiveness depends on the assumptions on which they were designed. This paper does an audit of defense mechanisms under jailbreak attacks on locally deployed models. Some defenses provide formal guarantees (SmoothLLM, Erase-and-Check, Sequential Monitors), while others rely on empirical detection results (Semantic Smoothing, Self-Denoised Smoothing, Perplexity Filtering). Instead of merely observing that defenses fail, we trace each failure back to the specific assumption: for every defense, we extract the condition it relies on, derive the empirical pattern a violation should produce, and test that prediction on six open-weight models (14B to 35B parameters) with a corpus of 100 jailbreak prompts taken from more than 40 public sources, totalling 13,800 evaluation records.

Bangla large language model (LLM) safety is difficult to evaluate with English-centric or standard-script benchmarks because Bangla users routinely write across scripts, spellings, code-mixed forms, and regional registers. This paper presents BanglaVeilGuard, a compact Bangla-first safety benchmark and lightweight prompt guard for six language forms: standard Bangla, Romanized Bangla, Banglish, code-mixed Bangla–English, noisy Bangla, and dialectal Bangla. The benchmark contains 2,366 quality-filtered prompts and a held-out 354-prompt evaluation split spanning unsafe, safe, and safe-sensitive requests. BanglaVeilGuard uses non-destructive multi-view normalization with a prompt-risk classifier and thresholded pre-generation gate, allowing it to screen prompts for heterogeneous target models without changing their weights. Across target-model families, guarded runs reduce attack success under deterministic response scoring from 93.8–100.0% to 6.3% for Claude Opus 4.8, BanglaLLama, and TituLLM; TigerLLM-1B with BanglaVeilGuard achieves 78.2% accuracy with 8.8% ASR. The prompt guard also attains 88.5% unsafe recall, substantially above the evaluated prompt-only guard baselines. The main remaining cost is over-refusal on dialectal and noisy benign prompts, revealing a concrete safety-helpfulness frontier for Bangla LLM deployment.

As machine learning (ML) models are increasingly deployed in high-stakes environments, explainable AI (XAI) methods like SHAP and LIME have become essential for regulatory compliance and trust. However, the current auditing paradigm relies on an implicit "chain of trust" where third-party auditors are assumed to be trusted. Recent research demonstrates that this assumption is flawed and adversarial auditors can manipulate XAI explanations through manipulation attacks such as output shuffling or scaffolding out-of-distribution (OOD) to conceal model biases while maintaining high prediction accuracy aiming for fairwashed explanation. In this paper, we introduce a novel defense framework, ExplainGuard, that leverages a Zero-Trust architecture (ZTA) design to be incorporated within the XAI explanation supply chain and ensures the integrity of the generated explanation. This framework would help us to replace the ambiguous default assumption of "auditor is trustworthy," with a continuous "verify-then-trust" approach. Our design architecture establishes a Policy Decision Point (PDP) that enforces three distinct pillars of verification before any explanation is released to the user: (1) asset integrity via behavioral fingerprint to detect model substitution, (2) semantic validity using axiomatic consistency checks to reject mathematically impossible explanations, and (3) feature faithfulness verification utilizing a ranking stability approach with minimal computational overhead. Finally, we evaluate how ExplainGuard can effectively neutralize state- of-the-art explanation manipulation attacks while transforming the auditing process into a verifiable operation.

Semantic communication has emerged as a promising paradigm for next-generation wireless systems by transmitting high-level semantic features rather than raw bits. However, collaborative devices and multimodal transmission increase privacy risks because sensitive information may leak through inter-device semantic fusion and cross-modal representations. To address this issue, we propose a privacy-preserving semantic communication framework for wireless edge networks. Leveraging a vision-language model (VLM), the framework extracts textual semantics from images and identifies privacy-sensitive entities using a privacy database maintained only at the edge server. Before image transmission, each device removes the identified private regions while preserving useful semantic content. The server then reconstructs the removed regions from the received masked images using textual embeddings and VLM-based semantic priors. To protect textual information, we design an encrypted semantic-channel transceiver using physical-layer keys generated from reciprocal wireless channels, without pre-shared keys. We also introduce a semantic information bottleneck to suppress redundant information across multiple devices. The framework is evaluated against a strong model-aware adversary that can intercept wireless transmissions and access edge-device model parameters but not server-side data. Simulation results show that the proposed method reduces privacy leakage by more than 50% compared with a semantic communication scheme without privacy protection, while the authorized server achieves a 48% improvement in perceptual reconstruction quality over the adversary. The estimated mutual information between transmitted representations approaches 0 bit, indicating effective suppression of cross-device semantic redundancy.

Conventional Quantum Key Distribution (QKD) requires the transmission of qubits proportional to or exceeding the length of the key, as protocols such as BB84 transmit more qubits than the final key size due to basis sifting and privacy amplification. Since quantum networks are still in their infancy and have limited capacity, this overhead puts significant pressure on network resources. To address this issue, we propose a Multi-Qubit Greenberger–Horne–Zeilinger (GHZ) State-based QKD scheme that reduces the number of qubits transmitted over the quantum channel. The proposed method transmits one GHZ qubit between endpoints and reuses the resulting entanglement to convey multiple classical key bits with the help of Quantum Non-Demolition (QND) measurements. Under the stated assumptions on authenticated classical communication, local reset verification, and bounded-error QND discrimination, one can transfer $L$ classical bits by generating an (L+1)-qubit GHZ state and transferring one qubit to the remote party. We verify correctness using the NetSquid quantum network simulator: the protocol achieves 100% raw-key fidelity for keys of length up to 12 bits under both ideal conditions and depolarizing noise up to p = 0.005 per round. We further show that the proposed QKD algorithm can be extended to multi-party QKD and server-client deployment. The proposed scheme offers a transmitted-qubit-efficient, noise-tolerant alternative for bandwidth-limited quantum networks.

Modern cyber–physical, Internet-of-Things (IoT), wearable, and edge systems increasingly require trust in three distinct entities: the human requesting access, the physical device executing the computation, and the hardware function that is permitted to operate. These requirements are usually studied in separate communities. Biometrics establish human identity but remain vulnerable to presentation attacks, intra-user variability, template leakage, and limited revocability. Physical unclonable functions (PUFs) provide device-specific physical identity and on-demand secret derivation, yet must address environmental instability, helper-data exposure, side channels, and modeling attacks. Hardware obfuscation and logic locking condition correct circuit behavior on an activation secret, but face oracle-guided, approximate, structural, removal, and physical attacks. This survey develops a unified human–device–function view of trust. We first decompose each primitive into its complete processing chain and identify the corresponding security assumptions, implementation mechanisms, and evaluation metrics. We then formalize pairwise compositions—biometric–PUF, PUF–obfuscation, and biometric–obfuscation—and a three-way architecture in which correct functionality is bound jointly to an authorized user and a genuine device. Particular attention is given to biometric key reconstruction, PUF stabilization and modeling resistance, logic-locking attack evaluation, cross-layer error propagation, enrollment trust, key lifecycle, and interface leakage. The survey concludes with a taxonomy and research agenda for revocable human–device credentials, compositional security, leakage-aware integration, reconfigurable activation, and standardized end-to-end evaluation.

TEE-shielded inference keeps sensitive state in a trusted execution environment (TEE) while offloading linear algebra to an untrusted accelerator. Wang et al., in Game of Arrows (USENIX Security 2025), showed that five widely adopted lightweight defenses preserve vector directions and introduced ArrowMatch to exploit this leakage. They then proposed ArrowCloak, which adds a different multiple of one shared mask direction to each vector and bases its weight-recovery hardness argument on Learning with Errors (LWE). ArrowCloak successfully reduces ArrowMatch to near-black-box levels. In this paper, we revisit ArrowCloak from cryptographic and structural perspectives. Its LWE formulation does not by itself establish standard LWE hardness: the reduction direction, quantized arithmetic, and joint instance distribution do not meet the required conditions. Reusing one mask direction leaves a recoverable rank-one component across the released matrix. We exploit this structure with our proposed attack, an end-to-end, query-free recovery attack. Given a public checkpoint and the obfuscated weights, the attack removes the masking subspace, recovers the hidden one-to-one correspondence, and reconstructs protected weights without transformation secrets, victim queries, or fine-tuning data. Across six model-task pairs spanning classification, segmentation, and diffusion, the attack recovers 99.92%-100% of hidden vector correspondences. Reconstructed classification models achieve 94.39%-99.54% victim agreement and differ by at most 1.59 percentage points in accuracy; the recovered segmentation model achieves 98.35% output agreement. These findings suggest that lightweight protection should address both per-vector geometry and joint structure across released weights.

The rapid development of artificial intelligence, including agents and deepfake techniques, has accelerated phishing attacks and lowered the threshold for attackers. Modern phishing attacks now blend multiple tactics, including social engineering, URL spoofing, and AI deepfakes enabling adversaries to craft highly convincing messages that exploit human vulnerabilities and bypass traditional detection systems. At the same time, current security awareness education struggles to keep up with the speed, sophistication, and complexity of these evolving threats. To address this challenge, we propose a two-stage anti-phishing framework, CyberGLA, that combines technical defense and user-centered security education. In the Detection stage, we introduce EmailKnight, a spoof detection tool that performs multi-level email analysis. To enhance user awareness, the Training stage incorporates a large language model (LLM)-based security coach that dynamically selects personalized training modules based on the outcomes of the Detection stage. This dual purpose design philosophy enables effective protection against the evolving threats of modern email phishing attacks.

Automated applicant tracking systems increasingly decide who advances in hiring, and litigation and regulation now demand that those decisions be auditable. Existing tools sit at two extremes. Group fairness metrics such as the disparate impact ratio summarize a whole population but cannot say which individual decisions were unfair or why, while local explainers such as SHAP attribute a single prediction but are not connected to the legal standard by which hiring bias is judged. We present the AI Bias Firewall (AIBF), a method that audits an applicant tracking system one decision at a time. AIBF neutralizes a candidate's protected-attribute proxies, re-scores the decision, and measures the resulting counterfactual shift, which yields a signed per-decision bias in score points, a flag for decisions the protected attributes changed, and a plain-language explanation naming the responsible factors. We evaluate on two real public datasets, Adult and COMPAS, rather than on synthetic data. The per-decision counterfactual shift is faithful, aggregating to reproduce the known group level disparity, for example a mean shift of +7.5 points for the privileged group and -8.0 for the disadvantaged group on Adult, consistent with the measured statistical parity difference. AIBF identifies the decisions that protected attributes flipped with an area under the ROC curve of 0.963 on Adult, against 0.672 for a baseline that flags by group membership, and it identifies the harmed candidates so precisely that reviewing only five percent of decisions surfaces fifty-five percent of them, against six percent under group based review. We also report a limitation: correcting flagged decisions raises the disparate impact ratio substantially but not to legal parity, because features labeled as merit carry residual proxy correlation. AIBF is released under the Apache 2.0 license with code and experiments.

A verification substrate is more credible when exposing errors in its own claims, not just model outputs. GBI-DCSE v3 falsified an architectural claim: the reported Fisher value epsilon ~ 0.066 satisfies the kappa^2 <= 10^4 budget only on the slice [epsilon, 3, 4, 5], while the full box [epsilon, 20]^4 requires epsilon ~ 0.326472. This erratum highlights whether an enterprise verification architecture can isolate interface failure, task competence, policy admissibility, and control integrity while keeping claims auditable. BoundaryBench v0.1 established the baseline: Qwen3-4B-Instruct-2507 completed 768 frozen executions, but 0% cleared the contract (369 failed parsing, 399 failed validation), limiting downstream selectivity metrics. This companion study evaluates three successive improvements. First, BeTaL-GBI v0.2 applies Benchmark Tuning with an LLM-in-the-loop over 2,218,750,380 grid points, separating format admission from conditional performance (rho_adm = N_admitted/N; rho_task = N_verified/N_admitted). Following schema repair, a model-free feedback search achieves a 2.87% mean held-out target gap, outperforming non-feedback baselines (13.61%, 11.46%). Second, GBI v2 swaps static keys for a reference-independent witness state W and policy P. Across 512 synthetic tasks, a 16-gate policy detects all 116 injected severe contradictions and accepts all 99 clean records (broad denominator: 4.27%). Hallucinator and evidence-forger surrogates are blocked with zero silent promotions. Third, GBI-DCSE v3 maps 99 claims to machine-readable evidence: 95 of 96 testable claims pass, with 148 standalone checks executed without failure. The harness exercises signed ledgers, PBFT quorums, and enclave forgery across 62 configurations. Under synthetic conditions, GBI-DCSE is a selective, policy-versioned, self-auditing test and routing substrate.

Prompt injection is listed as the \#1 threat to AI agents. When an agent accesses external data from websites, files, or emails, an attacker may inject a prompt into the data, saying, "Ignore all prior instructions and perform <an attacker's task>." To prevent arbitrary manipulation of agents, defenders try to train secure LLMs, which, however, still suffer from near 100% attack success rates (ASRs) against adaptive prompt injections. We note that this is because existing defensive finetuning recipes rely on sequence-level feedback signals (in DPO or GRPO). Treating an entire output equally prevents the model from learning precisely which output tokens are insecure. In this paper, we propose Secure On-Policy Distillation (SecOPD) that provides token-level feedback to guide defensive fine-tuning. The LLM receives an injected sample and produces a rollout, whose tokens are scored by the initialization model given the corresponding clean input. With more fine-grained training signals, our defended Qwen3.6-27B achieves a 9.0% ASR against the SoTA PISmith adaptive prompt injections, compared to 94.0% for the prior SoTA, Meta-SecAlign. The obtained security generalizes to domains completely unseen in training: in agentic tool calling, SecOPD achieves a 4.7% ASR compared to 5.5% for Meta-SecAlign. Code and the model are available at https://github.com/pppyb/SecOPD and https://huggingface.co/pybbb/Qwen3.6-27B-SecOPD.

Cloud environments built on Amazon Web Services face a structural security vulnerability: once a credential passes authentication, the resulting session is often treated as trusted for its entire duration. This assumption fails when credentials are stolen. We introduce the Explainable Adaptive Zero Trust Framework (EAZTF), a cloud-native security layer that continuously reevaluates the legitimacy of API actions throughout a session. EAZTF combines Isolation Forest and XGBoost to evaluate eight CloudTrail and IAM-derived behavioral features in real time and produce a Trust Risk Score (TRS) that determines whether a session continues, requires step-up MFA, or is restricted. Each decision is accompanied by a SHAP or LIME explanation, providing human-readable audit records for security analysis and compliance. The framework is also evaluated against four adversarial evasion strategies: credential theft, behavioral mimicry, API rate evasion, and privilege escalation. Experiments on an 8,500-record synthetic CloudTrail dataset show that Isolation Forest achieves 94.4% precision, 91.2% recall, and an F1 score of 0.928. Across the four adversarial scenarios, the mean detection rate is 91.0%, with behavioral mimicry being the most difficult at 83.9%. SHAP analysis identifies IP reputation, login-time deviation, and API call velocity as the three dominant features. A structured NIST SP 800-207 self-assessment gives EAZTF a mean compliance score of 93%, compared with 38% for a traditional perimeter baseline. Mean time to detect decreases from hours to under one minute. Because the evaluation uses synthetic data, these results should be interpreted as indicative rather than validated production performance.

Defenders of critical infrastructure (CI) subscribe to many public open-source intelligence (OSINT) feeds without an empirical basis for which feeds actually precede attacks. We provide one. Across 54 confirmed CI cyberattacks from 2010 through 2024 spanning twelve named CI sectors plus a cross-sector category (consolidation rules in Section IV), paired with 12 null-control vulnerability cases drawn from the same source space, we audit per-source attack coverage, null-case contamination, and signal lead time for ten public OSINT source classes that meet a minimum-volume threshold. Sources separate cleanly into three operationally distinct mission profiles (pooled Fisher exact p = 3.4x10^-8): precursor (six classes with zero observed null firings at coverage at or above 5%), disclosure-exposure (three classes whose null contamination meets or exceeds attack coverage), and one large broad-coverage class that mixes the two profiles but retains 91.3% within-corpus precision. The precision-side classification is stable across a 2019 temporal partition and across a US-versus-non-US geographic partition. Two sources, one broad-coverage and one precursor, cover 92.6% of corpus attacks; three cover 96.3%. The greedy portfolio at k = 3 outperforms the mean random three-source subset by 39.8 percentage points. Several source classes widely treated as canonical for industrial control system defense fall into the disclosure-exposure profile by operational mission, not by quality. Per-sector, per-actor, and per-jurisdiction portfolios diverge in rank order despite a shared rank-one source. The corpus, linkage protocol, and classification rules are released.

Agentic large language models are increasingly used in digital forensic analysis, yet their ability to infer relational structure inside undocumented mobile application databases remains poorly understood. In forensic contexts, structurally incorrect inferences can yield results that appear plausible while remaining evidentially unsound. This work evaluates agentic structural inference as an isolated capability, treating execution success and structural correctness as distinct evaluation axes. It examines how an agent reconstructs table relationships, linking attributes, and executable join paths when given only a raw database and a natural-language investigative prompt. We apply a fixed, deterministic evaluation pipeline to two contrasting SQLite repositories: Android's SMS database with stable identifier propagation, and Snapchat's database with irregular schemas, ephemeral identifiers, and polymorphic relationships. Using expert-verified SQL ground truth, we evaluate (i) structural correctness of inferred relational links, (ii) execution coherence under multi-table reasoning, and (iii) robustness and failure modes of inferred structure when execution succeeds but relational interpretation diverges from expert ground truth. Evaluation is performed independently of semantic interpretation, with full queries and execution traces provided in the Appendix. Results show that structural inference remains reliable in regular schemas but degrades sharply as schema ambiguity increases, frequently producing structurally plausible yet incorrect joins that execute successfully. These findings clarify where schema-agnostic agentic reasoning can support forensic analysis, how its robustness degrades under realistic schema irregularities, and why additional verification remains essential before inferred relationships can be treated as reliable evidence.

Discovering personally identifiable information (PII) in mobile forensic databases is difficult because the relevant table-column regions are unknown, distributed across heterogeneous SQLite schemas, and may contain values embedded in free-text or semi-structured fields. We present a hypothesis-driven framework that treats PII localization as bounded, adaptive search under uncertainty. An agent ranks candidate table-column regions, probes sampled values, and maintains a memory of prior evidence, confidence scores, and decisions to refine subsequent hypotheses. The framework separates lightweight PII exploration from targeted extraction, normalization, and deduplication over validated regions, thereby limiting exhaustive inspection to regions supported by sampled evidence. We evaluate the framework on 25 SQLite databases from 10 Android and iOS applications in the Cellebrite CTF corpus, targeting email addresses, phone numbers, domain names, person names, and postal addresses. Against a corpus-level distinct ground-truth set of 3,751 entities, Gemini 2.5 Pro achieves 94.5% F1 while reducing the effective extraction search space by 79.9% on average. Results across 12 model backends show strong performance among several frontier models, but substantial sensitivity to model capability.

Remote state preparation (RSP) is a central primitive in quantum cryptography, enabling classical parties to remotely construct quantum states using only classical communication. As a result, RSP serves as a key building block in numerous protocols involving classical clients and quantum servers, allowing classical parties to leverage the advantages offered by powerful quantum computers. All known constructions of RSP rely on strong cryptographic assumptions, typically variants of trapdoor claw-free functions (TCFs). In this work, we initiate the study of a weaker form of remote state preparation, which we call eavesdropper-blind remote state preparation (EB-RSP). Informally, EB-RSP requires blindness only against external observers who see the transcript of the honest protocol, rather than against the quantum server itself. Despite this relaxed adversarial model, the resulting notion remains sufficient for useful cryptographic applications. In particular, we show that two-message EB-RSP already suffices to construct quantum public-key encryption with classical public keys and quantum ciphertexts. We then construct two-message EB-RSP protocols from specific one-way group actions, yielding a first step toward RSP-type primitives based on assumptions that do not rely on trapdoors. Finally, we observe that existing RSP constructions are likely naturally adaptable to the two-message EB-RSP notion; we demonstrate this explicitly for a concrete TCF-based RSP construction.

Persistent memory makes false information durable: once a false statement is stored, it can be retrieved into future sessions that match it. We measure the cost of this failure mode using plainly worded false assertions generated in a single pass, with no instruction, trigger, or retriever optimization. Poisoning 1.2% of a LongMemEval corpus reduces accuracy from 0.850 to 0.300. A four-stage write-time screening pipeline that reaches 0.832 recall on indirect prompt injection while flagging 1.5% of trigger-word-laden benign text rejects 0 of 360 poisoned memories. We argue this exposes a boundary of content-only screening: distinguishing a false assertion from a true one generally requires external grounding beyond the text itself. We then evaluate provenance-weighted retrieval. The shipped weight is statistically indistinguishable from no defense (p=0.80), while a stronger weight recovers utility only by excluding untrusted content. In a mixed-provenance corpus where untrusted content is mostly benign, accuracy rises from 0.3167 to 0.7000; when the answer-bearing evidence itself arrives untrusted, evidence recall falls to zero and accuracy to 0.0417. Under the measured similarity regime, the additive provenance term has no usable setting: a weight strong enough to resist query-shaped poison is also strong enough to suppress legitimate untrusted evidence. We therefore argue for bounded occupancy constraints at retrieval rather than additive provenance penalties, and release the harnesses, corpora, and aggregate run reports.

Workplace surveillance is used by organizations to protect corporate assets and monitor employee productivity. This research presents two central arguments on workplace surveillance: although surveillance serves legitimate organizational purposes, over-surveillance can violate legal requirements and data privacy principles; and a primary security objective of workplace surveillance is the detection of insider threats (InT). InT are comprised of individuals with authorized resource access whose intentional or unintentional actions may damage or compromise corporate assets. This paper investigates InT personas to understand behavioral and psychological detection criteria. Employee surveillance tools and techniques are reviewed to characterize the employee surveillance landscape. Workplace privacy laws, examples of over-surveillance, and the resulting privacy harms are addressed. The review identifies research gaps related to over-surveillance, including the generation of excessive alerts that may obscure meaningful InT indicators. The paper concludes with recommendations to improve workplace surveillance transparency, implement InT training programs to improve organizational detection capabilities, and tune InT tools to detect relevant psychological and behavioral indicators.

Tool-using AI agents turn delegated tasks into provider effects, yet authorization often ends at admission while provider state, delivery, retry, and recovery evolve. A request may change before commit, or response loss may cause a replacement to create a second effect from one approval. We present AID-Guard, a stateful authorization-to-effect closure protocol. It revalidates the approved request and provider state at commit, retains one reservation under ambiguity, and permits release or one successor only after a terminal result or certified no effect with a delivery fence. For supported provider contracts, one reservation yields at most one effect across retry and recovery. To our knowledge, it is the first evaluated agent-authorization protocol to unify these controls in one lifecycle. We implement a Python/SQLite prototype. In a declared loopback MCP domain, 13 live mutations caused no unauthorized provider effects, three concurrent histories were linearizable, and evidence bundles supported public verification and replay. All 210 Stripe provider-contract trials matched predeclared outcomes. Across Stripe and Resend, 40 terminalize-successor schedules, 30 overlapping races, and 10 crash-recovery schedules completed without duplicate effects. Under complete proposer compromise, AID-Guard blocked 44/44 attacks and admitted 44/44 matched legitimate proposals. Its strict exact-manifest profile reduced benign utility by 35.4 to 43.8 percentage points; a typed frontier recovered 9-10 completions without observed unsafe effects. A composition study blocked 20/20 post-admission lifecycle attacks and preserved 8/8 valid or exact-retry executions. The results support authorization-to-effect binding under the evaluated effect-path inventory, provider contracts, and failure schedules.

Decentralized Federated Learning (DFL) promises trust-free collaborative learning by replacing the centralized parameter server with peer-to-peer model exchange. However, this architectural shift fundamentally reshapes the threat landscape. Without globally coordinated aggregation, DFL becomes particularly susceptible to backdoor attacks, in which malicious participants implant persistent hidden behaviors while maintaining high clean-task performance. In this paper, we argue that the robustness of DFL has been significantly overestimated. Existing studies rely on simplified threat models, non-adaptive adversaries, fragmented evaluation protocols, inconsistent communication topologies, and ad hoc training configurations, leading to an incomplete understanding of DFL security. To address these limitations, we present BackDFL, a unified benchmark for systematically evaluating DFL under realistic and adaptive backdoor attacks. Through extensive experiments, BackDFL exposes critical failure modes of decentralized learning. Our results demonstrate that both state-of-the-art Byzantine-robust DFL methods and adapted FL backdoor defenses fail under modest malicious participation rates (as low as 15%), especially in heterogeneous settings, while their robustness varies substantially across communication graph topologies.

Medical image-text data can expose protected health information (PHI) through both visible image content as well as accompanying text, creating a barrier to privacy-preserving medical AI systems. This risk is especially prominent in multimodal systems, where images, questions, reports, and clinical context may enter training, evaluation, or inference pipelines. Existing medical vision-language benchmarks primarily emphasize task utility, while de-identification methods are often evaluated separately from downstream reasoning. We introduce ClinX, an end-to-end multimodal PHI sanitization framework for medical image-text data. ClinX detects visible identifiers with optical character recognition (OCR), constructs binary PHI masks, and applies ClinX-PRISM, a no-skip generative restoration module with privacy-oriented post-processing for burned-in identifier suppression. In parallel, text-side PHI is reduced through progressive de-identification levels: regex masking, context-aware masking, and rewrite-based sanitization. We evaluate ClinX in medical visual question answering (MedVQA), jointly measuring PHI leakage and downstream utility across image-side, text-side, and combined de-identification settings. Results show that OCR-only masking is not sufficient as a standalone solution, and restoration-based sanitization better preserves clinically relevant visual context while sharply reducing recoverable PHI.

We build a system that answers "which database vectors are most similar to my query?" without the server ever seeing the query. The query is encrypted with fully homomorphic en- cryption (FHE); the server does all its scoring on ciphertexts and returns encrypted results that only the client can read. The challenge is speed: at a billion vectors, scoring every row under encryption is far too slow, so we combine two ideas - rank reduction (shrink each vector's dimen- sion) and a hierarchy (route to a small candidate set instead of scanning everything) - executed under encryption on a single GPU. We evaluate on three corpora at very different scales: a face corpus of 222 049 centroids clustered from ~10 M face images (512-dim), DataComp-1B (1.39 x 10^9 vectors, 512-dim CLIP), and Deep1B (10^9 vectors, 96-dim). On DataComp-1B we reach a recall@10 of 0.90 against the single labeled answer, or 0.95 when a near-duplicate im- age in the top-10 also counts as correct (the data is web-scraped and full of duplicates), at ~6 s per encrypted query on a GPU; a lighter configuration reaches 0.78/0.83 at ~1.8 s. These are warm (deployable) server-side latencies - client decryption and network transfer are excluded. On Deep1B we reach recall@10 0.90 under all-levels FHE (0.9045 measured over 2000 FHE queries, matching the 0.906 plaintext routing - the 96 -> 128 zero-pad is exact, correlation 1.0) at 2.3 s warm per query. We describe the full client-server protocol in enough detail to repro- duce it, and report accuracy and latency for every configuration. We also measure what this speed costs: the hierarchy's access pattern leaks the database geometry (an observer recovers 72% of the coarse-cell neighbor graph from access patterns alone), and we show that seeded (fixed-group) padding cuts this leak by ~35x (to ~2%), where naive padding is defeated by a repeated-query attack.