CyberSec Research

A watermark in a generative model's output is usually asked only whether a text is machine-made. The same mark can do more: attribute it to the user who produced it, extract a hidden payload, or localize the part that survives editing. These form a forensic ladder, and we ask what each rung costs in the sample length $n$. One object organizes the answers. Let $S$ be the secret the mark carries (a user's identity or payload), and let the information profile $ν(t)=I(S;X_t\mid X_{<t})$ record how much the $t$-th token reveals about $S$ given the earlier ones. Its total mass pays for attribution and extraction; how that mass is spread pays for localization; and detection alone is paid for not by information but by presence, the distance from the marked to the unmarked distribution. The literature's two quality models, a mark subtle on every token and one that stamps a few tokens loudly, are two incomparable ways of capping this profile. Our main theorem settles the ladder's entropy column. For statistically distortion-free schemes, attributing a text to one of $N$ users costs $Θ(\log N/h)$ tokens over every stationary-ergodic source of entropy rate $h$, sharp to a $(1+o(1))$ factor: to our knowledge the first tight entropy-rate law for multi-user attribution (via exact alignment). The natural collision-counting analysis overcharges without bound; only a decoder thresholding each candidate by its own realized surprisal attains the rate while almost never implicating an innocent user. A matching converse makes the law two-sided, and extraction of an $\ell$-bit payload costs $Θ(\ell/h)$. Two gaps are real, not modeling artifacts: a $Θ(\log N)$-token window in which a text is provably machine-made yet unattributable, and a footprint-resolution uncertainty principle. Experiments on GPT-2, Pythia-410M, and Qwen2.5 recover the predicted constants.

Jailbreak-robustness research typically evaluates safety through generated responses using an LLM-as-judge approach. Such evaluations, however, are sensitive to the benchmark's grading procedure and capture only observed behavior on a given set of attacks, without directly revealing the hidden fragility of the underlying safety mechanisms. This work proposes JADR (Jacobian Assessment of Danger Recognition), a protocol that measures a model's internal representation through Jacobian space (J-space, a recently proposed workspace of verbalizable concepts) before the first response token is generated. For every prompt and layer we record the top-k J-space tokens; these are grouped into six behavioral scenario axes and compared between a danger sample based on StrongREJECT and a safe control drawn from XSTest and OKTest. The method does not call on an external judge model: the computation runs entirely locally, on the activations of the model under evaluation, which lets us compare both different models against each other and modifications of a single model -- quantization and fine-tuning in particular -- on the same terms. The final comparison rests on the proposed SafetyAUC metric, complemented with bootstrap confidence intervals. The protocol is applied to six models (Qwen3-1.7B, Qwen3-4B, Qwen3-8B, Qwen3-Uncensored-4B, Qwen3-SafeRL-4B, Gemma 2 9B) across three weight-representation regimes -- BF16, INT8, and INT4 -- and checked against an independent behavioral evaluation with the StrongREJECT grader. The metric separates models with a strong versus a weak internal safety mechanism with statistical significance and captures substantively different effects across quantization regimes.

Quantum key distribution (QKD) offers unconditional security but existing QKD networks remain difficult to scale across heterogeneous infrastructures and administrative domains due to vendor-specific interfaces, trusted-node constraints, and limited interoperability. This work presents a flexible multi-domain and multi-site quantum-secure network architecture integrating vendor-agnostic QKD, SDN orchestration, and cloud-managed trust services. Communication is based on Zero Trust Network Access protocols featuring multi-level authentication mechanisms building upon post-quantum cryptography (PQC) signature and key encapsulation algorithms. The system is deployed on a real-world testbed with domains incorporating QKD nodes from 3 vendors, as well as domains without QKD infrastructure elements. Experimental results show that PQC and SDN overhead remain relatively low even on constrained devices, with the main bottleneck being QKD key retrieval and vendor-specific key streaming limitations. The proposed framework extends quantum-safe key transport beyond native QKD boundaries while preserving flexibility, interoperability, and compatibility with existing infrastructures.

Encrypted control lets a cloud coordinate a fleet of agents on fully homomorphically encrypted state, keeping their positions and commands private. The approximate scheme for real-valued control, CKKS, returns decryptions that carry the encryption noise, a key-recovery leak; the loop must decrypt to actuate, so the leak is unavoidable. Yet the security of approximate FHE is studied statically, encrypted control assumes an honest-but-curious cloud, and persistent-threat games never reach inside the cryptosystem. We model the loop's security under an advanced persistent threat as a two-phase game, passive reconnaissance then active manipulation, separated by a measured residual detector that sees only the manipulation. The passive phase reduces to the known flooding tradeoff; the active defense is re-keying, not bootstrapping, since only re-keying resets accumulated leakage. The active phase is a detection-evasion timing game: overt manipulation is caught, so the rational adversary stays stealthy, and at its Stackelberg equilibrium the defender re-keys on the laziest cadence that denies it, set by the control-theoretic fragility of the graph topology. The marginally-stable graph must re-key far more often than the well-connected one. A three-way tension among FHE precision, control accuracy, and re-key cadence sets where this game lives, between a securability floor and a static-suffices ceiling. The efficient secure point is that window, where re-keying is the price of precision efficiency. More broadly, security for an approximate cryptosystem in a feedback loop is a dynamic game whose defender's move is the scheme's own refresh, applying beyond control to any system that must repeatedly decrypt to act.

Filesystem isolation in container ecosystems is often weakened by cross-boundary path misresolution, causing path traversal (PaTra) vulnerabilities. These vulnerabilities stem from insecure host-container interactions and have become increasingly pervasive as cloud systems mount shared resources, such as GPUs and agent workspaces, into containers to support AI workloads. Existing defenses remain inadequate. Kernel-level protections are intrusive, can destabilize system calls, and have therefore not been accepted into the Linux mainline. Detection methods rely on static rule matching or manual code auditing. Static rules can flag path-related functions but fail to capture the semantics needed to determine whether a host-container interaction exists, causing many false positives. Manual review requires domain expertise, making it costly, inefficient, and difficult to scale. To address this threat, we present Bulkhead, an automated framework that integrates large language models (LLMs) with formal methods for semantic vulnerability discovery and remediation. Bulkhead uses a multi-agent system to identify and repair PaTra vulnerabilities through multi-dimensional knowledge patterns generalized from known cases. It first applies high-risk functional patterns to locate entry points for cross-boundary interactions in containerized code, then uses call-chain patterns to recover the corresponding execution paths at suitable depth. The Detection pipeline analyzes these call chains against the application scenarios and threat model, identifying vulnerabilities such as missing security checks and TOCTOU flaws in cross-boundary interactions, and generating proof-of-concept (PoC) exploits for validation. These PoCs then guide patch generation. To ensure remediation correctness, the Patch pipeline performs assertion-driven verification using predefined model-checking templates.

Full Disk Encryption (FDE) has become increasingly important in the last decades due to the evident confidentiality concerns. In most systems, encryption is provided by an operating system driver, through which the user can transparently access the encrypted disk after supplying the required keys (or the credentials from which those keys are derived). In this work, we explore an alternative approach: the use of an intermediate USB device placed between the host system and the external hard disk, where the encrypted data are stored. Although there are existing devices that follow this inline approach (such as RAID controllers and USB enclosures with built-in encryption), we explore the use of a general-purpose single-board computer running Linux with USB On-The-Go support (e.g. a Raspberry Pi), to provide FADE (Fully Authenticated Disk Encryption). Our system, named CC (Cryptographic Companion), performs the cryptographic operations (encryption/decryption and authentication), providing a standard USB mass-storage interface to the host (which is entirely unaware of the presence of encryption) while relying on the external USB hard disk to store the corresponding encrypted blocks. Our design provides several key advantages: flexibility, low cost, transparency, the use of generic hardware and free and open-source software, adaptability to emerging cryptographic schemes, and mitigations against malicious disk firmware. This paper presents the design and implementation of the CC and an experimental evaluation of our current research prototype, which indicates that it is sufficiently efficient for most common use cases.

Large language models (LLMs) are increasingly deployed as purpose-specific agents to handle domain-specific tasks such as customer service and code generation. These agents are expected to comply with not only generic safety guardrails but also purpose-specific restrictions tailored to their designated roles. Such additional restrictions enlarge the attack surface, particularly to prompt injection (PI) attacks. To defend against such attacks, existing detection methods primarily rely on analyzing input-output patterns, yet yield limited effectiveness. To address this limitation, we turn to analyzing the hidden activation space and discover that LLMs inherently retain latent policy-violation (PV) concepts when prompted with requests beyond their designated purpose. Particularly, PV concepts capture the semantics of conflicts between user queries and predefined restrictions, implicitly reflecting LLMs' intrinsic awareness of recognizing policy violations. Building on this insight, we propose PVDetector, a training-free framework that detects PI attacks during LLM inference by measuring hidden-state alignment with PV concepts, which are derived offline from the contrastive pairs of policy-violating and policy-compliant prompts. Experiments across multiple LLMs and datasets show that PVDetector achieves <1\% false negative rate with minimal auxiliary overhead, consistently outperforming state-of-the-art methods. Our code is available at https://github.com/Claresigle/PVDetector .

The rapid advancement of synthetic speech generation methods has made audio deepfake detection a critical challenge in multimedia forensics. While recent approaches achieve high detection accuracy, they typically rely on black-box architectures that offer limited interpretability and high computational complexity. In this paper, we propose an explainable-by-design audio deepfake detection framework based on Wiener-Hopf linear prediction, processed by a lightweight 2D Convolutional Neural Network (CNN). This design enables a direct and transparent connection between classification outcomes and the acoustic properties of the signal. Experimental results on benchmark datasets demonstrate competitive detection performance while maintaining significantly lower computational complexity compared to state-of-the-art solutions. The interpretability analysis using Grad-CAM reveals that the classifier focuses on low-order predictor coefficients and on silence and transitional regions, suggesting that the Wiener-Hopf predictor captures reverberation characteristics and subtle statistical inconsistencies in synthetic speech. Finally, robustness experiments show that fine-tuning effectively recovers detection performance under common post-processing degradations, including additive noise, MP3 compression, and telephone filtering.

AI agents are said to be forming an economy in which they pay, on their own, for the data, APIs, and compute they consume. x402, which settles a stablecoin payment on-chain for each purchase, is the most widely deployed protocol for this, and its hundreds of millions of settlements are read as proof that the economy has arrived. We show the count cannot be read as adoption: it is the one metric an interested party can manufacture almost for free, since the facilitator sponsors the gas and nothing on-chain marks who controls a payment. We give the first population-scale measurement of x402 on Base, supplemented with a coarser Solana census. Identifying settlements from their on-chain event and resolving the true payer through the meta-transaction layer, we sort each by what its trace can prove via a payment graph. Over a 280-day window Base carries 136{,}708{,}672 settlements worth \$44{,}121{,}383.81, concentrated on every axis we measure (payer, recipient, and value Gini all above 0.98), yet 21.20\% are fictitious and 63.78\% internal settlement within a linked cluster. What is genuinely independent is bounded: it lies between the \$187{,}861.35 that demonstrably reaches a nameable service and the \$20{,}258{,}746.09 (45.92\% of value) not provably manufactured. Finally, we resolve the count's manufacturable component, a coherent operator-driven economy, star-shaped, machine-timed, and gas-subsidized. Settlement count measures manufacturability, not adoption.

Adversarial robustness research has produced hundreds of defended models over the past decade, yet the literature almost universally reports robustness results in isolation: standard (clean) accuracy and adversarial accuracy of the robust model are shown, but the gap to the corresponding vanilla model is rarely quantified. We introduce VanillaBench, a systematic benchmark that makes this gap explicit. For every adversarially-trained model catalogued by RobustBench across four threat models, we compute the accuracy difference against multiple vanilla references from Papers with Code, computed over both all entries and no-extra-data entries, the best vanilla model as of the robust model's publication year, and an architecture-matched baseline. Across all 186 robust models, the mean delta clean relative to the best vanilla model ranges from -7.7 to -29.5 percentage points, and even the single most robust model per track still trails its temporal vanilla counterpart by 4.0-21.0 points. The architecture-matched comparison, which isolates the effect of adversarial training from architectural differences, reveals a mean gap of -3.5 to -17.5 points. Restricting this architecture-matched comparison to models whose vanilla accuracy is known for the exact same architecture, rather than approximated from a related one, narrows the gap to -4.0 to -14.0 points. These results demonstrate that the robustness-accuracy trade-off is substantially larger than what is typically conveyed by individual papers. This information is critical for practitioners and decision-makers. When deploying models in real-world settings, the accuracy cost of robustness directly affects business outcomes, yet current publications do not provide the vanilla baseline needed to assess it. We argue that future robustness evaluations should report vanilla-referenced accuracy gaps as a standard component.

Developers now draw code from two very different sources, the accumulated human answers on sites such as Stack Overflow and the output of large language models. We ask two questions about that split. First, can the provenance of a code snippet be recovered from the code itself, and second, do the two sources differ in the security patterns they adopt for the same task. Using only open sources, a public gateway of open-weight language models and the public Stack Overflow API, we build a fully reproducible pipeline that collects real implementations of 31 security-sensitive programming tasks, among them OAuth with PKCE, JWT verification, password hashing, and SQL access, from 9 language models and from human answers, and scores every sample with deterministic security and style detectors. On 528 real samples we train a cross-validated classifier that recovers human versus model provenance with 93 percent accuracy against a 78 percent baseline, and a 7-way classifier that attributes a sample to the specific model that wrote it at 48 percent. We then report where the sources diverge on security, which patterns models adopt more often than the human corpus and which they inherit from it. Running the same tasks in Python, JavaScript, Go, and Java, we find the security divergence holds in every language while the provenance boundary is partly language-specific and does not transfer symmetrically between them. A vulnerability repair case study, in which models are handed insecure code and asked to fix it, finds a 77 percent repair rate across 21 seeds and 12 weakness classes, but a recurring partial-fix failure in which the model removes the insecure pattern without adding the correct defense. The pipeline is data driven, so any new task or language is added as a single specification entry, and a fail-closed checker re-derives every number in this paper from the stored data.

A common intuition holds that a region's music mirrors the temperament of its people, so that melancholic melodies mark melancholic populations. We test the measurable half of that intuition and reject the inferential half. Using the Essen Folksong Collection, a corpus of thousands of notated folk melodies, we extract real melodic and affect-related features from 2393 deduplicated melodies spanning 16 countries and 7 geographic regions, with the analysis performed on symbolic scores rather than audio. The mode of each melody is computed with a key-finding algorithm rather than read from the file, because the collection's own documentation warns its major and minor labels are unreliable. Cross-country differences in melodic structure are large and highly significant. All 8 tested features differ across countries at p<0.001, with the leap-related features reaching p<10^-90, and China carries a distinctive wide-leap, high-activity signature (arousal composite +1.24 standard deviations, mean absolute interval 2.77 semitones against Germany's 2.17). We then test the inferential half. We correlate the regional musical-affect measures with two published, validated national indices, the World Happiness Report ladder score and the Hofstede individualism index. None of the 6 correlations is significant (0 of 6). The geography of musical affect is real and measurable, but it does not predict how happy or how individualist a population is, and any claim that it does is an ecological fallacy. We release the full extraction and analysis pipeline, and a fail-closed checker re-derives every number in this paper from the data.

LLM-assisted reverse-engineering (RE) systems analyze strings, decompiler output, and tool reports derived from ttacker-controlled binaries. A binary can make data look like instructions or records from one origin look like independent evidence. We call such failures Representation-Confusion Attacks in Reverse Engineering (RARE): the pipeline promotes a correctly extracted observation to instruction authority, claim-validating evidence, or trusted analysis state without the authority or support that role requires. RARE-Bench measures these failures with behavior-checked clean and adversarial binaries. After an exploratory 11,520-call study, we test RARE-Guard's authorization and evidence controls on 20 new programs and two models. Without runtime controls, the models propose a planted unsafe action in 35/40 adversarial cases and 0/40 clean cases. When binary-derived content is shown only as data (Data-Only rendering), they still make 15 unsafe proposals. Tool Authorization denies all 15 and authorizes all 40 matched analyst requests. On identical report drafts, Support Gate validates 23/40 false claims by counting records from one origin separately. Provenance Gate groups those records before counting support, validates 0/40 false claims, and retains all 40 supported claims. We then instrument Ghidra, r2pipe, and angr on 16 further programs. In a preselected eight-program subset, no single-tool draft reaches Support Gate's validation threshold for the false claim. In fused drafts across all 16 programs, Support Gate validates 32/32 false claims. Provenance Gate prevents validation of all 32 and retains all 32 supported claims. A deterministic renderer prevents downgraded claims from reappearing in the final report. Binary-derived content may therefore guide analysis without gaining authority over tools, and views from several tools do not necessarily provide independent evidence.

Post-Quantum Cryptography (PQC) is increasingly being integrated into TLS 1.3 to enhance resilience against quantum-enabled attacks. However, the additional computational and communication overhead introduced by PQC primitives during the handshake phase may also amplify the impact of TLS handshake exhaustion attacks, leading to more severe Distributed Denial-of-Service (DDoS) threats. In this study, we establish an empirical testbed consisting of one PQC-enabled TLS server and ten attacking nodes, generating over 16.5 GB of mixed traffic data that includes both legitimate browsing behavior and high-intensity handshake exhaustion attacks. Experimental results show that PQC-TLS can prolong periods of sustained high CPU utilization on the server by up to 88 times, significantly amplifying the effectiveness of such attacks. Furthermore, we evaluate state-of-the-art deep learning-based Intrusion Detection Systems (IDS) and observe a substantial decline in attack detection performance under PQC traffic conditions. In particular, exosphere achieves only around 50% recall, while HyperVision's AU-ROC degrades to near-random levels (0.49), revealing critical detection blind spots in existing IDS when operating in PQC environments. The main contributions of this work are threefold: (1) we systematically quantify and analyze the root causes of IDS detection blind spots in PQC settings; (2) we publicly release a comprehensive PQC-DDoS hybrid traffic dataset, including precise attack timestamps and server-side resource monitoring data; and (3) we open-source all experimental code and AWS deployment scripts, enabling a fully reproducible cloud-based testing environment. These resources aim to support both academia and industry in developing next-generation PQC-aware intrusion detection systems.

The increasing adoption of autonomous coding agents accelerates software development but also introduces scoped security risks within high-impact file paths that can outpace traditional human review capacity. While prior research has primarily evaluated these systems in terms of functional correctness and productivity, this paper presents a large-scale empirical study using the AIDev dataset to systematically characterize security code smells in agent-generated pull requests (PRs). Through a combination of a validated LLM-as-a-judge framework and manual qualitative analysis, we identify and classify security misconfigurations across 16,112 file changes spanning 4,022 pull requests. Our results reveal that 38.9% of agent-generated PRs contain at least one security smell, with supply chain integrity issues accounting for 82.3% of all detected security smells. Furthermore, hard-coded credentials constitute 99.6% of all critical-severity security smells. Crucially, we find that human collaborators are responsible for introducing 67.6% of genuine leaked secrets within these agent-assisted workflows, while existing automated and human review processes fail to detect 81.1% of these credentials prior to integration. These findings highlight substantial security risks in agent-assisted software development workflows and suggest a potential reduction in developer vigilance. They also underscore the urgent need for context-aware security guardrails implemented directly at the point of human-AI collaboration.

Text-to-SQL is increasingly deployed across trust boundaries between data providers and users. Such deployment must balance three competing requirements: policy compliance, answer coverage, and bounded cost. Existing approaches typically decide refusal based on which columns a query mentions and enforce it stochastically. Whether a query is compliant, however, depends not only on which columns appear but on how they are used, and stochastic enforcement cannot deterministically rule out violations. We formalize this requirement as a column-use policy over semantic use: output, filter condition, and aggregation argument. We integrate the policy by aligning each role with grammar productions tracked by the decoder. The resulting system, PCC-SQL, applies a per-token logits mask that deterministically eliminates single-query column-use violations on the supported SQL fragment in a single decoding pass. Across three benchmarks and three open-source models, PCC-SQL achieves 0% Leakage Rate and Coverage up to 88.7% on Spider-CU, while staying within +10% tokens of direct prompting. We additionally assess semantic alignment with execution accuracy.

LLM agents acquire new capabilities by downloading skills from open registries. Instead of browsing these catalogs manually, developers typically ask the agent to recommend and install a skill. This convenience hides a risk: agents frequently invent names for skills that exist in no registry. We term this flaw skill name hallucination. A fake name may seem harmless, but it opens the door to supply-chain attacks. Because registries rarely verify publishers, an adversary can prompt the agent, collect the fake names it returns, pre-register malicious skills under them, and wait for a victim to install the payload. We conducted the first large-scale measurement of skill name hallucination, evaluating 15,000 prompts across 12 configurations (4 standalone LLMs and 8 agents). We conservatively counted a name as hallucinated only if it was missing from all live registries and GitHub. The results reveal a systemic vulnerability: every configuration hallucinates. Rates average 36.0% for standalone LLMs and 36.9% for agents, rising to 43.1% on real-world developer questions. In total, the systems generated 5,669 distinct hallucinated names. Crucially, these names are not random noise. Agents repeat the same fake names across prompts and models, giving attackers highly reliable targets to hijack. Finally, we tested four model-level defenses and found a severe conflict between security and usability. The strongest, retrieval grounding, cut the hallucination rate from 40.8% to 3.2% but crippled usefulness: even the best-defended system recommended the correct skill only about one in six times. Skill name hallucination is thus a highly exploitable vulnerability requiring minimal attacker effort. Fixing it cannot rely on prompt engineering or model tuning alone. It demands ecosystem-wide structural changes: registry-level name reservations and verified recommendation pipelines.

Discovering vulnerabilities before attackers exploit them requires high recall and reliable automatic validation, but existing approaches struggle to achieve both without prohibitive cost. We present Antiproof, an end-to-end vulnerability discovery system that combines neuro-symbolic detector synthesis for high-recall discovery with proof-of-exploitability oracles for automatic validation. Antiproof learns and iteratively refines static detectors from vulnerability datasets, then validates candidates by verifying whether executable proofs demonstrate concrete attacker capabilities. Evaluated on BountyBench and our curated KEVBench dataset, Antiproof detects 64 of 66 vulnerabilities, improving recall by more than 60 percentage points over static-analysis and neuro-symbolic baselines. In a scan of 50 widely deployed systems, Antiproof uncovered several hundred previously unknown vulnerabilities. We are responsibly disclosing all confirmed zero-days and have received 12 CVE assignments to date, including remote code execution vulnerabilities in Ray, SGLang, vLLM, and LiteLLM that could allow attackers to take over LLM training and inference systems.