Papers

Developers may reference vulnerabilities in pull request discussions through both explicit identifiers, such as CVEs or GHSAs, and implicit security-related language (e.g., "unauthorized access" or "SQL injection"). Prior work has primarily focused on explicit identifiers, potentially overlooking vulnerability discussions that lack formal references. Bots and coding agents are becoming more common in pull requests, raising new questions about how different accounts communicate about vulnerabilities. In this registered report, we describe our planned study of vulnerability communication in pull requests by humans, bots, and coding agents. Building on the AIDev-pop dataset, we analyze explicit vulnerability references and implicit security-related signals across pull request titles, descriptions, review comments, commit messages, and timeline discussions. We further investigate whether these references are associated with vulnerabilities introduced or fixed in the modified code and how they relate to pull request review activity and outcomes. This study contributes a large-scale empirical investigation of vulnerability communication practices in modern software development.

The rapid proliferation of automated, multi-vector malware threats poses a significant risk to heterogeneous, resource constrained cyber-physical networks. Conventional epidemiological models often treat security defenses as static parameters, failing to capture the strategic, asymmetric maneuvers between an attacker and a defender. To address the gap, this paper proposes a Game-Theory-Integrated Modified Multi- Wireless Sensor Epidemic Malware Propagation (GTI-mSEMP) framework. This paper analyzed and compared the operational trajectories of Susceptible (S) and Recovered (R) node populations across three different operational regimes: Balanced Matchup, Exploit Surge and Hardened Defense. Numerical simulation results capture the real-time transient dynamics of the network state variables, demonstrating how the epidemic curve shifts when either the defensive or offensive scaling vectors hold an efficiency advantage. The proposed mathematical and numerical framework provides a rigorous foundation that can be deployed in highly adversarial network environments to evaluate dynamic malware propagation and predict localized node population states.

Large language models (LLMs) have increasingly moved from standalone text generation systems to agents that invoke external tools, access environments, and execute multi-step tasks. However, conventional function-calling benchmarks mainly evaluate task completion and API correctness, while privacy evaluation benchmarks typically focus on final responses or privacy judgments. Neither perspective captures purpose-bound information flow across an executed multi-tool trajectory. Motivated by this limitation in current agent evaluation, ToolPrivacyBench audits whether task-private atoms are routed only to authorized tools and downstream sinks, thereby evaluating both task completion and privacy over-disclosure during tool use. The benchmark contains 2,150 cases, including 1,150 fully synthetic privacy-sensitive business workflows and 1,000 cases adapted from existing multi-tool and function-calling benchmarks. Each case is represented by a policy knowledge base. After an agent executes against mock business backends, the evaluator compares recorded tool arguments and backend audit logs with this policy knowledge base. The evaluation covers nine widely used agents to characterize purpose-bound privacy over-disclosure. The results show that successful tool execution does not imply appropriate privacy disclosure: an agent may complete a task while transmitting unnecessary private information through intermediate tool calls. ToolPrivacyBench therefore formalizes a need-to-know disclosure boundary, under which each tool should receive only the information necessary for its stated purpose, and uses trajectory-level auditing to identify privacy over-disclosure in multi-tool workflows.

Cyber deception research has focused on improving honeypot deception capabilities to increase attacker engagement and extend their interactions to collect more and better intelligence. For SSH honeypots, this relies on the assumption that attackers log in, open a shell, and type. We tested whether this still held by deploying eleven SSH honeypots that served both interactive and non-interactive session requests for fifteen days. We collected 177,622 authenticated sessions and validated our results against an independent Cowrie dataset over the same time window. We found that 99.23% of sessions were non-interactive. Interactive sessions account for only 0.10%. The same pattern held in the comparative third-party dataset used for evaluation. This finding is important because a honeypot that focuses on interactive shells or evaluates success based on session length and the number of commands can miss most authenticated attacks and draw the wrong conclusions about what attackers do after login.

Threshold private set intersection (TPSI) allows parties to reveal their intersection only when its cardinality reaches a prescribed threshold. Existing quantum TPSI protocols typically rely on a third party (TP) to interpret the final results, which deviates from the cardinality-testing paradigm of TPSI. In this paper, we propose a quantum multiparty TPSI protocol with explicit cardinality testing. Our protocol develops a rotation-based quantum construction in which single-photon sequences are sequentially processed through participant-side data rotations, TP–participant masking rotations, and correlated aggregate rotations. This design produces hidden-label measurement vectors: TP can complete the final measurement, but cannot interpret the semantic meaning of the outcomes. Based on these hidden measurements, we further realize the threshold decision through an oblivious linear evaluation (OLE)-based inner product procedure and a lightweight garbled circuit, revealing only \(\mathbf 1[|\bigcap_i X_i|\ge \tau]\) before conditional intersection reconstruction. We prove the correctness and security of the proposed protocol, and further validate its feasibility through quantum-circuit simulations implemented on the IBM Qiskit platform.

Threshold private set intersection (TPSI) allows parties to reveal their intersection only when its cardinality reaches a prescribed threshold. Existing quantum TPSI protocols typically rely on a third party (TP) to interpret the final results, which deviates from the cardinality-testing paradigm of TPSI. In this paper, we propose a quantum multiparty TPSI protocol with explicit cardinality testing. Our protocol develops a rotation-based quantum construction in which single-photon sequences are sequentially processed through participant-side data rotations, TP–participant masking rotations, and correlated aggregate rotations. This design produces hidden-label measurement vectors: TP can complete the final measurement, but cannot interpret the semantic meaning of the outcomes. Based on these hidden measurements, we further realize the threshold decision through an oblivious linear evaluation (OLE)-based inner product procedure and a lightweight garbled circuit, revealing only \(\mathbf 1[|\bigcap_i X_i|\ge \tau]\) before conditional intersection reconstruction. We prove the correctness and security of the proposed protocol, and further validate its feasibility through quantum-circuit simulations implemented on the IBM Qiskit platform.

LLM-based SSH honeypots can generate believable interactions, but evaluations indicate they remain somewhat identifiable to determined attackers, indicating the need for a better scaffolding. We present a new LLM-based honeypot design that uses a multi-agent, multi-LLM architecture to address the limitations of the previous shelLM LLM honeypot. Our honeypot, called AdvancedShelLM, uses two LLM agents, a Manager and a Worker, that better understand the commands while reducing incorrect responses and increasing deception. It implements an advanced permanent filesystem, allowing many simultaneous attackers to see the same changing files for the first time. It was evaluated with: (i) unit tests for generative capabilities, (ii) an AI attacker (ARACNE) to assess realism and deception, (iii) human attackers to assess its deceptive capability, and (iv) an Internet deployment to evaluate deception in real-world attacks. In unit test results, AdvancedShelLM achieved a pass rate of up to 99.02%. The AI attacker ARACNE had issues making a decision if the system is honeypot or not, but showed slight bias towards saying honeypot, even for a real Ubuntu shell. With human attackers, AdvancedShelLM deceived more humans than Cowrie, but had similar results as shelLM. The Internet deployment showed concrete evidence that the output of AdvancedShelLM can influence the behaviour of real-life attackers.

Dense retrieval systems expose document geometry when vector stores are compromised, and a global protective transform can often be aligned from known pairs. We study SHARD, which splits PCA coordinates into a short routing prefix and a residual protected by independent cell-local orthogonal keys. It supports CKKS ciphertext–plaintext reranking but is evaluated as a leakage trade-off, not a cryptographic document-privacy guarantee. Corrected scoring uses centered document coordinates and an uncentered scoring query, preserving raw ranking up to a query-dependent constant. Across ten BEIR/MIRACL configurations it reproduces raw nDCG@10 and recall, whereas centering both sides loses up to 0.080 nDCG. Cell keys spread diffuse known-pair evidence across compartments, but minimum-norm alignment recovers useful signal far below full key rank, so there is no hard de-anonymization threshold. Real CKKS has maximum score error 2.29e-6 and no top-1 flips; block packing cuts query upload by 74–87% but raises in-process p50 latency by 14–26%. In a strengthened GTR case, an unknown key lowers token-F1 from 0.665 to 0.242; a wide prefix and eight pairs restore much. Under 25–90% release overlap, the unchanged prefix and clean residual norm link persistent rows with R@1 at least 0.9996, although cell-Gram linkage degrades under churn. A formally calibrated Gaussian release gives nDCG@10 at most 0.011 at epsilon=1; its only three strict utility matches occur at epsilon=32768 with linkage R@1 at least 0.995. SHARD preserves retrieval and compartmentalizes alignment evidence, but does not provide DP, unlinkability, or cancellable templates.

Cyber deception research often assumes that a decoy can be placed wherever there is attacker behavior. This work tests that assumption across MITRE ATT&CK v18.1. We introduce a four-criterion rubric for infrastructure deception and apply it to all 250 ATT&CK techniques. The rubric evaluates whether a defender-controlled decoy can be placed, whether an attacker is likely to interact with it, what intelligence that interaction can yield, and whether the interaction reliably indicates malice. The resulting deception surface is sparse: only 80 techniques (32%) admit a decoy the attacker could plausibly reach. For the remaining 170 techniques, there is no defender-controlled asset in the attacker's path that can be fabricated as a decoy. Decoy placement across those 80 techniques falls into two patterns we call Sweep and Seek. In Sweep, the attacker moves broadly through assets in range and encounters the decoy as part of that activity. In Seek, the attacker looks for a specific kind of asset and interacts with a fabricated version of it. These patterns give a simple placement rule: a decoy must either sit on a sweep path or imitate a sought asset. We also show that decoys usually have useful intelligence potential, but whether an attacker interacts with them at all, and whether that interaction reliably indicates malice, both vary. We release the rubric, decision rules, and per-technique assessment as an auditable baseline for future deception research and deployment planning, and show that infrastructure decoys cannot be assumed to apply to all attacker behavior.

Given \(H\leq G\) finite abelian groups, a transversal \(T\subseteq G\) for \(G/H\) has fixed size \(|G/H|\), but its ambient difference support \(D(T)=T-T\) can vary with the embedding of \(H\) in \(G\). We call $ \delta(G,H)=\min_T |D(T)| $ the transversal difference number of the pair \((G,H)\). This invariant is related to finite abelian factorisation, tiling complements, and small-sumset questions, and is motivated by recent work regarding ambient Galois labels in CRT transforms for cyclotomic-subfield homomorphic encryption. We prove various results regarding this invariant, including a general lower bound $\delta(G,H)\geq 2|G/H|-m(G,H), $ where \(m(G,H)\) is the largest order of a subgroup of \(G\) disjoint from \(H\). The bound is sharp for cyclic quotients, and Kneser's theorem gives a cross-transversal estimate leading to exact product families with one nonsplit cyclic coordinate and arbitrary split factors. These results isolate the first genuinely new residual obstruction, namely the same-prime square plane \[ G=(\mathbb Z/p^2\mathbb Z)^2,\qquad H=pG. \] For odd \(p\), this case is the technical core of the paper. Here transversals are graphs of functions \(\mathbb F_p^2\to \mathbb F_p^2\), and \(D(T)\) decomposes into carry-corrected finite-field derivative images. We conjecture that \[ \delta(G,H)=(2p-1)^2 \] for all odd primes \(p\), prove the unconditional lower bound \(3p^2-p-1\), and give small-prime, probabilistic, and fixed-polynomial evidence for the conjecture.

Phone-use Agents can execute complex tasks end to end across real mobile applications. By operating a real device on the user's behalf, they reach far more functionalities than CLI agents, which amplifies the real-world harm they can cause when driven for malicious purposes. We present the first study of this threat on real phones and 27 commercial apps, and find that agents built on 9 mainstream commercial and open-source models readily carry out serious misuse, ranging from procuring drug and explosive precursors to fraud, online harassment, and review manipulation. Across the agents we run on real devices, the average refusal rate to harmful requests stays low while the average task-completion rate reaches 68.8%, and in some scenarios an agent finishes a violation faster than a human would. These results suggest that Phone-use Agents already meet the practical conditions for automated misuse at scale. In one observed real-device execution, Claude-Opus-4.8 fabricated a medical history, deceived an online doctor into issuing a prescription, and completed the order and payment on its own to purchase a precursor for a highly toxic substance. To our knowledge, this is the first documented real-world case of an AI agent procuring controlled precursor materials. We trace this behavior to a Safety Awareness-Execution Gap, where an agent recognizes that a request is harmful yet still executes it. Simple defenses curb the overt cases, but the more covert and arguably more damaging threats, such as coordinated review manipulation and fake traffic, remain largely unsolved. We hope these findings push the community toward safer Phone-use Agents.

The widespread collection of fine-grained location data by commercial data brokers creates a re-identification risk that is not widely recognised by the public. While prior research has established that mobility traces are highly unique and that individuals can, in principle, be identified from a handful of spatio-temporal points, such attacks have historically required significant manual effort from skilled analysts, limiting their practical scale. In this feasibility study, we demonstrate in a real world setting that agentic AI fundamentally changes this threat model. We present an end-to-end pipeline in which large language model agents autonomously search the open web, cross-reference public records and social media, and resolve raw coordinate sequences to candidate identities - without human intervention. We evaluate the pipeline on a spatio-temporal dataset containing simulated location points anchored at and around true home and work addresses, focusing on a high-risk disclosure scenario. Our results demonstrate that, from spatio-temporal data and public sources alone, our agentic AI successfully re-identified 18 of the 25 re-identifiable individuals (72%) and 18 of 43 cases overall (41.9%). We discuss implications for Statistical Disclosure Control (SDC) practice and outline the near-future escalation that data custodians and regulators must anticipate. De facto anonymity - an implicit foundation of SDC practice - is shifting. Agentic AI strengthens the case that re-identification is reasonably likely by any means under the GDPR Recital-26 standard, at costs of minutes-and-dollars per target.

Performance numbers reported for hardware are accepted on trust: the reader cannot recompute them, the apparatus is gone, and the silicon itself can be silently wrong, with fleet studies reporting on the order of one core in a thousand returning incorrect arithmetic with no error raised. We make a reported hardware measurement a tamper-evident, independently checkable record. Every quantity in the text, a table, or a figure is bound, by its content hash, to the observation and the verification behind it; the whole is a hash-linked, append-only structure (a transparency log for measurement) that a verifier audits offline without trusting its producer. Matrix products are verified by a probabilistic identity (Freivalds) at O(k n^2) cost under a tolerance we derive from floating-point error analysis and calibrate to the device's own measured residual floor, so a wrong product is rejected with probability 1 - 2^(-k); quantities with no such identity carry an algebraic checksum and a measured reproducibility class. We then treat the check itself as a security object: a probe seed committed for offline reproducibility is an attack surface, and a probe-aware adversary can hide a corruption in the probe's null space, fooling even a quorum of bit-identical witnesses, while a Fiat-Shamir challenge derived from the claimed output closes this. Driving the device from an unprivileged tenant's reach, with a di/dt power virus and a thermal soak, neither moves the calibrated tolerance nor produces a silent error, placing the physical-fault threat at the rare defective part or the privileged attacker and marking the boundary at which the record must compose with a hardware root of trust. We demonstrate the construction across Blackwell and Hopper GPUs and report a residual-floor and reproducibility map by precision, size, and device.

Traditionally, the architecture of high-performance computing (HPC) systems is tailored for speed, while highly secure computer systems must sacrifice speed for security. However, a wide range of scientific domains, such as the life sciences, call for a combination of performance and security to allow processing sensitive data at scale. Here, we present RAMSES (Research Accelerator for Modeling and Simulation with Enhanced Security), an HPC system designed from the ground up to deliver high performance within a robust security framework. RAMSES integrates hardware-based memory encryption of AMD processors with state-of-the-art file encryption from IBM Storage Scale and the Thales CipherTrust manager, establishing an HPC platform that ensures continuous encryption throughout the data life cycle - at rest, in transit, and in use - in compliance with major data protection standards (European General Data Protection Regulation, ISO/IEC 27001 certification, and Federal Information Processing Standards). In addition, we implemented advanced operating system hardening, a multi-layered security architecture, and mandatory multi-factor authentication to adapt the HPC environment to increased security demands. Benchmark results from the biomedical sector demonstrate that the performance impact of the secure environment is limited and that integration of the conflicting requirements speed and security can be achieved while preserving a coherent, flexible, and user-friendly system.

Time-triggered communication protocols rely on trusted components known as guardians to enforce adherence to predetermined network schedules. Network-agnostic guardians offer an efficient and scalable distributed solution with reduced implementation cost and complexity compared to network-aware alternatives. However, this efficiency is based on the guardian's dependence on the controlled node for clock synchronization, which introduces a vulnerability: a malicious node can exploit this dependency to launch timing attacks against its guardian and eventually interfere with messages from other nodes on the network. In this paper, we establish a theoretical lower bound on the attainable clock synchronization precision between a node and its network-agnostic guardian. Building on this result, we introduce a timing attack that leverages the unavoidably imperfect clock synchrony to cause controlled and undetected de-synchronization of the guardian. The attack enables a malicious node to cause collisions with targeted critical network messages. We evaluate the effectiveness of the attack using a FlexRay field bus network model implemented in the OMNeT++ simulation framework. Our results show that the attack is able to remain undetected with 100% success and disrupts the transmission of the critical messages of the target node by causing collisions with them with 100% success.

Fuzzy Labeled Private Set Intersection (FLPSI) lets a receiver learn the labels of enrolled records similar to its query, and nothing else. Constructions based on a set-threshold reduction reach practical performance: a query matches a record when the two agree on a threshold number of components, and the private matching is delegated to an inner set-threshold kernel. We study its homomorphic form, which combines leveled-BFV homomorphic encryption (HE), a garbled circuit, and secret sharing to decide the match under encryption and release the record's label. We identify a composition gap in this kernel: efficiency is bought with a per-trial false-accept probability, but one query runs a trial for every record, so the error compounds with the database size into the kernel's realization soundness error (RSE), the rate at which it accepts a query the plaintext matcher would reject. The RSE is a reliability property of the cryptographic matching layer, not the matcher's accuracy, and a sound kernel must contribute zero or negligible RSE of its own. We formalize this as a composable security property, give a closed-form bound on the receiver's advantage, and close the gap with CSTPSI, a kernel that runs independent token rounds and raises the per-trial bound to a matching power. We prove CSTPSI secure in the semi-honest model. The bound sets the round count: two token rounds suffice for million-scale databases and three for billion-scale at the $10^{-6}$ engineering threshold. Our evaluation confirms this: at a million records the baseline kernel's RSE reaches 100% while CSTPSI holds it at 0 in every measured configuration. For large labels at small to moderate scale CSTPSI is more than 20x faster than the baseline, with up to 93% less communication, converging to the baseline only at million-scale. Our implementation, with a one-command reproducibility harness, is publicly available.

The rapid spread of fake news poses increasing threats to information ecosystems, especially as AI-generated misinformation under Generative Engine Optimization (GEO) poisoning allows adversarially crafted content to be systematically surfaced by retrieval systems, contaminating LLM reasoning. In this paper, we propose Tree of Evidence (ToE), a hierarchical evidence reasoning framework for automated fact-checking that models each claim as a dynamically expanding argument tree. ToE integrates a reinforcement learning-driven multi-source retrieval agent, an evidence evaluation agent, and an argument tree aggregation algorithm to iteratively decompose, retrieve, and verify claims through an explainable evidence chain. We further provide a theoretical analysis of the retrieval process, deriving a formal error bound that guarantees the learned policy converges to a neighborhood of the information-theoretically optimal policy. Experiments across multiple datasets and backbone LLMs demonstrate that ToE achieves improvements ranging from 4 to 24 percentage points over competitive baselines, with particularly pronounced gains on adversarially poisoned inputs.

TinyML models deployed on edge devices are increasingly adopted in safety/security-critical applications, making them a prime target for adversarial example (AE) attacks where inputs are modified to cause misclassifications. However, existing AE detection methods either require white-box model access, which is often unavailable in licensed black-box deployments, or rely on input pre-processing stages that add non-trivial latency and resource overhead, often exceeding what mission-critical applications can afford on their inference path. To address these challenges, we propose AdvScan, a runtime power analysis-based methodology for AE detection that operates in a black-box scenario while inducing minimal latency. AdvScan is based on the observation that AEs produce anomalous neuron activations, which in turn generate distinctive power-consumption signatures. The algorithm initially constructs a baseline distribution of power signatures from known benign inputs; then, at runtime, it applies a one-sample t-test to determine whether a test input's power signature significantly deviates from this baseline, thereby detecting AEs. We evaluated AdvScan using three adversarial example generation algorithms: Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and Carlini-Wagner (C&W), on three MLPerf Tiny benchmark models implemented on two target devices: the STM32F303RC (ARM Cortex-M4) and STM32L562RE (ARM Cortex-M33) microcontrollers. Across 318,400 total test inputs, AdvScan detects 99.984% of AEs with only 40 false negatives and zero false positives. These results demonstrate the viability of power-based AE detection for secure, accuracy-critical TinyML deployments in black-box environments.

While voice control is rapidly becoming a ubiquitous vector of human-AI communication, the risks facing these systems remain poorly understood. This is, in part, a product of the difficulties in scaling strictly digital adversarial workflows to the physical world. These scale barriers have led the community to abstract away key acoustic factors relating to detectability and the influence of geometry on acoustics. These methodological and metrological shortcomings undermine our understanding of risk. We illuminate these issues through real-world testing, conceptual discussions, and a novel, high-throughput reality simulation framework. By testing over 8 million adversarial evaluations, we demonstrate that acoustic awareness yields relative Word Error Rate increases of up to 94.5% under Whisper and wav2vec. We employ this framework to explore a formalize and operationalize a Dual-Form Signal to Noise Ratio to decouple source stealth from victim attack efficacy, resolving a crucial limitation in current works. This lays the groundwork for repeatable, verifiable research that embraces, rather than abstracts, the acoustic environment.

Automatic Speech Recognition systems are notoriously both sensitive to adversarial and benign perturbations. While this has been repeatedly demonstrated using reference datasets, detecting such behaviors in deployed systems is incredibly challenging, due to the absence of oracle knowledge of the true transcription. We demonstrate that employing a certification-inspired mechanism can significantly decrease WER, increase recall, and decrease the Spearman correlation between confidence and WER. We achieve this through a dual-gate diagnostic pipeline: a Two-Sided Atomic Audit that accumulates statistical wealth to certify both token existence and adversarial exclusion, and a Rank-Based Tournament that selects the winning sequence. Our evaluations across four diverse architectures demonstrate up to a 55% relative reduction in Word Error Rate, while also providing granular word- and sentence-level certifications to enhance acoustic security.

Randomized Smoothing (RS) provides rigorous robustness guarantees for neural networks without architectural constraints, yet its adoption is limited by extreme computational costs. Standard RS requires tens of thousands of model evaluations per input and forces practitioners to commit to fixed sample sizes a priori. In this work, we present a novel meta-learning framework for anytime-valid certified robustness that adaptively deploys computational resources. By using a lightweight meta-learner to predict image-specific priors for a sequential E-process, we achieve a 20-fold reduction in sample complexity compared to traditional methods while maintaining rigorous statistical guarantees. Beyond raw efficiency, we demonstrate how anytime-validity enables adaptively allocating compute based upon application-specific risk thresholds, a form of resource triage impossible under classic certification frameworks. That this is achievable while also providing similar certification performance demonstrates that our approach provides a pathway for real-time, safety-critical certification deployments.

LLM-based agents for automated code repair have received significant attention in recent years from both research and software engineering practice perspectives. However, limited attention has been paid to patches that pass syntactic and functional verification but still retain or introduce security vulnerabilities. The aim of this research is to systematically identify and categorize such silent failures in LLM-based agentic code repair. We conducted an empirical study using 1,030 valid execution traces produced by seven agent frameworks with GPT-4o-mini across two security-focused datasets, SecurityEval and CVEfixes. Through three iterations of qualitative coding and manual verification, 170 confirmed silent failures were identified. The key results are: (i) Three main categories of silent failures were identified: Omission, Introduction, and Inadequacy. Omission accounts for 48.2% of the confirmed failures, Introduction for 30.6%, and Inadequacy for 21.2%. (ii) Ten fine-grained failure codes were classified under these three categories, showing how agents omit required security controls, apply incomplete defenses, or introduce new vulnerabilities during repair. (iii) Current test-passing evaluation and LLM-based reviewer roles were insufficient to expose or intercept these failures in the confirmed cases. (iv) Similar insecure solutions appeared across different frameworks, suggesting possible shared model-, prompt-, or task-level influences, while single-agent and multi-agent systems showed different failure profiles. The results of this study will assist researchers and practitioners in improving the evaluation of LLM-based agentic code repair and developing targeted verification methods that go beyond functional correctness and cover all generated artifacts.

Increasingly autonomous agentic AI systems pose novel multi-agent risks, such as secret collusion via covert communication channels. The natural defence to these collusion attempts is to monitor plain-text communication, but the efficacy of monitors has been called into doubt by increasingly sophisticated model steganography; indeed, some theoretical schemes have been proposed that are information-theoretically or computationally indistinguishable from good-faith plain-text communication. In this paper, we demonstrate that the complexity of these schemes is no longer a safety barrier, as agentic coding models can already produce undetectable stegosystems when given realistic tool usage, such as code execution or accessing research papers through web searches. Agents also adapt when key ingredients are missing, for example, by adding model-sampling components or implementing related keyed coding schemes. We then frame tacit steganographic coordination between agents as a Schelling-point problem and introduce coordination metrics for estimating when two agents are likely to select compatible schemes without explicit prior agreement. Our results suggest a shift in the threat model for covert communication between AI agents, where the main barrier is no longer whether frontier agents can understand and implement sophisticated stegosystems, but coordination: whether independently acting agents can converge on compatible schemes, keys, and parameters. We find substantial convergence on broad scheme families but limited strict one-shot coordination, suggesting that shared artefacts, repeated interaction, and tool-mediated search are the settings where covert communication risks are most acute. Overall, our findings provide empirical grounding for the recent strategic confinement hypothesis, which assumes that capable agents can construct covert channels that survive monitoring.

Prompt injection is the top security risk for LLM-integrated applications, yet every defense proposed so far has been broken. We prove this is not a coincidence: in shared-embedding architectures that lack enforced control-data separation, perfect prompt-injection prevention is mathematically impossible. We formalize prompted systems as Prompted Action Models whose outputs include control-authoritative actions: refusal decisions, tool authorization, policy routing, and memory writes. We define Semantic-Faithful Control (SFC), the property that such behavior depends only on the meaning of untrusted input, not on how it is encoded. We then prove SFC is unachievable within the shared pipeline, via three results: a provenance-recovery impossibility (shared representations make trusted and untrusted content statistically inseparable, bounded by total variation distance); control-path exposure (untrusted tokens enter control-relevant computation through the same attention value-aggregation that determines outputs); and a finite-coverage invariance gap (finite training cannot certify invariance over infinite semantic-equivalence classes). We ground each quantity in measurements on production tokenizers and models. The result is structural, not a gap in current defenses. It mirrors the code-data confusion in Von Neumann machines that gives rise to buffer overflows, a vulnerability class that took decades of layered defenses (DEP, Write-XOR-Execute, ASLR, stack canaries, and ultimately memory-safe languages) to contain, because no single mechanism sufficed. The implication is the same: prompt injection cannot be eliminated by better in-pipeline classification or alignment alone. It requires architectural separation of instruction and data channels. We identify the root cause and the class of solution it demands.

Fairness measurements in the form of disaggregated evaluations often rely on demographic signals that are legally constrained or culturally sensitive. Race and ethnicity signals are among the more difficult signals to curate and use for this task. This paper presents Privacy-Preserving Probabilistic Race/Ethnicity Estimation (PPRE) as a method for enabling fairness measurements with respect to race/ethnicity for U.S.\ LinkedIn members in a privacy-preserving manner. PPRE applies privacy technologies (specifically: secure two-party computation, differential privacy, and additive homomorphic encryption) on top of two race/ethnicity demographic signal sources (the Bayesian Improved Surname Geocoding estimator and a sparse golden survey set of self-reported demographics) to power a fairness measurement solution with respect to US-based race/ethnicity demographics. We detail its privacy guarantees and demonstrate its application on candidate- and viewer-side fairness measurements. We close with a transferable framework for institutions seeking to implement similar privacy-preserving measurement infrastructure.

Large Language Model (LLM)-based question-answering (QA) systems are increasingly deployed in sensitive domains such as healthcare, mental health counseling, and legal consultation. Federated learning (FL) enables collaborative training without sharing raw client data, for which locally trained models are aggregated at a central server (i.e., a cloud service provider) to obtain a global model. In this paper, we explore the potential vulnerability where a malicious aggregator, who may collude with a third-party vendor, stealthily implants advertisement-type backdoors into federated QA models, without ever accessing client data. The attacker's goals are twofold: (1) preserve clean QA fidelity (i.e., the poisoned model behaves like a clean model on non-triggered queries); and (2) generate highly natural, contextually relevant responses with target advertisements when a trigger appears. Achieving these two goals simultaneously is highly challenging, as naive backdoor injection without knowledge about private data may degrade model's clean performance or fail to inject the target. Motivated by this, we propose to leverage clients' uploaded gradients during training, and develop a two-stage framework for data-free and stealthy poisoning: (1) recover representative training samples from client gradients, and (2) construct poisoning datasets utilizing recovered samples and trigger phrases to inject backdoors into the global model. Experiments across representative QA datasets and LLM families under full fine-tuning and LoRA settings demonstrate that, our method achieves nearly 100% Attack Success Rate (ASR) while incurring negligible degradation on clean tasks. Crucially, reconstructing only 5-20% of gradients suffices to mount a reliable attack, exposing a practical blind spot in the pipeline of federated training of QA LLMs.

Decentralized Finance (DeFi) applications rely heavily on the order in which transactions are executed, making them susceptible to reordering attacks that enable adversaries to extract Blockchain Extractable Value (BEV). While linear blockchain systems such as Ethereum have inspired extensive research into fair ordering mechanisms, DAG-based consensus protocols have remained largely unprotected despite their growing adoption for scalability and performance. In this paper, we introduce Tilikum, a DAG-based ledger protocol that ensures fair transaction ordering without relying on weak edges. Tilikum achieves ordering linearizability by leveraging median-based timestamp aggregation, or batch order fairness, while maintaining low data redundancy and robust garbage collection. We implemented Tilikum in Rust and evaluated it against representative baselines, namely Narwhal/Tusk, Pompē, Themis and FairDAG. Our results show that Tilikum achieves up to $39\times$ higher throughput than other fair-ordering baselines, while fully blocking state-of-the-art DAG-specific reordering attacks.

Impagliazzo's five worlds classify computational assumptions along a single axis, the existence of cryptographic primitives. All five worlds implicitly assume that every party, including the adversary, observes the full input, that the observer is always $O_{top}$. This assumption is so natural that it is never stated. This work makes it explicit and relaxes it by introducing a second, orthogonal axis, the observational axis, defined by the observer hierarchy introduced in previous work. Relaxing the assumption reveals structural phenomena, such as the collapse $P^{O_{prof}} = NP^{O_{prof}} \subset P$, that the five-world framework cannot express. We prove that this collapse holds unconditionally in all five worlds, showing that observational blindness and computational hardness are independent. We define the Observer World $W_O$, classify all world-observer pairs, identify the labeled cells (a)–(d), and introduce a parametric family $W_O^{\varepsilon}$ modelling partial violations of observational invariants. The framework also interfaces with physical information limits, including thermodynamic, quantum, and cosmological bounds.

We introduce PRISM (PE Relational Inter-Section Matrix), an open dataset and feature representation for static Windows PE malware detection. Existing benchmarks such as EMBER, BODMAS, and SOREL-20M represent each PE file as a flat one-dimensional feature vector, discarding the ordering of sections and the relational context between them. PRISM instead encodes every binary as a two-dimensional matrix whose rows are individual PE sections in file order, with a global summary row that preserves compatibility with EMBER-style models. We build the corpus from four malware sources (BODMAS, MalwareBazaar, VirusShare, and CAPE) together with SOREL-20M benign software, yielding 83,633 deduplicated matrices and a family-filtered analysis corpus of 49,204 samples across 684 malware families. A formal separability analysis (Fisher Discriminant Ratio, mutual information, and inter-section information gain) shows that the per-section positional structure carries discriminative information that flat representations cannot capture. Under a strictly controlled, sample-matched comparison, a gradient-boosted classifier on the compact PRISM representation recovers nearly all of the binary-detection performance of the same classifier on the much larger EMBER vector, at roughly one-sixth the dimensionality; EMBER retains only a small, consistent advantage confined to the extreme low-false-positive regime, the two being operationally indistinguishable at the decision threshold. We are explicit that this binary task is saturated, so the structural content PRISM preserves is reserved for tasks with greater metric headroom, such as family classification and architectures that exploit the 2D structure directly. The dataset, extraction library, trained models, and full analysis pipeline are released under CC BY-NC-SA and MIT licences.

We present a novel approach for applying Large Language Models (LLMs) to threat assessment in the context of foreign peacekeeping missions. Building on the PINPOINT project and its use case, the EU Monitoring Mission in Georgia, we combine an interdisciplinary risk-model with OSINT-based media collection and LLM-supported threat extraction. The proposed workflow maps media contents to mission-relevant threats, extracts structured information and applies several additional LLM-based processing steps to improve relevance and grounding. An evaluation of threats extracted from media documents shows high agreement between automatically generated results and human judgment for core aspects such as threat and mission relevance. These results indicate that LLMs provide a promising approach to support analysts in the context of peacekeeping missions.