CyberSec Research

As artificial intelligence (AI) systems are increasingly deployed across critical domains, their security vulnerabilities pose growing risks of high-profile exploits and consequential system failures. Yet systematic approaches to evaluating AI security remain underdeveloped. In this paper, we introduce AVISE (AI Vulnerability Identification and Security Evaluation), a modular open-source framework for identifying vulnerabilities in and evaluating the security of AI systems and models. As a demonstration of the framework, we extend the theory-of-mind-based multi-turn Red Queen attack into an Adversarial Language Model (ALM) augmented attack and develop an automated Security Evaluation Test (SET) for discovering jailbreak vulnerabilities in language models. The SET comprises 25 test cases and an Evaluation Language Model (ELM) that determines whether each test case was able to jailbreak the target model, achieving 92% accuracy, an F1-score of 0.91, and a Matthews correlation coefficient of 0.83. We evaluate nine recently released language models of diverse sizes with the SET and find that all are vulnerable to the augmented Red Queen attack to varying degrees. AVISE provides researchers and industry practitioners with an extensible foundation for developing and deploying automated SETs, offering a concrete step toward more rigorous and reproducible AI security evaluation.

Phishing attacks remain one of the most prevalent threats to online security, with the Anti-Phishing Working Group reporting over 890,000 attacks in Q3 2025 alone. Traditional password-based authentication is particularly vulnerable to such attacks, prompting the development of more secure alternatives. This paper examines passkeys, also known as FIDO2, which claim to provide phishing-resistant authentication through asymmetric cryptography. In this approach, a private key is stored on a user's device, the authenticator, while the server stores the corresponding public key. During authentication, the server generates a challenge that the user signs with the private key; the server then verifies the signature and establishes a session. We present passkey workflows and review state-of-the-art attack vectors from related work alongside newly identified approaches. Two attacks are implemented and evaluated: the Infected Authenticator attack, which generates attacker-known keys on a corrupted authenticator, and the Authenticator Deception attack, which spoofs a target website by modifying the browser's certificate authority store, installing a valid certificate, and intercepting user traffic. An attacker relays a legitimate challenge from the real server to a user, who signs it, allowing the attacker to authenticate as the victim. Our results demonstrate that successful attacks on passkeys require substantial effort and resources. The claim that passkeys are phishing-resistant largely holds true, significantly raising the bar compared to traditional password-based authentication.

LLM agents have begun to find real security vulnerabilities that human auditors and automated fuzzers missed for decades, in source-available targets where the analyst can build and instrument the code. In practice the work is split among several agents, wired together by a harness: the program that fixes which roles exist, how they pass information, which tools each may call, and how retries are coordinated. When the language model is held fixed, changing only the harness can still change success rates by several-fold on public agent benchmarks, yet most harnesses are written by hand; recent harness optimizers each search only a narrow slice of the design space and rely on coarse pass/fail feedback that gives no diagnostic signal about why a trial failed. AgentFlow addresses both limitations with a typed graph DSL whose search space jointly covers agent roles, prompts, tools, communication topology, and coordination protocol, paired with a feedback-driven outer loop that reads runtime signals from the target program itself to diagnose which part of the harness caused the failure and rewrite it accordingly. We evaluate AgentFlow on TerminalBench-2 with Claude Opus 4.6 and on Google Chrome with Kimi K2.5. AgentFlow reaches 84.3% on TerminalBench-2, the highest score in the public leaderboard snapshot we evaluate against, and discovers ten previously unknown zero-day vulnerabilities in Google Chrome, including two Critical sandbox-escape vulnerabilities (CVE-2026-5280 and CVE-2026-6297).

Post-quantum cryptographic (PQC) accelerators for ML-KEM (FIPS 203) and ML-DSA (FIPS 204) rely on pipelined Number Theoretic Transform (NTT) stages over $\mathbb{Z}_q$. Our prior work established structural dependency analysis at scale [1] and quantified the security margin of partial NTT masking [2]. Whether per-stage arithmetic masking guarantees pipeline-level security had no prior machine-checked answer for the r-bearing case: composition frameworks (ISW, t-SNI, PINI, DOM) were formalized exclusively for Boolean masking over $\mathrm{GF}(2)$; no proof assistant artifact addresses the NTT butterfly over $\mathbb{Z}_q$. We present three machine-checked results in Lean 4 with Mathlib, all zero sorry. First, we close a stated limitation of prior work: value-independence implies constant marginal distribution under fresh randomness (via an algebraic MutualInfoZero proxy). Second, butterfly per-context uniformity: for any Cooley-Tukey butterfly with fresh output mask over $\mathbb{Z}/q\mathbb{Z}$ ($q > 0$), each output wire has exactly one mask value producing each output, a uniform marginal independent of secrets, universal over all moduli, twiddle factors, and inputs. Third, a k-stage NTT pipeline with fresh per-stage masking satisfies per-context uniformity at every stage under the ISW first-order probing model. We document a named warning: pointwise value-independence is false for butterfly outputs. The Adams Bridge accelerator (CHIPS Alliance Caliptra) fails the fresh masking hypothesis, masking active only in INTT round 0, architecturally explaining its structural insecurity. Artifact: nine theorems, 1,738 build jobs, zero sorry. Composition for nonlinear gadgets (Barrett) is addressed in forthcoming manuscripts proving Barrett's PF-PINI(2) satisfaction ('one-bit barrier') [3] and k-stage composition for PF-PINI gadgets under fresh-mask renewal [4].

The Internet of Vehicles (IoV) is advancing modern transportation by improving safety, efficiency, and intelligence. However, the reliance on the Controller Area Network (CAN) introduces critical security risks, as CAN-based communication is highly vulnerable to cyberattacks. Addressing this challenge, we propose DAIRE (Detecting Attacks in IoV in REal-time), a lightweight machine learning framework designed for real-time detection and classification of CAN attacks. DAIRE is built on a lightweight artificial neural network (ANN) where each layer contains Ni = i x c neurons, with Ni representing the number of neurons in the ith layer and c corresponding to the total number of attack classes. Other hyperparameters are determined empirically to ensure real-time operation. To support the detection and classification of various IoV attacks, such as Denial-of-Service, Fuzzy, and Spoofing, DAIRE employs the sparse categorical cross-entropy loss function and root mean square propagation for loss minimization. In contrast to more resource-intensive architectures, DAIRE leverages a lightweight ANN to reduce computational demands while still delivering strong performance. Experimental results on the CICIoV2024 and Car-Hacking datasets demonstrate DAIRE's effectiveness, achieving an average detection rate of 99.88%, a false positive rate of 0.02%, and an overall accuracy of 99.96%. Furthermore, DAIRE significantly outperforms state-of-the-art approaches in inference speed, with a classification time of just 0.03 ms per sample. These results highlight DAIRE's effectiveness in detecting IoV cyberattacks and its practical suitability for real-time deployment in vehicular systems, underscoring its vital role in strengthening automotive cybersecurity.

Critical vulnerabilities with Common Vulnerability Scoring System scores of 9.0 or higher pose severe risks to organisations' information systems. Timely detection and remediation are essential to minimise economic and reputational damage from cyberattacks. This paper provides a thorough analysis of the identification and resolution timelines of such critical vulnerabilities. A mixed-methods approach is employed, integrating quantitative data from global vulnerability databases analysing 245,456 Common Vulnerabilities and Exposures records spanning from 2009 to 2024, of which 12.8 % were critical, with qualitative case studies of notable incidents. This methodical combination of quantitative and qualitative data sources enables the identification of patterns and delay factors in vulnerability management. The findings indicate significant delays in public disclosure and patch deployment, influenced by industry-specific factors, resource availability and organisational processes. The paper concludes with a series of actionable recommendations to improve the efficiency of vulnerability responses. Despite faster disclosure, the remediation gap for critical vulnerabilities remains a systemic risk, driven by organisational inertia and system complexity.

Adversarial robustness evaluation underpins every claim of trustworthy ML deployment, yet the field suffers from fragmented protocols and undetected gradient masking. We make two contributions. (1) Structured synthesis. We analyze nine peer-reviewed corpus sources (2020--2026) through seven complementary protocols, producing the first end-to-end structured analysis of the field's consensus and unresolved challenges. (2) Auto-ART framework. We introduce Auto-ART, an open-source framework that operationalizes identified gaps: 50+ attacks, 28 defense modules, the Robustness Diagnostic Index (RDI), and gradient-masking detection. It supports multi-norm evaluation (l1/l2/linf/semantic/spatial) and compliance mapping to NIST AI RMF, OWASP LLM Top 10, and the EU AI Act. Empirical validation on RobustBench demonstrates that Auto-ART's pre-screening identifies gradient masking in 92% of flagged cases, and RDI rankings correlate highly with full AutoAttack. Multi-norm evaluation exposes a 23.5 pp gap between average and worst-case robustness on state-of-the-art models. No prior work combines such structured meta-scientific analysis with an executable evaluation framework bridging literature gaps into engineering.

Autonomous vehicles (AVs) increasingly rely on multi-sensor perception pipelines that combine data from cameras, lidar, radar, and other modalities to interpret the environment. This SoK systematizes 48 peer-reviewed studies on perception-layer attacks against AVs, tracking the field's evolution from single-sensor exploits to complex cross-modal threats that compromise multi-sensor fusion (MSF). We develop a unified taxonomy of 20 attack vectors organized by sensor type, attack stage, medium, and perception module, revealing patterns that expose underexplored vulnerabilities in fusion logic and cross-sensor dependencies. Our analysis identifies key research gaps, including limited real-world testing, short-term evaluation bias, and the absence of defenses that account for inter-sensor consistency. To illustrate one such gap, we validate a fusion-level vulnerability through a proof-of-concept simulation combining infrared and lidar spoofing. The findings highlight a fundamental shift in AV security: as systems fuse more sensors for robustness, attackers exploit the very redundancy meant to ensure safety. We conclude with directions for fusion-aware defense design and a research agenda for trustworthy perception in autonomous systems.

Federated learning (FL) enables training of a global model while keeping raw data on end-devices. Despite this, FL has shown to leak private user information and thus in practice, it is often coupled with methods such as differential privacy (DP) and secure vector sum to provide formal privacy guarantees to its participants. In realistic cross-device deployments, the data are highly heterogeneous, so vanilla federated learning converges slowly and generalizes poorly. Clustered federated learning (CFL) mitigates this by segregating users into clusters, leading to lower intra-cluster data heterogeneity. Nevertheless, coupling CFL with DP remains challenging: the injected DP noise makes individual client updates excessively noisy, and the server is unable to initialize cluster centroids with the less noisy aggregated updates. To address this challenge, we propose PINA, a two-stage framework that first lets each client fine-tune a lightweight low-rank adaptation (LoRA) adapter and privately share a compressed sketch of the update. The server leverages these sketches to construct robust cluster centroids. In the second stage, PINA introduces a normality-driven aggregation mechanism that improves convergence and robustness. Our method retains the benefits of clustered FL while providing formal privacy guarantees against an untrusted server. Extensive evaluations show that our proposed method outperforms state-of-the-art DP-FL algorithms by an average of 2.9% in accuracy for privacy budgets (epsilon in {2, 8}).

As DRAM scaling exacerbates RowHammer, DDR5 introduces per-row activation counting (PRAC) to track aggressor activity. However, PRAC indiscriminately increments counters on every activation -- including benign refreshes -- while relying solely on explicit RFM operations for resets. Consequently, counters saturate even in an idle bank, triggering cascading mitigations and degrading performance. This vulnerability arises from a fundamental mismatch: PRAC tracks the aggressor but aims to protect the victim. We present Per-Victim-row hAmmered Counting (PVAC), a victim-based counting mechanism that aligns the counter semantics with the physical disturbance mechanism of RowHammer. PVAC increments the counters of victim rows, resets the activated row, and naturally bounds counter values under normal refresh. To enable efficient victim-based updates, PVAC employs a dedicated counter subarray (CSA) that performs all counter resets and increments concurrently with normal accesses, without timing overhead. We further devise an energy-efficient CSA layout that minimizes refresh-induced counter accesses. Through victim-based counting, PVAC supports higher hammering tolerance than PRAC while maintaining the same worst-case safety guarantee. Across benign workloads and adversarial attack patterns, PVAC avoids spurious Alerts, eliminates PRAC timing penalties, and achieves higher performance and lower energy consumption than prior PRAC-based defenses.

The April 2026 Claude Mythos sandbox escape exposed a critical weakness in frontier AI containment: the infrastructure surrounding advanced models remains susceptible to formally characterizable arithmetic vulnerabilities. Anthropic has not publicly characterized the escape vector; some secondary accounts hypothesize a CWE-190 arithmetic vulnerability in sandbox networking code. We treat this as unverified and analyze the vulnerability class rather than the specific escape. This paper presents COBALT, a Z3 SMT-based formal verification engine for identifying CWE-190/191/195 arithmetic vulnerability patterns in C/C++ infrastructure prior to deployment. We distinguish two classes of contribution. Validated: COBALT detects arithmetic vulnerability patterns in production codebases, producing SAT verdicts with concrete witnesses and UNSAT guarantees under explicit safety bounds. We demonstrate this on four production case studies: NASA cFE, wolfSSL, Eclipse Mosquitto, and NASA F Prime, with reproducible encodings, verified solver output, and acknowledged security outcomes. Proposed: a four-layer containment framework consisting of COBALT, VERDICT, DIRECTIVE-4, and SENTINEL, mapping pre-deployment verification, pre-execution constraints, output control, and runtime monitoring to the failure modes exposed by the Mythos incident. Under explicit assumptions, we further argue that the publicly reported Mythos escape class is consistent with a Z3-expressible CWE-190 arithmetic formulation and that pre-deployment formal analysis would have been capable of surfacing the relevant pattern. The broader claim is infrastructural: frontier-model safety cannot depend on behavioral safeguards alone; the containment stack itself must be subjected to formal verification.

Machine learning-based static malware detectors remain vulnerable to adversarial evasion techniques, such as metamorphic engine mutations. To address this vulnerability, we propose a certifiably robust malware detection framework based on randomized smoothing through feature ablation and targeted noise injection. During evaluation, our system analyzes an executable by generating multiple ablated variants, classifies them by using a smoothed classifier, and identifies the final label based on the majority vote. By analyzing the top-class voting distribution and the Wilson score interval, we derive a formal certificate that guarantees robustness within a specific radius against feature-space perturbations. We evaluate our approach by comparing the performance of the base classifier and the smoothed classifier on both clean executables and ablated variants generated using PyMetaEngine. Our results demonstrate that the proposed smoothed classifier successfully provides certifiable robustness against metamorphic evasion attacks without requiring modifications to the underlying machine learning architecture.

Approximate nearest neighbor (ANN) search in AI systems increasingly handles sensitive data on third-party infrastructure. Trusted execution environments (TEEs) offer protection, but cost-efficient deployments must rely on external SSDs, which leaks user queries through disk access patterns to the host. Oblivious RAM (ORAM) can hide these access patterns but at a high cost; when paired with existing disk-based ANN search techniques, it makes poor use of SSD resources, yielding high latency and poor cost-efficiency. The core challenge for efficient oblivious ANN search over SSDs is balancing both bandwidth and access count. The state-of-the-art ORAM-ANN design minimizes access count at the ANN level and bandwidth at the ORAM level, each trading-off the other, leaving the combined system with both resources overutilized. We propose inverting this design, minimizing bandwidth consumption in the ANN layer and access count in the ORAM layer, since each component is better suited for its new role: ANN's inherent approximation allows for more bandwidth efficiency, while ORAM has no fundamental lower bounds on access count (as opposed to bandwidth). To this end, we propose a cost-efficient approach, Onyx, with two new co-designed components: Onyx-ANNS introduces a compact intermediate representation that proactively prunes the majority of bandwidth-intensive accesses without hurting recall, and Onyx-ORAM proposes a locality-aware shallow tree design that reduces access count while remaining compatible with bandwidth-efficient ORAM techniques. Compared to the state-of-the-art oblivious ANN search system, Onyx achieves $1.7-9.9\times$ lower cost and $2.3-12.3\times$ lower latency.

The rapid evolution and use of Large Language Models (LLMs) in professional workflows require an evaluation of their domain-specific knowledge against industry standards. We introduceCyberCertBench, a new suite of Multiple Choice Question Answering (MCQA) benchmarks derived from industry recognized certifications. CyberCertBench evaluates LLM domain knowledgeagainst the professional standards of Information Technology cybersecurity and more specializedareas such as Operational Technology and related cybersecurity standards. Concurrently, we propose and validate a novel Proposer-Verifier framework, a methodology to generate interpretable,natural language explanations for model performance. Our evaluation shows that frontier modelsachieve human expert level in general networking and IT security knowledge. However, theiraccuracy declines in questions that require vendor-specific nuances or knowledge in formalstandards, like, e.g., IEC 62443. Analysis of model scaling trend and release date demonstratesremarkable gains in parameter efficiency, while recent larger models show diminishing returns.Code and evaluation scripts are available at: https://github.com/GKeppler/CyberCertBench.

Memory analysis is a crucial technique in digital forensics that enables investigators to examine the runtime state of a system through physical memory dumps. While significant advances have been made in memory forensics, the detection and analysis of Thread Local Storage (TLS) callbacks remain challenging due to their dual nature as both legitimate Windows constructs and potential vectors for malware execution. An early version of the TlsCheck plugin received recognition in the Volatility Plugin Contest 2024. In this paper, we present an enhanced version of TlsCheck for Volatility 3, designed to detect and analyze TLS callbacks in process memory. It implements precise detection of TLS callback tables through analysis of PE headers and memory structures, combined with disassembly of identified callback routines. The plugin supports both 32-bit and 64-bit architectures, offering investigators insights into callback locations, assembly behavior, and potential signs of suspicious activity. To enhance detection, we incorporate pattern matching using custom regular expressions and YARA rules, helping analysts identify specific code patterns or suspicious constructs within TLS callbacks. The framework also includes instruction-level analysis to highlight behavior often linked to malware, such as anti-debugging, code injection, and process manipulation. This implementation significantly improves defenders' ability to detect and investigate TLS-based threats during memory forensics, supporting more effective malware analysis and incident response operations.

With the popularity of the large language models (LLMs), text steganography has achieved remarkable performance. However, existing methods still have some issues: (1) For the white-box paradigm, this steganography behavior is prone to exposure due to sharing the off-the-shelf language model between Alice and Bob.(2) For the black-box paradigm, these methods lack flexibility and practicality since Alice and Bob should share the fixed codebook while sharing a specific extracting prompt for each steganographic sentence. In order to improve the security and practicality, we introduce a black-box text steganography with a dynamic codebook and multimodal large language model. Specifically, we first construct a dynamic codebook via some shared session configuration and a multimodal large language model. Then an encrypted steganographic mapping is designed to embed secret messages during the steganographic caption generation. Furthermore, we introduce a feedback optimization mechanism based on reject sampling to ensure accurate extraction of secret messages. Experimental results show that the proposed method outperforms existing white-box text steganography methods in terms of embedding capacity and text quality. Meanwhile, the proposed method has achieved better practicality and flexibility than the existing black-box paradigm in some popular online social networks.

Modern ransomware exhibits polymorphic and evasive behaviors by frequently modifying execution patterns to evade detection. This dynamic nature disrupts feature spaces and limits the effectiveness of static or predefined models. To address this challenge, we propose TL-RL-FusionNet, a reinforcement learning (RL)-guided hybrid framework that integrates frozen dual transfer learning (TL) backbones as feature extractors with a lightweight residual multilayer perceptron (MLP) classifier. The RL agent supervises training by adaptively reweighting samples in response to variations in observable ransomware behavior. Through reward and penalty signals, the agent prioritizes complex cases such as stealthy or polymorphic ransomware employing obfuscation, while down-weighting trivial samples including benign applications with simple file I/O operations or easily classified ransomware. This adaptive mechanism enables the model to dynamically refine its strategy, improving resilience against evolving threats while maintaining strong classification performance. The framework utilizes dynamic behavioral features such as file system activity, registry changes, network traffic, API calls, and anti-analysis checks, extracted from sandbox-generated JSON reports. These features are transformed into RGB images and processed using frozen EfficientNetB0 and InceptionV3 models to capture rich feature representations efficiently. Final classification is performed by a lightweight residual MLP guided by an RL (Q-learning) agent. Experiments on a balanced dataset of 1,000 samples (500 ransomware, 500 benign) show that TL-RL-FusionNet achieves 99.1% accuracy, 98.6% precision, 99.6% recall, and 99.74% AUC, outperforming non-RL baselines by up to 2.5% in accuracy and 3.1% in recall. Efficiency analysis shows 55% lower training time and 59% reduced RAM usage, demonstrating suitability for real-world deployment.

Fundamental rate-distortion-perception (RDP) trade-offs arise in applications requiring maintained perceptual quality of reconstructed data, such as neural image compression. When compressed data is transmitted over public communication channels, security risks emerge. We therefore study secure RDP under negligible information leakage over both noiseless channels and broadcast channels, BCs, with correlated noise components. For noiseless channels, the exact secure RDP region is characterized. For BCs, an inner bound is derived and shown to be tight for a class of more-capable BCs. Separate source-channel coding is further shown to be optimal for this exact secure RDP region with unlimited common randomness available. Moreover, when both encoder and decoder have access to side information correlated with the source and the channel is noiseless, the exact RDP region is established. If only the decoder has correlated side information in the noiseless setting, an inner bound is derived along with a special case where the region is exact. Binary and Gaussian examples demonstrate that common randomness can significantly reduce the communication rate in secure RDP settings, unlike in standard rate-distortion settings. Thus, our results illustrate that random binning-based coding achieves strong secrecy, low distortion, and high perceptual quality simultaneously.