CyberSec Research

The use of ML in cybersecurity has long been impaired by generalization issues: Models that work well in controlled scenarios fail to maintain performance in production. The root cause often lies in ML algorithms learning superficial patterns (shortcuts) rather than underlying cybersecurity concepts. We investigate contrastive multi-modal learning as a first step towards improving ML performance in cybersecurity tasks. We aim at transferring knowledge from data-rich modalities, such as text, to data-scarce modalities, such as payloads. We set up a case study on threat classification and propose a two-stage multi-modal contrastive learning framework that uses textual vulnerability descriptions to guide payload classification. First, we construct a semantically meaningful embedding space using contrastive learning on descriptions. Then, we align payloads to this space, transferring knowledge from text to payloads. We evaluate the approach on a large-scale private dataset and a synthetic benchmark built from public CVE descriptions and LLM-generated payloads. The methodology appears to reduce shortcut learning over baselines on both benchmarks. We release our synthetic benchmark and source code as open source.

Hamming Quasi-Cyclic (HQC) was chosen for the latest post-quantum cryptography standardization. A concatenated Reed-Muller (RM) and Reed-Solomon (RS) code is decoded during the HQC decryption. Soft-decision RS decoders achieve better error-correcting performance than hard-decision decoders and accordingly shorten the required codeword and key lengths. However, the only soft-decision decoder for HQC in prior works is an erasure-only decoder, which has limited coding gain. This paper analyzes other hardware-friendly soft-decision RS decoders and discovers that the generalized minimum-distance (GMD) decoder can better utilize the soft information available in HQC. Extending the Agrawal-Vardy bound for the scenario of HQC, it was found that the RS codeword length for HQC-128 can be reduced from 46 to 36. This paper also proposes efficient GMD decoder hardware architectures optimized for the short and low-rate RS codes used in HQC. The HQC-128 decryption utilizing the proposed GMD decoder achieves 20% and 15% reductions on the latency and area, respectively, compared to the decryption with hard-decision decoders.

Getting a real cybersecurity risk assessment for a small organization is expensive -- a NIST CSF-aligned engagement runs $15,000 on the low end, takes weeks, and depends on practitioners who are genuinely scarce. Most small companies skip it entirely. We built a six-agent AI system where each agent handles one analytical stage: profiling the organization, mapping assets, analyzing threats, evaluating controls, scoring risks, and generating recommendations. Agents share a persistent context that grows as the assessment proceeds, so later agents build on what earlier ones concluded -- the mechanism that distinguishes this from standard sequential agent pipelines. We tested it on a 15-person HIPAA-covered healthcare company and compared outputs to independent assessments by three CISSP practitioners -- the system agreed with them 85% of the time on severity classifications, covered 92% of identified risks, and finished in under 15 minutes. We then ran 30 repeated single-agent assessments across five synthetic but sector-realistic organizational profiles in healthcare, fintech, manufacturing, retail, and SaaS, comparing a general-purpose Mistral-7B against a domain fine-tuned model. Both completed every run. The fine-tuned model flagged threats the baseline could not see at all: PHI exposure in healthcare, OT/IIoT vulnerabilities in manufacturing, platform-specific risks in retail. The full multi-agent pipeline, however, failed every one of 30 attempts on a Tesla T4 with its 4,096-token default context window -- context capacity, not model quality, turned out to be the binding constraint.

Large Language Models (LLMs) have been widely deployed, especially through free Web-based applications that expose them to diverse user-generated inputs, including those from long-tail distributions such as low-resource languages and encrypted private data. This open-ended exposure increases the risk of jailbreak attacks that undermine model safety alignment. While recent studies have shown that leveraging long-tail distributions can facilitate such jailbreaks, existing approaches largely rely on handcrafted rules, limiting the systematic evaluation of these security and privacy vulnerabilities. In this work, we present EvoJail, an automated framework for discovering long-tail distribution attacks via multi-objective evolutionary search. EvoJail formulates long-tail attack prompt generation as a multi-objective optimization problem that jointly maximizes attack effectiveness and minimizes output perplexity, and introduces a semantic-algorithmic solution representation to capture both high-level semantic intent and low-level structural transformations of encryption-decryption logic. Building upon this representation, EvoJail integrates LLM-assisted operators into a multi-objective evolutionary framework, enabling adaptive and semantically informed mutation and crossover for efficiently exploring a highly structured and open-ended search space. Extensive experiments demonstrate that EvoJail consistently discovers diverse and effective long-tail jailbreak strategies, achieving competitive performance with existing methods in both individual and ensemble level.

Forecasting plays a crucial role in modern safety-critical applications, such as space operations. However, the increasing use of deep forecasting models introduces a new security risk of trojan horse attacks, carried out by hiding a backdoor in the training data or directly in the model weights. Once implanted, the backdoor is activated by a specific trigger pattern at test time, causing the model to produce manipulated predictions. We focus on this issue in our \textit{Trojan Horse Hunt} data science competition, where more than 200 teams faced the task of identifying triggers hidden in deep forecasting models for spacecraft telemetry. We describe the novel task formulation, benchmark set, evaluation protocol, and best solutions from the competition. We further summarize key insights and research directions for effective identification of triggers in time series forecasting models. All materials are publicly available on the official competition webpage https://www.kaggle.com/competitions/trojan-horse-hunt-in-space.

In traditional runtime verification, a system is typically observed by a monolithic monitor. Enforcing privacy in such settings is computationally expensive, as it necessitates heavy cryptographic primitives. Therefore, privacy-preserving monitoring remains impractical for real-time applications. In this work, we address this scalability challenge by distributing the monitor across multiple parties -- at least one of which is honest. This architecture enables the use of efficient secret-sharing schemes instead of computationally intensive cryptography, dramatically reducing over-head while maintaining strong privacy guarantees. While existing secret-sharing approaches are typically limited to one-shot executions which do not maintain an internal state, we introduce a protocol tailored for continuous monitoring that supports repeated evaluations over an evolving internal state (kept secret from the system and the monitoring entities). We implement our approach using the MP-SPDZ framework. Our experiments demonstrate that, under these architectural assumptions, our protocol is significantly more scalable than existing alternatives.

Autonomous coding agents are increasingly integrated into software development workflows, offering capabilities that extend beyond code suggestion to active system interaction and environment management. OpenClaw, a representative platform in this emerging paradigm, introduces an extensible skill ecosystem that allows third-party developers to inject behavioral guidance through lifecycle hooks during agent initialization. While this design enhances automation and customization, it also opens a novel and unexplored attack surface. In this paper, we identify and systematically characterize guidance injection, a stealthy attack vector that embeds adversarial operational narratives into bootstrap guidance files. Unlike traditional prompt injection, which relies on explicit malicious instructions, guidance injection manipulates the agent's reasoning context by framing harmful actions as routine best practices. These narratives are automatically incorporated into the agent's interpretive framework and influence future task execution without raising suspicion.We construct 26 malicious skills spanning 13 attack categories including credential exfiltration, workspace destruction, privilege escalation, and persistent backdoor installation. We evaluate them using ORE-Bench, a realistic developer workspace benchmark we developed. Across 52 natural user prompts and six state-of-the-art LLM backends, our attacks achieve success rates from 16.0% to 64.2%, with the majority of malicious actions executed autonomously without user confirmation. Furthermore, 94% of our malicious skills evade detection by existing static and LLM-based scanners. Our findings reveal fundamental tensions in the design of autonomous agent ecosystems and underscore the urgent need for defenses based on capability isolation, runtime policy enforcement, and transparent guidance provenance.

Wireless networks are highly vulnerable to spoofing attacks, especially when attackers transmit consecutive spoofing packets. Conventional physical layer authentication (PLA) methods have mostly focused on single-packet spoofing attack. However, under consecutive spoofing attacks, they become ineffective due to channel evolution caused by device mobility and channel fading. To address this challenge, we propose a channel prediction-based PLA framework. Specifically, a Transformer-based channel prediction module is employed to predict legitimate CSI measurements during spoofing interval, and the input of channel prediction module is adaptively updated with predicted or observed CSI measurements based on the authentication decision to ensure robustness against sustained spoofing. Simulation results under Rayleigh fading channels demonstrate that the proposed approach achieves low prediction error and significantly higher authentication accuracy than conventional benchmark, maintaining robustness even under extended spoofing attacks.

Privacy-preserving aggregation is a cornerstone for AI systems that learn from distributed data without exposing individual records, especially in federated learning and telemetry. Existing two-server protocols (e.g., Prio and successors) set a practical baseline by validating inputs while preventing any single party from learning users' values, but they impose symmetric costs on both servers and communication that scales with the per-client input dimension $L$. Modern learning tasks routinely involve dimensionalities $L$ in the tens to hundreds of millions of model parameters. We present TAPAS, a two-server asymmetric private aggregation scheme that addresses these limitations along four dimensions: (i) no trusted setup or preprocessing, (ii) server-side communication that is independent of $L$ (iii) post-quantum security based solely on standard lattice assumptions (LWE, SIS), and (iv) stronger robustness with identifiable abort and full malicious security for the servers. A key design choice is intentional asymmetry: one server bears the $O(L)$ aggregation and verification work, while the other operates as a lightweight facilitator with computation independent of $L$. This reduces total cost, enables the secondary server to run on commodity hardware, and strengthens the non-collusion assumption of the servers. One of our main contributions is a suite of new and efficient lattice-based zero-knowledge proofs; to our knowledge, we are the first to establish privacy and correctness with identifiable abort in the two-server setting.

Formal patterns are formally specified solutions to frequently occurring distributed system problems that are generic, executable, and come with strong qualitative and/or quantitative formal guarantees. A formal pattern is a generic system transformation which transforms a usually infinite class of systems in need of the pattern's solution into enhanced versions of such systems that solve the problem in question. In this paper we demonstrate the application of formal patterns to protocol dialects. Dialects are methods for hardening protocols so as to endow them with light-weight security, especially against easy attacks that can lead to more serious ones. A lingo is a dialect's key security component, because attackers are unable to ''speak'' the lingo. A lingo's ''talk'' changes all the time, becoming a moving target for attackers. In this paper we present several formal patterns for both lingos and dialects. Lingo formal patterns can make lingos stronger by both transforming them and by composing several lingos into a stronger lingo. Dialects themselves can be obtained by the application of a single dialect formal pattern, generic on both the chosen lingo and the chosen protocol.

Penetration testing, the practice of simulating cyberattacks to identify vulnerabilities, is a complex sequential decision-making task that is inherently partially observable and features large action spaces. Training reinforcement learning (RL) policies for this domain faces a fundamental bottleneck: existing simulators are too slow to train on realistic network scenarios at scale, resulting in policies that fail to generalize. We present NASimJax, a complete JAX-based reimplementation of the Network Attack Simulator (NASim), achieving up to 100x higher environment throughput than the original simulator. By running the entire training pipeline on hardware accelerators, NASimJax enables experimentation on larger networks under fixed compute budgets that were previously infeasible. We formulate automated penetration testing as a Contextual POMDP and introduce a network generation pipeline that produces structurally diverse and guaranteed-solvable scenarios. Together, these provide a principled basis for studying zero-shot policy generalization. We use the framework to investigate action-space scaling and generalization across networks of up to 40 hosts. We find that Prioritized Level Replay better handles dense training distributions than Domain Randomization, particularly at larger scales, and that training on sparser topologies yields an implicit curriculum that improves out-of-distribution generalization, even on topologies denser than those seen during training. To handle linearly growing action spaces, we propose a two-stage action decomposition (2SAS) that substantially outperforms flat action masking at scale. Finally, we identify a failure mode arising from the interaction between Prioritized Level Replay's episode-reset behaviour and 2SAS's credit assignment structure. NASimJax thus provides a fast, flexible, and realistic platform for advancing RL-based penetration testing.

Devices employing cryptographic approaches have to be resistant to physical attacks. Side-Channel Analysis (SCA) and Fault Injection (FI) attacks are frequently used to reveal cryptographic keys. In this paper, we present a combined SCA and laser illumination attack against an Elliptic Curve Scalar Multiplication accelerator using a differential probe from Teledyne LeCroy. Our experiments show that laser illumination increases the power consumption of the chip, especially its static power consumption but the success of the horizontal power analysis attacks was changed insignificantly. We assume that using a laser with a high laser beam power and concentrating on measuring and analysing only static current can improve the attack success significantly. The horizontal attacks against public key cryptosystems exploiting the Static Consumption under Laser Illumination (SCuLI attacks) are novel and their potential is not investigated yet. These attacks can be especially dangerous against cryptographic chips manufactured in scaled technologies. If such attacks are feasible, appropriate countermeasures have to be proposed in the future.

The ability to simulate human privacy decisions has significant implications for aligning autonomous agents with individual intent and conducting cost-effective, large-scale privacy-centric user studies. Prior approaches prompt Large Language Models (LLMs) with natural language user statements, data-sharing histories, or demographic attributes to simulate privacy decisions. These approaches, however, fail to balance individual-level accuracy, prompt usability, token efficiency, and population-level representation. We present Narriva, an approach that generates text-based synthetic privacy personas to address these shortcomings. Narriva grounds persona generation in prior user privacy decisions, such as those from large-scale survey datasets, rather than purely relying on demographic stereotypes. It compresses this data into concise, human-readable summaries structured by established privacy theories. Through benchmarking across five diverse datasets, we analyze the characteristics of Narriva's synthetic personas in modeling both individual and population-level privacy preferences. We find that grounding personas in past privacy behaviors achieves up to 88% predictive accuracy (significantly outperforming a non-personalized LLM baseline), and yields an 80-95% reduction in prompt tokens compared to in-context learning with raw examples. Finally, we demonstrate that personas synthesized from a single survey can reproduce the aggregate privacy behaviors and statistical distributions (TVComplement up to 0.85) of entirely different studies.

Serverless computing abstracts infrastructure management but also obscures system-level behaviors that can introduce security risks. Prior work has shown that serverless platforms are vulnerable to attacks exploiting shared execution environments, including attacker--victim co-location and denial-of-service through resource contention, yet analyzing these risks on production platforms is difficult due to limited observability, high cost, and lack of experimental control, while existing simulators primarily focus on performance and cost rather than security. We present Kumo, a security-focused simulator for serverless platforms that enables controlled, reproducible analysis of security risks arising from scheduling and resource sharing decisions. Kumo models invocation arrivals, scheduler placement, container reuse, resource contention, and queuing within a discrete-event framework, explicitly representing attackers and victims as first-class entities and providing metrics such as co-location probability, time to first co-location, invocation drop rate, and tail latency. Through two case studies, we show that scheduler choice is a first-order factor for co-location attacks, inducing orders-of-magnitude differences under identical workloads, while Denial-of-Service behavior is largely governed by system-level factors such as service time, queuing policy, and cluster capacity once contention dominates. These results highlight the need to distinguish scheduler-driven isolation risks from broader resource exhaustion vulnerabilities and position Kumo as a flexible foundation for systematic, security-aware exploration of serverless platforms.

At SAC 2013, Berger et al. first proposed the Extended Generalized Feistel Networks (EGFN) structure for the design of block ciphers with efficient diffusion. Later, based on the Type-2 EGFN, they instantiated a new lightweight block cipher named Lilliput (published in IEEE Transactions on Computers, Vol. 65, Issue 7, 2016). According to published cryptanalysis results, Lilliput is sufficiently secure against theoretical attacks such as differential, linear, boomerang, and integral attacks, which rely on the statistical properties of plaintext and ciphertext. However, there is a lack of analysis regarding its resistance to physical attacks in real-world scenarios, such as fault attacks. In this paper, we present the first systematic differential fault analysis (DFA) of Lilliput under three nibble-oriented fault models with progressively relaxed adversarial assumptions to comprehensively assess its fault resilience. In Model I (multi-round fixed-location), precise fault injections at specific rounds recover the master key with a 98% success rate using only 8 faults. Model II (single-round fixed-location) relaxes the multi-round requirement, demonstrating that 8 faults confined to a single round are still sufficient to achieve a 99% success rate by exploiting Lilliput's diffusion properties and DDT-based constraints. Model III (single-round random-location) further weakens the assumption by allowing faults to occur randomly among the eight rightmost branches of round 27. By uniquely identifying the fault location from ciphertext differences with high probability, the attack remains highly feasible, achieving over 99% success with 33 faults and exceeding 99.5% with 36 faults. Our findings reveal a significant vulnerability of Lilliput to practical fault attacks across different adversary capabilities in real-world scenarios, providing crucial insights for its secure implementation.

As the Internet of Things (IoT) becomes an integral part of critical infrastructure, smart cities, and consumer networks, there has been an increase in the number of software attacks on the microcontrollers (MCUs) that constitute such networks. Runtime firmware attestation, i.e., the verification of a firmware's integrity, has become instrumental, and prior work focuses on lightweight IoT MCUs, offloading the verification task to capable remote verifiers. However, modern IoT devices feature large flash and volatile memory, on-device TinyML inference, and Trusted Execution Environments (TEE). Leveraging these capabilities, this paper presents a verifier-less, hybrid Self-Attestation (SA) framework called LiteAtt, which is based on TinyML execution in the Arm TrustZone of an IoT MCU for quick, on-device evaluation of the IoT firmware's SRAM footprint. LiteAtt takes a step towards ubiquitous intelligence and decentralized trust in IoT networks. It eliminates the need for firmware copies for attestation, and protects the privacy of user SRAM data by leveraging twin devices to train the TinyML models. The proposed framework achieves an average accuracy of 98.7%, F1 score of 99.33%, TPR of 98.72%, and TNR of 97.45% on SRAM attestation datasets collected from real devices. LiteAtt operates with a latency of 1.29ms, an energy consumption of 42.79uJ, and a runtime memory overhead of up to 32KB, which is suitable for battery-operated Arm Cortex-M devices. A security analysis is provided for the protocol regarding mutual authentication, confidentiality, integrity, SRAM privacy, and defense against replay and impersonation attacks. Practical deployment scenarios and future works are also discussed.

Graph pattern counting serves as a cornerstone of network analysis with extensive real-world applications. Its integration with local differential privacy (LDP) has gained growing attention for protecting sensitive graph information in decentralized settings. However, existing LDP frameworks are largely ad hoc, offering solutions only for specific patterns such as triangles and stars. A general mechanism for counting arbitrary graph patterns, even for the subclass of acyclic patterns, has remained an open problem. To fill this gap, we present the first general solution for counting arbitrary acyclic patterns under LDP. We identify and tackle two fundamental challenges: generalizing pattern construction from distributed data and eliminating node duplication during the construction. To address the first challenge, we propose an LDP-tailored recursive subpattern counting framework that incrementally builds patterns across multiple communication rounds. For the second challenge, we apply a random marking technique that restricts each node to a unique position in the pattern during computation. Our mechanism achieves strong utility guarantees: for any acyclic graph pattern with $k$ edges, we achieve an additive error of $\tilde{O}(\sqrt{N}d(G)^k)$, where $N$ is the number of nodes and $d(G)$ is the maximum degree of the input graph $G$. Experiments on real-world graph datasets across multiple types of acyclic patterns demonstrate that our mechanisms achieve up to $46$-$2600\times$ improvement in utility and $300$-$650\times$ reduction in communication cost compared to the baseline methods.

Advanced Persistent Threats (APTs) remain difficult to detect due to their stealthy nature and long-term persistence. To tackle this challenge, provenance-based threat hunting has gained traction as a proactive defense mechanism. This technique models audit logs as a whole-system provenance graph and searches for subgraphs that match APT patterns recorded in Cyber Threat Intelligence (CTI) reports. However, several limitations persist: 1) significant memory and time overhead due to the extremely large provenance graphs; 2) imprecise segmentation of APT activities from provenance graphs due to their intricate entanglement with benign operations; and 3) poor alignment of attack representations between CTI-derived query graphs and provenance graphs due to their substantial semantic gaps. To address these limitations, this paper presents ProHunter, an efficient and accurate provenance-based APT hunting system with a platform-independent design. To minimize system overhead, ProHunter creates a compact data structure that efficiently stores long-term provenance graphs using semantic abstraction and bit-level hierarchical encoding strategies. To segment APT behaviors, a heuristic-driven threat graph sampling algorithm is designed, which can extract precise attack patterns from provenance graphs. Furthermore, to bridge the semantic gaps between CTI-derived graphs and provenance graphs, ProHunter proposes adaptive graph representation and feature enhancement methods, enabling the extraction of consistent attack semantics at both localized and globalized levels.Extensive evaluations on real-world APT campaigns from DARPA TC E3, E5 and OpTC datasets demonstrate that ProHunter outperforms state-of-the-art threat hunting systems in terms of efficiency and accuracy. Our code is available at https://github.com/xueboQiu/ProHunter.