The National Institute of Standards and Technology has recently published the new post-quantum standards for key encapsulation mechanism (KEM) and digital signature. The new standard for key encapsulation is called ML-KEM, while ML-DSA, FN-DSA, and SLH-DSA are the new digital signaturesdigital signatures. Following the standardization process, it is important to analyze the weaknesses and potential information leakages of these new standards, so that they can be prevented. In this work, we study some of the most important side-channel attacks based on artificial intelligence (AI) models against the original proposals on which the standards are based: CRYSTALS-Kyber, CRYSTALS-Dilithium, FALCON, and SPHINCS + . Moreover, we examine the AI algorithms used in these attacks and their configurations, discussing which parameters and setting are suitable, and identifying different tools that might be useful.

Abstract In recent years, the rapid development of the intelligent connected vehicle industry has highlighted the critical importance of in-vehicle network security. The Controller Area Network(CAN) protocol is widely used as the core communication framework within in-vehicle systems. However, the CAN protocol lacks encryption and authentication mechanisms, making the network vulnerable to various security threats. Additionally, the rapid advancement of Vehicle-to-Everything communication has increased the frequency of interactions between vehicles and external environments, further exacerbating the security risks of the Internet of Vehicles (IoV) and highlighting the necessity of effective intrusion detection methods. Existing detection methods often face challenges like slow detection speeds and low accuracy. To overcome these limitations, this paper proposes a novel lightweight neural network model that integrates the Mamba architecture with Convolutional Neural Networks (CNNs) for IoV intrusion detection. In the in-vehicle scenario, a two-stage detection method combines single-message detection with message sequence detection. Single-message detection processes each message individually, enabling fine-grained intrusion detection. In contrast, message sequence detection leverages the periodic characteristics of CAN bus messages to detect anomalies in consecutive message sequences. In a complex external network environment, we employ a stacking ensemble approach and knowledge distillation to enhance and compress the model, resulting in a compact, high-performance, lightweight solution. Experimental results on four publicly available datasets show that our method has good detection performance and provides an effective solution to the security challenges in the IoV.

Abstract Extracting Chinese Cyber Threat Intelligence (CTI) under increasingly complex advanced persistent threat scenarios is crucial, yet challenging due to domain-specific term ambiguity and frequent long, nested entities. To address polysemy, nested-label conflicts, and cross-sentence semantic discontinuity, we propose an enhanced Transformer-based entity recognition method formulated as a pointer network. On the encoder side, we build a RoBERTa model with Rotary Positional Embeddings. To handle complex positions and boundaries of heterogeneous entity types, we introduce tokenization compensation and positional-parameter compression to sharpen boundary sensitivity. In the decoder, we refine GlobalPointer and model recognition as 2D head–tail span matching, enabling direct detection of overlapping and nested entities. To mitigate long-tail bias, we introduce an entity-frequency-aware dynamic threshold and a reweighted zero-boundary log-loss to improve recall for rare entities. Experiments demonstrate an overall F1 improvement of 6.32% over baselines on Chinese CTI datasets, with absolute gains reaching 19.7% specifically on nested and long entities. These results validate the model’s effectiveness in Chinese-specific named entity recognition and its utility for high-accuracy automated CTI analysis.

Abstract Accuracy remains the dominant evaluation metric in Intrusion Detection System (IDS) research, yet an IDS that detects attacks too late is functionally equivalent to one that fails—particularly in Internet of Things (IoT) environments. In operational settings, the timing of detection shapes both the scope of adversarial activity and the feasibility of effective response. To the best of our knowledge, latency (the speed at which intrusions are identified) has received no systematic attention. Our analysis of published IDS papers reveals that latency is defined inconsistently—often referring to inference, communication, computation time, or combinations of these-leading to incomparable performance claims. To close this gap, we formally define end-to-end latency as the sum of three measurable components: Attacker-Controlled Latency (ACL), IDS-Determined Latency (IDL), and Post-Detection Latency (PDL). The framework further decomposes IDL into environment-specific sub-components, linking network topology and detection methodology to overall responsiveness. The framework provides a systematic mechanism to parameterize and vary these components across deployment environments (edge, cloud, federated learning). Using a Cyber-Physical dataset, we demonstrate the applicability of our latency framework to an existing dataset. By providing a mechanism to quantify detection timeliness, the proposed framework enables analysts to estimate attacker dwell time, benchmark real-time responsiveness, and standardize latency reporting across IDS studies.

Abstract Source code vulnerability detection is critical to software security. Existing methods either rely on external tools to construct program graphs, which are constrained by environmental limitations, incur high computational costs, and are prone to failure, or adopt lightweight sequence models that overlook code structural features. To address this, we propose LiteVul, a lightweight framework that constructs compact directed graphs directly from raw source code without relying on external tools, capturing multi-scale lexical co-occurrence relationships and fine-grained structural dependencies. By leveraging structure-aware graph neural networks, LiteVul aggregates multi-layer features to achieve precise vulnerability prediction. Experiments conducted on the Devign, Reveal, and DiverseVul benchmark datasets show that LiteVul substantially surpasses nine representative baseline methods in performance. Additionally, its graph construction efficiency is substantially improved, providing a feasible technical pathway for vulnerability detection in large-scale codebases.

Abstract Quantum commitment schemes remain challenging to deploy in practice due to the high cost and fragility of quantum communication. Although recent advances have leveraged quantum channels to improve security and efficiency, these schemes often require substantial quantum interaction, limiting their practicality. To address these challenges, we follow the emerging paradigm of quantum-computation classical-communication (QCCC) protocols, which minimize quantum communication while preserving robust security guarantees. In this work, we construct a QCCC commitment scheme that achieves statistically hiding and computationally collapse-binding-a strong binding notion previously attainable only via collapsing hash functions, which are believed to be stronger than quantum collision-resistant hash functions. In order to circumvent the obstacle of quantum rewinding for an entirely quantum adversary, our construction introduces a novel cryptographic primitive, the quantum inaccessible entropy generator (qIEG), as a quantum analogue of the classical IEG framework developed by Haitner et al. [STOC ’09]. Notably, our approach relies on potentially weaker assumptions, marking a significant step toward practically deployable and theoretically robust quantum commitments in communication-constrained quantum settings.

Abstract Fully Homomorphic Encryption (FHE) enables arbitrary computations on encrypted data, a paradigm that Multi-Key FHE (MKFHE) extends to the decentralized setting by supporting operations on ciphertexts encrypted under multiple, distinct keys. However, the high computational cost of bootstrapping remains a major bottleneck, especially in the multi-key scenario where blind rotation is the dominant overhead. To address this, we propose a novel and parallel-friendly blind rotation scheme based on the NTRU assumption for efficient MKFHE bootstrapping. Our core technical contribution is a grouped inner product algorithm optimized for automorphism-based blind rotation, which reorganizes hybrid product storage and extends the external product to be compatible with both NTRU and MK-RLWE ciphertexts. Our parallelized algorithm reduces the time complexity from O ( n ) to $$O(\sqrt{n})$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>O</mml:mi> <mml:mo>(</mml:mo> <mml:msqrt> <mml:mi>n</mml:mi> </mml:msqrt> <mml:mo>)</mml:mo> </mml:mrow> </mml:math> . Our scheme demonstrates significant improvements over prior MKFHE works in both computational efficiency and storage requirements. At a 100-bit security level with $$k=8$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>k</mml:mi> <mml:mo>=</mml:mo> <mml:mn>8</mml:mn> </mml:mrow> </mml:math> participants, our scheme achieves a ciphertext bootstrapping time of 0.048 seconds, representing a $$6.8 \times$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mn>6.8</mml:mn> <mml:mo>×</mml:mo> </mml:mrow> </mml:math> speedup compared to Kwak et al.’s state-of-the-art work. Furthermore, our scheme substantially reduces storage overhead, requiring only 81.5MB for evaluation keys ( $$1.7 \times$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mn>1.7</mml:mn> <mml:mo>×</mml:mo> </mml:mrow> </mml:math> smaller) and 64KB for re-linearization keys ( $$6.0 \times$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mn>6.0</mml:mn> <mml:mo>×</mml:mo> </mml:mrow> </mml:math> smaller) relative to Kwak et al.’s implementation.

Abstract The proliferation of Cyber-Physical Systems (CPS) across critical infrastructure has created an unprecedented attack surface where digital threats may precipitate catastrophic physical consequences. As conventional centralized security paradigms fail to address the scale and complexity of these environments, Federated Learning (FL) has emerged as a transformative approach, enabling collaborative, edge-native anomaly detection without centralizing sensitive data. This paper presents a comprehensive survey and critical analysis of the state-of-the-art in securing CPS through advanced FL. We introduce a novel multi-axis taxonomy that systematically categorizes the field by architecture, detection methodology, application domain, and privacy-preservation scheme. Building on this analysis, we synthesize these findings into a prescriptive framework to guide the selection of appropriate security archetypes for different CPS domains. Through this lens, we deconstruct the—the trade-off between accuracy, communication, and privacy- that governs every FL design. Our analysis synthesizes the dominant trends, including the convergence of deep learning with edge computing and the increasing sophistication of privacy-enhancing technologies. We further identify critical research gaps, including the scarcity of physical testbeds, limited resilience against advanced adversarial attacks, and underdeveloped explainability. The paper concludes by defining the critical frontiers for future research, emphasizing the need to resolve the inherent tension between FL’s privacy goals and the transparency requirements of Explainable AI (XAI) to build truly trustworthy systems.

Abstract Program options provide flexible software functionality control but complicate fuzz testing, as triggering many behaviors require specific option combinations. Although existing option-aware fuzzing approaches attempt to mutate options as inputs or leverage AI technologies to extract option relationships from documentation, these methods have limitations. Documentation is often incomplete, and some option dependencies are embedded deeply within program logic via data or control flows, making these methods challenging to detect all possible dependencies. This paper introduces OSmartPro , an advanced option-fuzzing approach that directly extracts options and infers option dependencies from source code. Given LLM’s capabilities to interpret program semantics, OSmartPro employs LLM-assisted static analysis to handle diverse option-parsing structures and extract comprehensive options. Through control and data dependency analysis, it constructs option impact graph , which it uses to guide fuzzing strategies. The tool successfully extracted complete options from all 59 programs in our test set, uncovering undocumented options in over 66% of them. Additionally, OSmartPro inferred 14,701 option combinations, identified 45.03% more execution paths compared to AFL++, and uncovered 54 zero-day vulnerabilities, of which 18 awarded CVE IDs. Lastly, in a benchmark comparison against four option-aware fuzzers, OSmartPro achieved higher line coverage in 66.7% (20 out of 30) of the programs.