Research
Academic papers, new techniques, benchmarks, and theoretical findings in AI/LLM security.
Academic papers, new techniques, benchmarks, and theoretical findings in AI/LLM security.
2,033 items
The paper proposes SCZ-HA, a seamless cross-zone authentication scheme for vehicle-to-infrastructure (V2I) communication. It combines long-term and short-term credentials, distributing encapsulated short-term credentials to neighboring zones with Shamir's secret sharing so vehicles present only encapsulations during handovers. Experiments report an average 70.54% reduction in authentication latency compared with state-of-the-art schemes.
MalPurifier is a new adversarial purification framework for Android malware detection, designed to defend ML-based detectors against evasion attacks. It combines a diversified adversarial perturbation mechanism, a protective noise injection strategy, and a Denoising AutoEncoder with a dual-objective loss. Tested on two large-scale datasets, it defended against 37 perturbation-based evasion attacks, keeping robust accuracy above 90.91%.
Researchers propose an image sharing method based on bilateral compressive sensing that provides multilevel privacy preservation. Data owners compress original images with designed measurement matrices using the proposed T ℓ1-B2DLDA algorithm, producing dimension-reduced data that supports classification for level I data users and reconstruction for level II data users. Level I users can complete their tasks without obtaining private or confidential information from the original images. The paper reports experiments on feasibility, performance and robustness, plus theoretical and practical security analysis, and releases its source code publicly.
The source is a bibliographic entry for a paper titled "DV2PDA: Decentralized and verifiable privacy-preserving data aggregation scheme for IIoT," published July 2026 in the Journal of Information Security and Applications, Volume 100, by Juan Wu, Lunzhi Deng, Yan Gao, Na Wang, Yanli Chen and Siwei Li. The provided text contains no abstract, method details or findings.
Li, Wang and Guo published a novel image steganography method for copyright marking in the Journal of Information Security and Applications, Volume 99, in June 2026. The source text gives only the title, date, journal and authors, so the method, its findings and any numbers are not described here.
Researchers Do Thi Thu Hien, Bao Pham-Thai, Nguyen Tan Cam and Van-Hau Pham published a multimodal approach for detecting Windows malware using comprehensive analysis of called APIs in the Journal of Information Security and Applications, Volume 99, in June 2026. The source text provided contains only publication metadata, so the method and findings cannot be described from it.
The source is bibliographic metadata only: a June 2026 publication in the Journal of Information Security and Applications, Volume 99, by Akhilesh Sharma and Preeti Chandrakar. It gives the title, which names the LECASC framework for secure and anonymous e-bidding using lightweight signcryption on blockchain. No method details, results, or figures are included.
This study applies the information-theoretic Privacy Funnel model to face recognition, using an end-to-end trainable framework for privacy-preserving representation learning. It introduces the Generative Privacy Funnel (GenPF) and the deep variational PF (DVPF) model, and reports that the framework yields a controllable privacy-utility trade-off while substantially reducing leakage about sensitive attributes on modern face recognition systems. A PyTorch implementation is released.
The paper compares adaptive active-defense methods for hardening ML-based network intrusion detection systems against reinforcement-learning-driven adversaries with static defenses. It was published in the Journal of Information Security and Applications, Volume 100, in July 2026, by Iacovos Ioannou, Christophoros Christophorou, Andreas Andreou, Marios Raspopoulos, Constandinos Mavromoustakis, Vasos Vassiliou and Fabrizio Granelli.
The paper QAEAS: A quantum adaptive ensemble attack system against robust deep neural networks was published in July 2026 in the Journal of Information Security and Applications, Volume 100. It is authored by Ali Mohammadi Ruzbahani, Abbas Yazdinejad and Hadis Karimipour. The source text provided contains only publication metadata and no abstract, method description, or findings.
Yao, Liu, Zhao, Deng, Guo and Ding publish a comprehensive survey of deep video inpainting and video inpainting detection in the Journal of Information Security and Applications, volume 99, in June 2026. The source text provides only the publication metadata, so no method, findings or numbers are described.