A dual-stream CodeBERT architecture is presented that addresses keyword bias —by combining pre-trained Transformer representations with a 50-dimensional hand-engineered security feature vector, supported by targeted data augmentation and two-stage adversarial fine-tuning.
The increasing deployment of Large Language Models (LLMs) in critical infrastructure has introduced a class of security risks that remain insufficiently characterised and poorly tooled in practice. Attack vectors including prompt injection, jailbreaking, code execution facilitation, covert data exfiltration, and traini...
M. Yamin· International Conference on...· 0 citations
This work presents GraftyVul, a system that constructs vulnerable programs by grafting real-world vulnerabilities into open-source projects, and introduces a language- and context-agnostic semantic embedding that compares vulnerabilities by sink, mechanism and host-feature rather than surface code.
Omri Ram, Mitchell Horner, R. van der Meyden et al.· 0 citations
The proposed machine learning-based system for detecting and mitigating SQL injection attacks is designed not just to protect against conventional SQLi attacks but also against JSON-based SQLi attacks, NoSQL injection attacks, hybrid attacks, and conventional WAF evasion techniques.
Nisha P. Shetty, Vinayak Kothari, Eva Hemantkumar Shah et al.· Journal of Mobile Multimedia· 0 citations
Findings show that insecure code generation is not merely a collection of independent defects, but a structured and prompt-conditioned phenomenon, motivating cluster-aware verification and prompt-level intervention for safer LLM-assisted programming.
Extensive evaluations across three representative vulnerability cases and ten backdoor attacks, along with sixteen competitive baselines, demonstrate that CodeTracer consistently achieves high forensic accuracy, low false identification rates, and strong robustness against adaptive attacks.
Anjun Gao, Yueyang Quan, Zhuqing Liu et al.· arXiv.org· 2 citations
Experimental results demonstrate the superiority of HFE-DDL over the state-of-the-art method in vulnerability detection tasks and show that HFE-DDL has potential as a tool for proactively identifying vulnerabilities during development, aiming to produce more secure Android applications.
Ahmed Abdu, Hakim A. Abdo, Mohsen Ali Alawami et al.· Cluster Computing· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
