Tags → #large language models

Research Highlights Risks in General-Purpose AI Deployment

Research by Mario Fritz and his team explores the significance of Data-Instruction Separation (DIS) in enhancing the security of AI-driven systems against evolving cybersecurity threats.
New Tool Enhances Grading in Cybersecurity Education

A recent study introduces OpenCompass, a tool aimed at improving the grading of student responses in cybersecurity education by evaluating completeness, accuracy, and relevance.
Security Vulnerabilities in LLM-based Code Completion Tools

Cybersecurity experts are increasingly scrutinizing the vulnerabilities of Large Language Model-based Code Completion Tools, highlighting significant security risks that could affect user privacy and software integrity.
Advancements in Privacy-Preserving Synthetic Data Generation

Recent research has made significant strides in synthetic data generation by integrating Large Language Models and Differential Privacy to enhance data privacy while maintaining utility for machine learning applications.
Proactive Defense Strategies for Cloud Cybersecurity Explored

Recent research highlights innovative strategies for enhancing cybersecurity in Kubernetes environments, focusing on mutation-enabled proactive defense mechanisms to counter sophisticated attacks.
Vulnerabilities in Large Language Models: Data Poisoning Threats

Recent research highlights vulnerabilities in large language models, particularly concerning data poisoning and jailbreak-tuning, emphasizing the need for improved data integrity and security measures.
SecBench: Benchmarking Dataset for LLMs in Cybersecurity

Automated threat detection, powered by artificial intelligence, is emerging as a crucial defense mechanism against evolving cyber threats, while also highlighting the importance of integrating human expertise in cybersecurity practices.
CS-Eval Benchmark Introduced for Cybersecurity Language Models

The CS-Eval benchmark aims to enhance the evaluation of large language models in cybersecurity, focusing on improving their capabilities in threat detection and response.
Large Language Models in Software Supply Chain Security

Recent research highlights the potential of Large Language Models to improve software supply chain security while also identifying challenges that need to be addressed for effective implementation.
Large Language Models Enhance Cybersecurity Incident Response

Recent research highlights the potential of large language models to improve collaboration and efficiency in incident response scenarios within cybersecurity.