Tags → #ai security

New AI Model Released Amid Security Vulnerability Concerns

A new AI reasoning model developed by a Chinese lab has garnered attention for its advanced capabilities, while also raising concerns about potential security vulnerabilities such as prompt injection and Cross-Site Scripting (XSS).
Research Introduces Trapdoor Defense Against Model Inversion Attacks

A recent research paper discusses the challenges of Model Inversion attacks on Deep Neural Networks and introduces a new defense mechanism called Trapdoor-based Model Inversion Defense (Trap-MID) to enhance privacy protection.
Research Highlights Vulnerabilities in Large Language Models Security

A new paper examines the vulnerabilities of large language models in enterprise settings, emphasizing the need for a comprehensive understanding of security risks across software and hardware components in compound AI systems.
New Concept of Defendability Against Backdoors in Machine Learning

Researchers from the Alignment Research Center have proposed a new formal concept of defendability against backdoors in machine learning models, framing it as a strategic game between an attacker and a defender.
Jailbreaking Techniques and Security Risks in Large Language Models

The practice of jailbreaking Large Language Models (LLMs) involves manipulating AI systems to produce unintended outputs, raising concerns about potential security vulnerabilities and prompting research into enhancing their security.
Research Paper Proposes New Method to Combat Backdoor Attacks

A research paper from The Chinese University of Hong Kong, Shenzhen, introduces a novel method called Distance-Driven Detoxification (D3) to enhance defenses against backdoor attacks in machine learning models.
Study Identifies Vulnerabilities in Google’s Vertex AI Platform

A recent study by Palo Alto Networks identified two significant vulnerabilities in Google’s Vertex AI platform that could allow attackers to escalate privileges and exfiltrate sensitive machine learning models.
HackerOne Report Highlights AI's Impact on Cybersecurity

The 2024 Hacker-Powered Security Report by HackerOne highlights the growing intersection of artificial intelligence and cybersecurity, revealing concerns among security leaders about AI risks and the increasing focus on AI safety vulnerabilities.