"Gang of AI Bots Can Hack Websites With a 53% Success Rate, Researchers Show"

Researchers from the University of Illinois Urbana-Champaign found that GPT-4-powered teams of bots can scan websites for zero-day vulnerabilities and attack them with 53 percent success. According to the researchers, Large Language Models (LLMs) can collaborate and work more effectively than a single instance of chatbot in exploiting real-world vulnerabilities. Their paper shows up to 4.5 times improvement over a standalone Artificial Intelligence (AI) agent. They created "the first multi-agent system to successfully accomplish meaningful cybersecurity exploits." The framework called "Hierarchical Planning and Task-Specific Agents" (HPTSA) involves a planner bot, a manager bot, and task-specific AI-powered agents. This article continues to discuss the study "Teams of LLM Agents Can Exploit Zero-Day Vulnerabilities."

Cybernews reports "Gang of AI Bots Can Hack Websites With a 53% Success Rate, Researchers Show"

Submitted by grigby1