Cybersecurity researchers conducted a study to assess the effectiveness of large language models (LLMs) like ChatGPT in launching cyber attacks. They found that ChatGPT was able to exploit one-day vulnerabilities with an 87% success rate, outperforming other methods. However, the success rate dropped significantly when the CVE code was not provided. The researchers highlighted the potential for LLMs to become more powerful tools for cyber criminals in the future. They emphasized the need for the cybersecurity community to carefully consider the integration of LLM agents in defensive measures. The study concluded that while LLMs have the capability to exploit vulnerabilities autonomously, there is a need for caution and strategic deployment to mitigate potential risks.
Source link
Source link: https://securityintelligence.com/articles/chatgpt-4-exploits-87-percent-one-day-vulnerabilities/
ChatGPT 4 exploits 87% of one-day vulnerabilities #cybersecurity
