Google’s AI-powered bug bounty tool finds many troubling open-source security flaws

Google’s OSS-Fuzz has discovered more than two dozen vulnerabilities in various open source projects
Among them is a vulnerability in OpenSSL that can lead to RCE.
Google sees this as a major milestone in automated error detection

Google found 26 vulnerabilities in various open source code repositories, including a moderate-severity bug in the “critical OpenSSL library that underlies much of the Internet’s infrastructure.”

This wouldn’t be such news (Google has helped find thousands of bugs over the years) if the method by which the flaws were discovered wasn’t “artificial,” since the bugs were discovered using its AI-powered fuzzing tool. OSS-Fuzz.

“These specific vulnerabilities represent a major milestone in automated vulnerability hunting: each was discovered by AI using AI-generated and advanced fuzzy targets,” Google explained in a blog post.

Major improvements to the LLM program

Among these 26 flaws is an OpenSSL bug tracked as CVE-2024-9143. It has a severity of 4.3 and is described as an out-of-bounds write error that can crash the application or allow attackers to enable remote code execution (RCE). malware attacks. OpenSSL has since been updated to versions 3.3.3, 3.2.4, 3.1.8, 3.0.16, 1.1.1zb and 1.0.2zl to fix this bug.

To make things even more interesting, Google said that the vulnerability had likely existed for two decades “and would not have been detected by existing fuzzy targets written by humans.”

The discovery of the bug resulted from two major improvements, the company explained. The first is the ability to automatically generate more relevant context in tooltips, making LLM “less likely.” hallucinate missing details in his answer.” The second relates to LLM’s ability to simulate the entire workflow of a typical developer, including writing, testing, and iterating on a fuzz target, as well as triaging detected failures.

“This makes it possible to automate more parts of the fuzzing workflow. This additional iterative feedback, in turn, also resulted in improved quality and an increase in the number of correct fuzz targets.”

By using Hacker news