Security researchers from Calif have successfully exploited a vulnerability in macOS on M5 silicon, achieving this in a matter of five days. This exploit leverages Anthropic’s Mythos Preview model, which allowed the team to combine two bugs with various techniques to compromise the Mac’s memory, accessing restricted areas of the device.
Apple's security efforts, particularly with its Memory Integrity Enforcement (MIE), have faced significant challenges. Introduced last year, MIE is designed to enhance memory safety using Arm’s Memory Tagging Extension (MTE), a specification aimed at preventing memory corruption exploits. While MIE aims to strengthen security, researchers argue that it has vulnerabilities that can be exploited.
The Calif team shared insights into their approach, including a brief video demonstrating the kernel memory corruption exploit. Although Apple has invested considerable resources into developing MIE, the researchers noted that they identified a path to breach this security within a remarkably short time frame.