MIT's Fractal OS, a bare-metal research kernel developed at MIT CSAIL, has uncovered the first published evidence of Phantom speculation on Apple's M1 chip — a class of speculative execution attack previously observed only on AMD and Intel processors. The research also found that the M1's conditional branch predictor has no privilege isolation between user-space and kernel-space, directly contradicting earlier published findings that were likely skewed by macOS scheduler noise. While full arbitrary code execution was not demonstrated, observable instruction cache side-channel leakage was confirmed across user-to-user, user-to-kernel, and cross-ASID configurations. Apple's response has been atypical — requesting access to Fractal itself rather than issuing a CVE or advisory — leaving developers without official guidance. The findings challenge the long-held assumption that Apple Silicon is inherently safer from Spectre-class attacks due to its post-Spectre design.

5m read timeFrom latesthackingnews.com
Post cover image
Table of contents
Where the “Safer Apple Silicon” Idea Came FromWhat Apple M1 Phantom Speculation Actually ShowsThe Conditional Branch Predictor Finding Is More Immediately ConcreteApple’s Response Is StrangeWhat Still Holds, and What Doesn’t
319 Impressions