FORCEDENTRY: Sandbox Escape

Google Project Zero analyzes FORCEDENTRY's sandbox escape: attackers embedded a serialized NSFunctionExpression inside a JBIG2-derived object to trigger deserialization and evaluation of malicious NSPredicate/NSExpression graphs. By abusing NSXPC type-flexibility, library-loading side effects (e.g., AVSpeechSynthesisVoice -> AXSpeechImplementation -> PrototypeTools) and initWithCoder behaviors (PTRow/PTSection), the exploit coerced CommCenter to load frameworks, fetch and decrypt a next-stage payload, and evaluate it — achieving post-sandbox arbitrary code execution without memory corruption and largely unaffected by pointer authentication or memory-tagging mitigations.