Malware-model reliability
98.0% overall accuracy hid an 85.2% false-negative rate for one malware family.
Aggregate calibration looked excellent. Family-level reliability told a very different story.
Hi, I'm Tara! I'm a CS student at Stanford interested in security, especially where the security we expect on paper breaks down in practice. Recently, I've been working on malware-model reliability, vulnerability intelligence, enterprise security, and security for autonomous AI systems.
Malware-model reliability
Aggregate calibration looked excellent. Family-level reliability told a very different story.
CVE enrichment
Rare classes and vulnerability details that had to be inferred remained much harder.
Government security guidance
Even close security allies differed substantially in what they recommended enterprises prioritize.