Personalizing Security Education: LLMs Inject Vulnerabilities Into Students' Own Code for Learning

arXiv·low signalXBlueskyLinkedInCopy link

Researchers use LLMs to automatically inject realistic security vulnerabilities into students' own code, grounding secure programming education in constructivist theory — students learn security more effectively when examples are drawn from their personal projects rather than generic templates. The approach generates personalized vulnerability exercises at scale, addressing a persistent pedagogical gap in cybersecurity education. Potentially useful beyond education — the same technique could generate security-focused test cases for production code review training.