Protecting software through compiler plugins (GCC, LLVM) and Linux kernel customizations.
Static and dynamic analysis to detect vulnerabilities, buffer overflows, and software defects.
Leveraging deep learning and GNNs to facilitate security-oriented program analysis and malware detection.
Defending LLM agents against prompt injection attacks with polymorphic prompt protections.
Identifying critical data in binaries through execution trace analysis and deep neural models.
Understanding trade-offs between granularity, performance and timeliness in CFI mechanisms.