A brief introduction to Agentic Reinforcement Learning. And some of my thoughts.

An inspection of SMT solvers, typically exploring their capabilities and limitations in various domains such as program verification, symbolic execution, and optimization.

Analyzing and deobfuscating Polaris Obfuscator's BogusControlFlow pass, which inserts cloned basic blocks with opaque predicates and junk instructions.

Analyzing and deobfuscating Polaris Obfuscator's IndirectCall pass, which converts direct calls into indirect calls through computed addresses.

Ad-hoc analysis of binary obfuscation techniques using Polaris Obfuscator, examining effectiveness against reverse engineering and interaction with compiler optimizations.

Course notes on behavioral cloning and interactive imitation learning, covering loss functions, distribution shift (DAgger), and key theoretical foundations.

Security assessment of the University of Utah CHPC, covering software package management, patching policies, and the Protected Environment.

Reading notes on FIBER (USENIX Security '18), a source-to-binary patch presence test for identifying missing security patches in firmware.

Reading notes on PS^3 (ICSE '24), a binary-level patch presence test using semantic symbolic signatures and theorem proving for cross-compiler verification.