Arnab DevResearch
I research software security with a focus on fuzz testing, static analysis, and variability-aware testing. My work aims to improve software reliability and security through automated testing techniques.
SentState: Incremental Event-Driven Tracking of Financial Market Sentiment with Small Language Models
In this work, we address the gap between memoryless news classifiers and slow agent-memory systems by building a streaming architecture that tracks financial-market sentiment as a compact, continuously-updated per-asset state. We combine a small language model with a marked Hawkes process to update this state in constant time per headline, and evaluate it with a confound-aware incremental-regression protocol that isolates what the incremental memory genuinely adds.
FuzzBench++: Configuration-Aware Fuzzing Benchmarks
In this work, we identify critical gaps in current benchmarking infrastructure that prevent fair and systematic comparison of configuration-aware fuzzers. We extend Google’s FuzzBench to support configuration-aware fuzzing, using Docker containers to orchestrate scalable experiments.
Return to Home.