Current Projects

🔹 ChainSentinel: AI-Powered Multi-Agent Framework for Smart Contract Security

Objective: Design a domain-specific, multi-agent framework that integrates large language models (LLMs) with static and dynamic analysis tools (e.g., Slither, Mythril) to enhance security auditing, reduce false positives, and improve explainability in smart contracts.

  • Implements Auditor–Critic–Validator agent collaboration for iterative and explainable vulnerability detection.
  • Uses annotated datasets from SmartBugs Curated and NotSoSmartContracts for fine-tuning and benchmarking.
  • Combines formal verification with natural-language reasoning to generate human-readable security reports.

🔹 Benchmarking Study: Static Analysis vs. LLM-Based Auditing Tools

Conducting an extensive comparative study evaluating traditional smart contract security tools against emerging LLM-based systems. The goal is to understand capability gaps, false-positive trends, explainability quality, and adaptability to evolving Ethereum Improvement Proposals (EIPs).

  • Benchmarks tools including Slither, Mythril, FTSmartAudit, and Smart-LLaMA-DPO using standardized evaluation pipelines.
  • Computes precision, recall, and F1 scores with fuzzy line matching for cross-tool output normalization.
  • Forms the foundation for an ICSE 2026 submission titled “Auditing Smart Contracts with Language Models: Benchmarking Domain-Fine-Tuned LLMs vs. Static Analysis Tools.”

🔹 Capabilities of Generative AI for Smart Contract Development

Empirical multi-phase study investigating how generative AI tools such as ChatGPT, ChainGPT, and Gemini influence blockchain development. The research focuses on performance, security, and developer trust in AI-generated Solidity code.

  • Surveyed 114 developers to capture perceptions, usage patterns, and trust dynamics.
  • Evaluated AI-generated contracts for compilation success, unit testing, and static-analysis compliance.
  • Published in BSCI 2025 and forms the empirical component of dissertation Study 3.

Other Research Themes

My broader research explores the intersection of software engineering, artificial intelligence, and blockchain security. I have conducted empirical investigations and authored multiple peer-reviewed publications across topics such as:

  • Developer Behavior and AI Integration — analyzing how AI assistance impacts secure software practices.
  • Human-Centric Blockchain Engineering — understanding developer collaboration, challenges, and workflow design (SDS 2023, VL/HCC 2024).
  • Crypto Signal Analysis — examining how social media signals influence blockchain developer activity and repository behavior (IEEE Blockchain 2024, SANER 2025).
  • Rishan Biju — Undergraduate Researcher, Fall 2025
  • Joaquin Tuckett — Undergraduate Researcher, Summer 2025
  • Huayu Liang — M.S. Research Assistant, Spring–Fall 2022

Future Directions

My upcoming research will advance ChainSentinel into a continuously learning AI-auditor ecosystem, integrating active alignment for evolving smart contract standards and improved natural language explanations for developers.

Long-term, I aim to develop transparent and adaptive AI systems that enable human-AI collaboration in security-critical software engineering domains.