03 RESEARCHpublications
Non-EVM smart contract security and vehicular network privacy.
1
PUBLISHED (SPRINGER LNCS)
216
LLM EVALUATION EXPERIMENTS RUN
24
CONTRACT VULN. BENCHMARK
Springer LNCS, METAVERSE 2026 · published
LLM-Assisted Smart Contract Vulnerability Detection for Metaverse Blockchain Infrastructure
Full title: "Securing Metaverse Blockchain Infrastructure: LLM-Assisted Vulnerability Detection on Solana and Algorand Smart Contracts." Investigated LLM-driven vulnerability detection for non-EVM blockchain ecosystems, addressing security challenges in metaverse infrastructure, decentralized applications, and smart contract auditing.
- Developed a benchmark of 24 vulnerable and patched Solana and Algorand smart contracts covering eight vulnerability classes.
- Designed and executed 216 experiments comparing Zero-Shot, Chain-of-Thought (CoT), and Retrieval-Augmented Generation (RAG) prompting strategies for vulnerability detection and remediation.
- Evaluated detection accuracy, false positive rates, reasoning quality, and hallucination reduction across frontier LLMs (GPT-4o, Claude Sonnet 4, Llama-3.3-70B) for automated smart contract security auditing.
- Demonstrated the effectiveness of structured prompting techniques for improving LLM-assisted security analysis and scalable smart contract auditing.
LLMs (GPT-4o, Claude Sonnet 4, Llama-3.3-70B)SolanaAlgorandRustRAG
Assisted on the technical side
A Blockchain-based Privacy-Preserving Authentication for Secure Vehicular Communications
Acted as an integral researcher addressing critical security and privacy gaps in Vehicular Ad Hoc Networks (VANETs): anonymity, unlinkability, and data integrity.
- Designed an Inter-Platoon Communication Protocol (IPCP) enabling secure coordination (lane changes, accident avoidance) between vehicle platoons.
- Engineered a scalable, RSU-less architecture removing reliance on fixed Road Side Units, reducing cost and complexity.
- Implemented a distributed blockchain for immutable event logging and dynamic trust-score management, for accountability.
- Developed a custom, low-latency consensus mechanism (replacing PoW) tailored for real-time vehicular operations.
- Built specialized protocols for large vehicle dynamics and emergency vehicle priority.
- Ensured user privacy via a pseudonym-based system managed by a central Trust Authority (TA) for traceability.
PythonC++SolidityPlexeRShell
want the vulnerability disclosures behind the CVE count? see security →