Smart Contract Vulnerability Detection Using Large Language Models

Smart contracts are immutable, self-executing programs deployed on blockchain networks that automate decentralized financial transactions and governance. Despite securing vast amounts of digital assets, they remain critically vulnerable to reentrancy attacks, integer overflows, access-control flaws, and logic errors that have caused cumulative losses exceeding several billion dollars. Traditional security approaches (e.g., manual auditing, rule-based static analyzers such as Slither and Mythril, and formal verification) struggle with scalability, evolving attack patterns, and the semantic complexity of modern contract logic. The rapid advancement of large language models (LLMs) with strong code-comprehension capabilities has opened a transformative research frontier at the intersection of artificial intelligence and blockchain security, offering scalable, context-aware, and explainable vulnerability analysis.This Research Topic aims to consolidate and accelerate emerging research on leveraging LLMs and AI/ML techniques for smart contract security. While the field remains nascent, early frameworks (including SmartLLMSentry, LLM-BSCVM, and SmartGuard, alongside comparative LLM evaluation studies) have shown strong promise for automating vulnerability detection and classification. Critical open challenges remain; standardized benchmarks are lacking, prompt-engineering strategies for security tasks are underexplored, fine-tuning methodologies on contract corpora are immature, and the integration of LLMs with established static and dynamic analysis pipelines is not yet well understood. This collection seeks to establish community-shared best practices, evaluation methodologies, benchmark datasets, and architectural frameworks, bridging the gap between state-of-the-art AI capabilities and the security demands of decentralized systems.We invite Original Research, Review, Mini-Review, Perspective, and Methods articles addressing:(1) LLM-based vulnerability detection and classification across Solidity, Vyper, and Rust-based contract languages;(2) prompt-engineering and instruction-tuning strategies for code security tasks;(3) hybrid frameworks combining LLMs with static analysis, fuzzing, or formal verification;(4) construction and evaluation of smart contract vulnerability benchmarks and datasets;(5) explainability and interpretability of AI-driven security tools;(6) privacy-preserving and federated approaches for contract analysis;(7) real-world deployment case studies and auditing-pipeline integration; and(8) adversarial robustness of LLM-based detectors.Contributions on both Ethereum-compatible contracts and emerging multi-chain environments are welcome. Topic editors: Mohammed Tawfik, Ali Al-madani, Warda M. Shaban, Islam S. Saeed Source: https://www.frontiersin.org/research-topics/84206/smart-contract-vulnerability-detection-using-large-language-models