An Explainable AI Model for the Detecting Malicious Smart Contracts Based on EVM Opcode Based Features

High detection accuracy with minimal false alarms: The proposed system successfully identifies 99% of malicious smart contracts while maintaining a false positive rate of only 0.01, meaning legitimate contracts are rarely incorrectly flagged as dangerous.

Addresses real financial threats on blockchain: The research tackles concrete security vulnerabilities in Ethereum smart contracts (reentrancy attacks, randomness exploits, etc.) that have historically resulted in fund theft and service disruptions, making detection a practical necessity before deployment.

Explainability enables human oversight: By incorporating the LIME algorithm, the system can explain why it classifies a contract as malicious based on specific code patterns (EVM opcodes), allowing developers and auditors to understand and verify the AI's reasoning rather than treating it as a "black box."

Technical preprocessing improves fairness: The use of SMOTE algorithm to balance the training dataset and entropy-based feature transformation helps prevent the model from being biased toward overrepresented malicious contract types, supporting more equitable detection across different attack patterns.

Practical deployment pathway: The focus on EVM opcodes (the low-level instructions executed on Ethereum) provides a concrete, implementable approach that could be integrated into blockchain development workflows to catch threats before contracts go live.