AmirhosseinHonardoust/How-AI-Detects-Rugpulls

Static analysis via regex pattern matching on Solidity source code extracts behavioral features (owner minting, fee manipulation, trading locks, blacklists) and converts them into a weighted numeric risk score—mimicking ML feature importance without requiring labeled training data. The approach prioritizes interpretability over complexity, mapping scores to categorical labels (safe/suspicious/rugpull_candidate) that mirror real auditor assessment language, while acknowledging gaps in detecting runtime exploits, liquidity dynamics, and multi-contract interactions.