Paper ID: 2211.10724

Deep Smart Contract Intent Detection

Youwei Huang, Sen Fang, Jianwen Li, Jiachun Tao, Bin Hu, Tao Zhang

In recent years, researchers in the software security field have focused on detecting vulnerabilities in smart contracts to avoid significant losses of crypto assets on the blockchain. Despite early successes in this domain, detecting developers' intents in smart contracts is a more pressing issue, as malicious intents have resulted in substantial financial losses. Unfortunately, existing research lacks effective methods for detecting development intents in smart contracts. To address this gap, we propose \textsc{SmartIntentNN} (Smart Contract Intent Neural Network), a deep learning model designed to automatically detect development intent in smart contracts. \textsc{SmartIntentNN} utilizes a pre-trained sentence encoder to generate contextual representations of smart contract code, a K-means clustering model to identify and highlight prominent intent features, and a bidirectional LSTM-based deep neural network for multi-label classification. We trained and evaluated \textsc{SmartIntentNN} on a dataset comprising over 40,000 real-world smart contracts, employing self-comparison baselines in our experimental setup. The results demonstrate that \textsc{SmartIntentNN} achieves an F1-score of 0.8633 in identifying intents across 10 distinct categories, outperforming all baselines and filling the gap in smart contract detection by incorporating intent analysis.

Submitted: Nov 19, 2022