HGS-RF: Heuristic-Guided Selective Random Forest For Enhancing Digital Forensic Investigations Based On Detecting Vulnerabilities In Smart Contracts

In recent years, Blockchain technology has stood out and widely spread. Blockchain is a decentralized structure that executes operations and transactions automatically. These automatic operations mainly depend on smart contracts, which are among the most important components of Blockchain. Smart contracts involve a set of business transactions. They can be described in simpler terms as an agreement between two parties, similar to any sales contract that is entered into between two people. Once the pre-defined terms of the contract are met, the contract is automatically verified and executed. The automated execution nature of smart contracts makes them vulnerable to numerous security threats. Attackers try to exploit vulnerabilities within the smart contract to steal funds. This leads to serious financial and operational consequences for companies and institutions that depend on Blockchain networks in their operations. This research proposes the HGS-RF (Heuristic-Guided Selective Random Forest) framework, which is a hybrid model that enhances digital forensic investigations by detecting vulnerabilities in smart contracts. The methodology integrates Natural Language Processing (NLP) for feature extraction using the RoBERTA model. Then it applies a binary classification process using a Multilayer Perceptron (MLP) network to verify if the contract is secure or vulnerable. If a contract is vulnerable, a multi-class classification process is performed. This multi-class classification process is based on the HGS-RF model. The multilayer classification process addresses the critical gap, which is multi-label detection, where a single node contains multiple security vulnerabilities. The MLP achieved a performance with 98–99% F1-score, and the HGS-RF model achieved an overall accuracy of 96.7–97% across datasets with up to 37 vulnerability types. The proposed model aims to improve the accuracy and efficiency of smart contract vulnerability detection and contribute to finding a solution to this problem in line with current security needs and long-term sustainability goals.