CBGRU: A Detection Method of Smart Contract Vulnerability Based on a Hybrid Model

Area of Science:

• Computer Science
• Blockchain Technology
• Artificial Intelligence

Background:

• Smart contracts are increasingly used across various sectors like IoT, finance, and healthcare.
• The proliferation of smart contracts has led to a rise in security vulnerabilities, causing significant financial losses.
• Current vulnerability detection tools face scalability issues due to reliance on expert-defined hard rules, slowing detection with increasing contract complexity.

Purpose of the Study:

• To propose a novel hybrid deep learning model, CBGRU, for enhanced smart contract vulnerability detection.
• To address the limitations of existing rule-based detection methods by leveraging advanced machine learning techniques.
• To improve the accuracy and efficiency of identifying security flaws in smart contracts.

Main Methods:

• Developed a hybrid deep learning model named CBGRU.
• Integrated multiple word embedding techniques (Word2Vec, FastText) with diverse deep learning architectures (LSTM, GRU, BiLSTM, CNN, BiGRU).
• Employed feature extraction and combination from various deep learning models for vulnerability detection.

Main Results:

• The CBGRU model demonstrated significant performance in detecting smart contract vulnerabilities on the SmartBugs Dataset-Wild.
• Experimental results confirmed the effectiveness of the hybrid approach in identifying security flaws.
• Comparative analysis showed superior performance of CBGRU over existing methods.

Conclusions:

• The proposed CBGRU hybrid model offers a powerful and efficient solution for smart contract vulnerability detection.
• This deep learning approach overcomes the limitations of traditional rule-based systems.
• CBGRU represents a significant advancement in securing blockchain applications through improved smart contract analysis.