Raptor vision-invasive hunt optimization enabled light graph attention coupled gated graph sequence neural network for vulnerability detection

Raptor vision-invasive hunt optimization enabled light graph attention coupled gated graph sequence neural network for vulnerability detection

연구 분야: Infrastructure

학회: Journal of Computer Virology and Hacking Techniques

Vulnerabilities in the Ethereum smart contract data affect the network’s hardware and software, as well as the configuration, through several malware attacks, and even human errors. Thus, the detection of vulnerabilities remains critical due to the dynamic environments of attacks in recent decades. Still, there emerged several existing research that worked to enhance the efficiency of vulnerability detection but ended up with certain drawbacks. The challenges addressed are high false positive rates, limited scope with complex vulnerabilities, prioritization, interpretation of vulnerabilities, and so on. To tackle the limitation addressed by the existing research, and to achieve the précised outcomes in vulnerability detection, Raptor Vision-Invasive hunt Optimization enabled Light Graph Attention coupled Gated Graph Sequence Neural Network (RVIhO-LGAtt-G2SN) is proposed in the research. The G2SN model utilized in the research is efficient when integrated with the LGAtt mechanism,as it discovers the semantic structure or the patterns of vulnerabilities that aid in achieving the most robust detection outcomes. Further, the RVIhO algorithm utilized in the research aids the model in choosing the optimal outcomes with improved operational efficiency. Therefore, the RVIhO-LGAtt-G2SNprovides effective performance outcomes when compared with other conventional methods, which are evaluated with metrics such as precision, recall, and accuracy that attained values 97.16%, 95.54%, and 96.30% respectively.