Attention mechanism-based vulnerability detection method for graph neural network source code

With the development of artificial intelligence and machine learning, automatic code vulnerability detection methods based on machine learning and deep learning have been gradually applied. However, the false alarm rate and leakage rate of vulnerability detection are still a concern. To this end, a graph neural network source code vulnerability detection method based on the attention mechanism is proposed, which fully captures the syntactic and semantic information related to the vulnerability through the code attribute graph, and at the same time combines the attention mechanism to enhance the model's ability to understand the features. Experimental results show that compared with existing methods, our accuracy and precision are improved by 5.7% and 4.2%, respectively, which significantly improves the vulnerability detection capability.