As software complexity increases, integrating vulnerability detection becomes essential to ensure the security and integrity of modern systems. Traditional static and dynamic analysis methods face limitations in efficiency and accuracy, particularly for large-scale vulnerability detection, while existing deep learning methods struggle to fully capture structural information and dependencies in code, leading to incomplete identification of vulnerabilities. In this paper, we propose VulEPEDE, an innovative function-level vulnerability detection method. VulEPEDE leverages Program Dependency Graphs (PDG) to represent function code and constructs a Vulnerability Semantic Dependency Graph (VSDG) using slicing techniques, introducing function parameter nodes as slicing candidates to capture more comprehensive vulnerability trigger chains. It integrates two core modules: the Enhanced Positional Encoding (EPE) module and the Dependency Embedding (DE) module. The EPE module combines node attribute encoding with positional encoding using a Transformer and multi-head attention mechanism to capture complex features and contextual semantics of code, while the DE module learns dependency embeddings between code nodes through convolutional neural networks. We evaluate VulEPEDE on three widely used datasets, comparing its performance against state-of-the-art deep learning-based methods. Experimental results demonstrate that VulEPEDE outperforms the best baseline methods by 1.66%, 17.54%, and 28.96% in F1-score across the three datasets, with considerable computational efficiency.

Vulnerabilities in software and distributed systems are increasing, and system security becomes a challenge for developers. Giving a quick vulnerability classification for newly discovered vulnerabilities is helpful for developers to quickly analyze the vulnerabilities and complete the vulnerability fix. Therefore, Res-CNN-BiTLSTM is proposed for vulnerability classification in this paper. The model consists of Text Convolutional Neural Network (TextCNN), Bidirectional improved gating mechanism (BiTLSTM) and residual blocks. Firstly, TextCNN is employed to capture local features from vulnerability description information, according to different convolutional kernel sizes. Secondly, an improved gating mechanism (TLSTM) is constructed to enhance the expressive power of Long Short-Term Memory (LSTM). BiTLSTM is employed to extract long-text features of vulnerability descriptions through both forward and backward directions. Thirdly, the residual block is used for enhancing and fusing the local features with the long text features to retain important text features. Finally, a fully connected layer is employed to classify vulnerabilities. The results show that Res-CNN-BiTLSTM has the highest macro precision (MacroP), macro recall (MacroR), and macro F1 (MacroF1) on National Vulnerability Database (NVD) dataset. It has a high prediction precision on CWE-352 and CWE-416. The performance of Res-CNN-BiTLSTM is better than the other comparative models. It is effective for vulnerability automatic classification. Meanwhile, the effectiveness of each part of the model is proved by ablation experiments.