Devign: Effective Vulnerability Identification by Learning Comprehensive Program Semantics via Graph Neural Networks
By Yaqin Zhou, Shangqing Liu, Jing Kai Siow, Xiaoning Du, Yang Liu. 2019
1,186 citations recorded by the source corpus.
Vulnerability identification is crucial to protect the software systems from attacks for cyber security. It is especially important to localize the vulnerable functions among the source code to facilitate the fix. However, it is a challenging and tedious process, and also requires specialized security expertise. Inspired by the work on manually-defined patterns of vulnerabilities from various code representation graphs and the recent advance on graph neural networks, we propose Devign, a general graph neural network based model for graph-level classification through learning on a rich set of code semantic representations. It includes a novel Conv module to efficiently extract useful features in the learned rich node representations for graph-level classification. The model is trained over manually labeled datasets built on 4 diversified large-scale open-source C projects that incorporate high complexity and variety of real source code instead of synthesis code used in previous works. The results of the extensive evaluation on the datasets demonstrate that Devign outperforms the state of the arts significantly with an average of 10.51% higher accuracy and 8.68% F1 score, increases averagely 4.66% accuracy and 6.37% F1 by the Conv module.
Subfields: Information Systems, Software, Program Analysis & Verification, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Information Systems
In Software
- Chakkrit Kla Tantithamthavorn
- Michael Fu
- Shaohua Wang
- Yi Li
- Yanqiao Zhu
- Yichen Xu
- Tien Nhut Nguyen
- Ge Li
In Program Analysis & Verification
- Chakkrit Kla Tantithamthavorn
- Michael Fu
- Tien Nhut Nguyen
- Shaohua Wang
- Yi Li
- Ge Li
- Saurabh Pujar
- Zhi Jin
In Signal Processing
Citations in the corpus
- CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation (2021, methodology): “We observe all our model variants significantly outperform prior work with either an encode-only (RoBERTa, CodeBERT, DOBF) or encoder-decoder framework (PLBART).”
- Unified Pre-training for Program Understanding and Generation (2021, background): “, 2020), Bug Detection (Ray et al., 2016; Li et al., 2018b; Russell et al., 2018; Zhou et al., 2019; Chakraborty et al., 2020), Program Repair (Chen et al.”
- A Survey on Pretrained Language Models for Neural Code Intelligence (2022, methodology): “Devign (Zhou et al., 2019) a defect detection dataset created from four largescale open-source C projects, comprising 48,687 samples with diverse vulnerabilities.”
- CoCoMIC: Code Completion by Jointly Modeling In-file and Cross-file Context (2022, neutral): “…to pretrain Transformer language models using unlabeled source code (Feng et al., 2020; Ahmad et al., 2021; Wang et al., 2021b; Guo et al., 2022; Ding et al., 2022b) to facilitate software engineering applications (Husain et al., 2019; Iyer et al., 2018; Tufano et al., 2019; Zhou et al., 2019).”
- A C/C++ Code Vulnerability Dataset with Code Changes and CVE Summaries (2020, methodology): “There are several existing vulnerability datasets created by previous studies [2, 5, 11].”
- Project CodeNet: A Large-Scale AI for Code Dataset for Learning a Diversity of Coding Tasks (2021, background): “Such tasks include clone detection, vulnerability detection [7, 8], cloze test [9], code completion [10, 11], code repair [12], code-to-code translation, natural language code search [13], text-to-code generation [14], and code summarization [13].”
- Integrating Tree Path in Transformer for Code Representation (2021, background): “[Zhou et al., 2019] proposed a novel graph neural network with composite programming representation for vulnerability identification.”
- LineVul: A Transformer-based Line-Level Vulnerability Prediction (2022, methodology): “For instance, VulDeePecker [33] leverages symbolic representations on program slices, Devign [65] uses graph embedding on code property graphs (i.”
- Combining Graph Neural Networks With Expert Knowledge for Smart Contract Vulnerability Detection (2021, background): “[34] develops Devign, a general graph neural”
- Deep Learning Based Vulnerability Detection: Are We There Yet? (2020, methodology): “[6] used regularized cross entropy loss.”
- CVEfixes: automated collection of vulnerabilities and their fixes from open-source software (2021, methodology): “The Devign [21] dataset includes four real-world open-source C/C++ projects, Linux, FFmpeg, Qemu and Wireshark, where the labeling is performed using security-related keyword filtering.”
- NoFunEval: Funny How Code LMs Falter on Requirements Beyond Functional Correctness (2024, methodology): “Prior works have considered code-comprehension tasks like Clone-detection (Svajlenko et al., 2014; Lu et al., 2021), Defect-detection (Zhou et al., 2019; Li et al., 2021; Lu et al., 2021), Code-explaination (Muennighoff et al., 2023; Leinonen et al., 2023), and Question Answering over Code (Huang…”
- CoTexT: Multi-task Learning with Code-Text Transformer (2021, methodology): “The task uses the Devign dataset (Zhou et al., 2019), which contains C programming language from open-source projects.”
- HELoC: Hierarchical Contrastive Learning of Source Code Representation (2022, background): “…years, many deep learning based approaches to software engineering tasks, such as code retrieval [1, 2], code clone detection [3–6], code comment generation [7–9], software defect prediction [10–12], and code summarization [13–16], have emerged to improve software development and maintenance.”
- Vulnerability detection with fine-grained interpretations (2021, methodology): “For the method-level VD, our results show that IVDetect outperforms the existing ML/DL-based approaches [11, 19, 20, 27, 37] by 43%–84% and 105%–255% at the top 10 list for two ranking scores nDCG and MAP, respectively.”