Poison Attack and Poison Detection on Deep Source Code Processing Models

Article Computer Science, Artificial Intelligence

Combining Graph Neural Networks With Expert Knowledge for Smart Contract Vulnerability Detection

Zhenguang Liu et al.

Summary: Smart contract vulnerability detection has gained significant attention recently. Existing methods heavily rely on expert-defined rules, which are labor-intensive and non-scalable. This paper proposes a method using graph neural networks and expert knowledge for smart contract vulnerability detection. Experimental results show improved accuracy compared to state-of-the-art methods on different types of vulnerabilities.

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING (2023)

添加到收藏夹

Article Computer Science, Software Engineering

DeepWukong: Statically Detecting Software Vulnerabilities Using Deep Graph Neural Network

Xiao Cheng et al.

Summary: Static bug detection is effective in detecting well-defined memory errors, but modern software systems have diverse and complex vulnerabilities. DeepWukong utilizes deep learning to embed code fragments and detect software vulnerabilities in C/C++ programs, showing promising results in addressing general static code analysis challenges.

ACM TRANSACTIONS ON SOFTWARE ENGINEERING AND METHODOLOGY (2021)

添加到收藏夹

Article Computer Science, Theory & Methods

Combining Graph-Based Learning With Automated Data Collection for Code Vulnerability Detection

Huanting Wang et al.

Summary: This paper presents FUNDED, a novel learning framework that leverages graph neural networks to develop a graph-based learning method. Unlike prior work, Funded operates on a graph representation of the program source code, capturing the program syntax, semantics, and flows.

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

A Targeted Attack on Black-Box Neural Machine Translation with Parallel Data Poisoning

Chang Xu et al.

Summary: This study highlights the potential risks of targeted attacks on black-box NMT systems, where attackers can exploit system vulnerabilities by poisoning a small fraction of training data. Even on systems trained with massive amounts of data, these attacks can still be successful with surprisingly low poisoning budgets.

PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE 2021 (WWW 2021) (2021)

添加到收藏夹

Proceedings Paper Computer Science, Software Engineering

A Syntax-Guided Edit Decoder for Neural Program Repair

Qihao Zhu et al.

Summary: Recoder, a novel syntax-guided edit decoder with placeholder generation, outperforms traditional APR methods by efficiently repairing program errors, achieving significant improvements on various benchmark datasets.

PROCEEDINGS OF THE 29TH ACM JOINT MEETING ON EUROPEAN SOFTWARE ENGINEERING CONFERENCE AND SYMPOSIUM ON THE FOUNDATIONS OF SOFTWARE ENGINEERING (ESEC/FSE '21) (2021)

添加到收藏夹

Proceedings Paper Computer Science, Software Engineering

CURE: Code-Aware Neural Machine Translation for Automatic Program Repair

Nan Jiang et al.

Summary: CURE is a new NMT-based APR technique that outperforms existing approaches by pre-training a programming language model, designing a code-aware search strategy, and using subword tokenization technique to generate a smaller search space with more correct fixes.

2021 IEEE/ACM 43RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE 2021) (2021)

添加到收藏夹

Proceedings Paper Computer Science, Software Engineering

DeepPayload: Black-box Backdoor Attack on Deep Learning Models through Neural Payload Injection

Yuanchun Li et al.

Summary: This paper introduces a highly practical backdoor attack achieved with reverse-engineering techniques over compiled deep learning models, showing its effectiveness and vulnerability of real-world mobile deep learning apps.

2021 IEEE/ACM 43RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE 2021) (2021)

添加到收藏夹

Article Computer Science, Software Engineering