Dong Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Dong Li has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Meng Yan
Chongqing University
4 shared publications - Chao Liu
Chongqing University, Chongqing, China
3 shared publications - Xiaohong Zhang
Chongqing University
3 shared publications - David Lo
School of Computing and Information Systems, Singapore Management University, Singapore, Singapore
2 shared publications - Zhongxin Liu
Zhejiang University
2 shared publications - Dan Yang
School of Information Science and Technology, Southwest Jiaotong University, Chengdu, China
1 shared publication
Latest Publications
- 1 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 51 (9) , 2634-2650
Improving Co-Decoding Based Security Hardening of Code LLMs Leveraging Knowledge Distillation
Large Language Models (LLMs) have been widely adopted by developers in software development. However, the massive pretraining code data is not rigorously filtered, allowing LLMs to learn unsafe coding patterns. Several prior studies have demonstrated that code LLMs tend to generate code with potential vulnerabilities. The widespread adoption of intelligent programming assistants poses a significant threat to the software development process. Existing approaches to mitigating this risk primarily involve constructing secure data that are free of vulnerabilities and then retraining or fine-tuning the models. However, such an effort is resource intensive and requires significant manual supervision. When the model parameters are too large (e.g., more than 1 billion) or multiple models with the same parameter scale have the same optimization needs (e.g., to avoid outputting vulnerable code), the above work will become unaffordable. To address this challenge, in previous work, we proposed CoSec, an approach to improve the security of code LLMs with different parameters by utilizing an independent and very small parametric security model as a decoding navigator. Despite CoSec’s excellent performance, we found that there is still room for improving: 1) its ability to maintain the functional correctness of hardened targets, and 2) the security of the generated code. To address the above issues, we propose CoSec+, a hardening framework consisting of three phases: 1) Functional Correctness Alignment, which improves the functional correctness of the security base with knowledge disstillation; 2) Security Training, which yields an independent, but much smaller security model; and 3) Co-decoding, where the security model iteratively reasons about the next token along with the target model. Due to the higher confidence that a well-trained security model places in secure and correct tokens, it guides the target base model to generate more secure code, even as it improves the functional correctness of the target base model. We have conducted extensive experiments in several code LLMs (i.e., CodeGen, StarCoderBase, DeepSeekCoder and Qwen2.5-Coder), and the results show that our approach is effective in improving the functional correctness and security of the models. The evaluation results show that CoSec+ can deliver a 0.8% to 37.7% improvement in security across models of various parameter sizes and families; moreover, it preserves the functional correctness of the target base models—achieving functional-correctness gains of 0.7% to 51.1% for most of those models.
- 1 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 51 (9) , 2634-2650
Improving Co-Decoding Based Security Hardening of Code LLMs Leveraging Knowledge Distillation
Large Language Models (LLMs) have been widely adopted by developers in software development. However, the massive pretraining code data is not rigorously filtered, allowing LLMs to learn unsafe coding patterns. Several prior studies have demonstrated that code LLMs tend to generate code with potential vulnerabilities. The widespread adoption of intelligent programming assistants poses a significant threat to the software development process. Existing approaches to mitigating this risk primarily involve constructing secure data that are free of vulnerabilities and then retraining or fine-tuning the models. However, such an effort is resource intensive and requires significant manual supervision. When the model parameters are too large (e.g., more than 1 billion) or multiple models with the same parameter scale have the same optimization needs (e.g., to avoid outputting vulnerable code), the above work will become unaffordable. To address this challenge, in previous work, we proposed CoSec, an approach to improve the security of code LLMs with different parameters by utilizing an independent and very small parametric security model as a decoding navigator. Despite CoSec’s excellent performance, we found that there is still room for improving: 1) its ability to maintain the functional correctness of hardened targets, and 2) the security of the generated code. To address the above issues, we propose CoSec+, a hardening framework consisting of three phases: 1) Functional Correctness Alignment, which improves the functional correctness of the security base with knowledge disstillation; 2) Security Training, which yields an independent, but much smaller security model; and 3) Co-decoding, where the security model iteratively reasons about the next token along with the target model. Due to the higher confidence that a well-trained security model places in secure and correct tokens, it guides the target base model to generate more secure code, even as it improves the functional correctness of the target base model. We have conducted extensive experiments in several code LLMs (i.e., CodeGen, StarCoderBase, DeepSeekCoder and Qwen2.5-Coder), and the results show that our approach is effective in improving the functional correctness and security of the models. The evaluation results show that CoSec+ can deliver a 0.8% to 37.7% improvement in security across models of various parameter sizes and families; moreover, it preserves the functional correctness of the target base models—achieving functional-correctness gains of 0.7% to 51.1% for most of those models.
- Open Access
- 23 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1198-1202
CoSEFA: An LLM-Based Programming Assistant for Secure Code Generation via Supervised Co-Decoding
- 23 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1198-1202
Open AccessCoSEFA: An LLM-Based Programming Assistant for Secure Code Generation via Supervised Co-Decoding
- Open Access
- 11 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1428-1439
CoSec: On-the-Fly Security Hardening of Code LLMs via Supervised Co-decoding
Large Language Models (LLMs) specialized in code have shown exceptional proficiency across various programming-related tasks, particularly code generation. Nonetheless, due to its nature of pretraining on massive uncritically filtered data, prior studies have shown that code LLMs are prone to generate code with potential vulnerabilities. Existing approaches to mitigate this risk involve crafting data without vulnerability and subsequently retraining or fine-tuning the model. As the number of parameters exceeds a billion, the computation and data demands of the above approaches will be enormous. Moreover, an increasing number of code LLMs tend to be distributed as services, where the internal representation is not accessible, and the API is the only way to reach the LLM, making the prior mitigation strategies non-applicable. To cope with this, we propose CoSec, an on-the-fly Security hardening method of code LLMs based on security model-guided Co-decoding, to reduce the likelihood of code LLMs to generate code containing vulnerabilities. Our key idea is to train a separate but much smaller security model to co-decode with a target code LLM. Since the trained secure model has higher confidence for secure tokens, it guides the generation of the target base model towards more secure code generation. By adjusting the probability distributions of tokens during each step of the decoding process, our approach effectively influences the tendencies of generation without accessing the internal parameters of the target code LLM. We have conducted extensive experiments across various parameters in multiple code LLMs (i.e., CodeGen, StarCoder, and DeepSeek-Coder), and the results show that our approach is effective in security hardening. Specifically, our approach improves the average security ratio of six base models by 5.02%-37.14%, while maintaining the functional correctness of the target model.
- 11 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1428-1439
Open AccessCoSec: On-the-Fly Security Hardening of Code LLMs via Supervised Co-decoding
Large Language Models (LLMs) specialized in code have shown exceptional proficiency across various programming-related tasks, particularly code generation. Nonetheless, due to its nature of pretraining on massive uncritically filtered data, prior studies have shown that code LLMs are prone to generate code with potential vulnerabilities. Existing approaches to mitigate this risk involve crafting data without vulnerability and subsequently retraining or fine-tuning the model. As the number of parameters exceeds a billion, the computation and data demands of the above approaches will be enormous. Moreover, an increasing number of code LLMs tend to be distributed as services, where the internal representation is not accessible, and the API is the only way to reach the LLM, making the prior mitigation strategies non-applicable. To cope with this, we propose CoSec, an on-the-fly Security hardening method of code LLMs based on security model-guided Co-decoding, to reduce the likelihood of code LLMs to generate code containing vulnerabilities. Our key idea is to train a separate but much smaller security model to co-decode with a target code LLM. Since the trained secure model has higher confidence for secure tokens, it guides the generation of the target base model towards more secure code generation. By adjusting the probability distributions of tokens during each step of the decoding process, our approach effectively influences the tendencies of generation without accessing the internal parameters of the target code LLM. We have conducted extensive experiments across various parameters in multiple code LLMs (i.e., CodeGen, StarCoder, and DeepSeek-Coder), and the results show that our approach is effective in security hardening. Specifically, our approach improves the average security ratio of six base models by 5.02%-37.14%, while maintaining the functional correctness of the target model.
- Open Access
- 15 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 133-137
AW4C: A Commit-Aware C Dataset for Actionable Warning Identification
Excessive non-actionable warnings generated by static program analysis tools can hinder developers from utilizing these tools effectively. Leveraging learning-based approaches for actionable warning identification has demonstrated promise in boosting developer productivity, minimizing the risk of bugs, and reducing code smells. However, the small sizes of existing datasets have limited the model choices for machine learning researchers, and the lack of aligned fix commits limits the scope of the dataset for research. In this paper, we present AW4C, an actionable warning C dataset that contains 38,134 actionable warnings mined from more than 500 repositories on GitHub. These warnings are generated via Cppcheck, and most importantly, each warning is precisely mapped to the commit where the corrective action occurred. To the best of our knowledge, this is the largest publicly available actionable warning dataset for C programming language to date. The dataset is suited for use in machine/deep learning models and can support a wide range of tasks, such as actionable warning identification and vulnerability detection. Furthermore, we have released our dataset1 and a general framework for collecting actionable warnings on GitHub2 to facilitate other researchers to replicate our work and validate their innovative ideas.
- 15 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 133-137
Open AccessAW4C: A Commit-Aware C Dataset for Actionable Warning Identification
Excessive non-actionable warnings generated by static program analysis tools can hinder developers from utilizing these tools effectively. Leveraging learning-based approaches for actionable warning identification has demonstrated promise in boosting developer productivity, minimizing the risk of bugs, and reducing code smells. However, the small sizes of existing datasets have limited the model choices for machine learning researchers, and the lack of aligned fix commits limits the scope of the dataset for research. In this paper, we present AW4C, an actionable warning C dataset that contains 38,134 actionable warnings mined from more than 500 repositories on GitHub. These warnings are generated via Cppcheck, and most importantly, each warning is precisely mapped to the commit where the corrective action occurred. To the best of our knowledge, this is the largest publicly available actionable warning dataset for C programming language to date. The dataset is suited for use in machine/deep learning models and can support a wide range of tasks, such as actionable warning identification and vulnerability detection. Furthermore, we have released our dataset1 and a general framework for collecting actionable warnings on GitHub2 to facilitate other researchers to replicate our work and validate their innovative ideas.