Yanjing Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yanjing Yang has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- He Zhang
Software Institute, Nanjing University, Nanjing, China
7 shared publications - Bohan Liu
State Key Laboratory of Novel Software Technology, Software Institute, Nanjing University, Nanjing, China
5 shared publications - Xin Zhou
State Key Laboratory of Novel Software Technology, Software Institute, Nanjing University, Nanjing, China
5 shared publications - Jinwei Xu
State Key Laboratory of Novel Software Technology, Software Institute, Nanjing University, Nanjing, China
4 shared publications - Lanxin Yang
State Key Laboratory of Novel Software Technology, Software Institute, Nanjing University, Nanjing, China
3 shared publications - Haifeng Shen
Faculty of Science and Engineering, Southern Cross University, Gold Coast, Australia
2 shared publications
Latest Publications
- Open Access
- 28 August 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Software Engineering and Methodology
Leveraging Self-Paced Learning for Software Vulnerability Detection
Vulnerabilities are major risks to software systems. Researchers have proposed many deep learning approaches to detect software vulnerabilities. However, their accuracy is limited in practice. One of the main causes is low-quality samples ( i . e ., source code and its label) in the training set. To this end, we propose a new approach: SPLVD ( S elf- P aced L earning for S oftware V ulnerability D etection). SPLVD dynamically selects samples based on the training stage, which simulates the easy-to-hard human learning process. SPLVD has a data selector specifically designed for vulnerability detection. Before each training epoch, SPLVD uses the data selector to calculate the difficulty of each sample, select parts of samples for training, and update the data selector. When evaluating SPLVD, we first use three well-known datasets (BigVul, Devign, and ReVeal) containing more than 239K samples, of which 25K are vulnerable. Experimental results demonstrate that SPLVD achieves the highest F1 of 89.2% on BigVul, 68.7% on Devign, and 43.5% on ReVeal, respectively, outperforming the state-of-the-art approaches. Moreover, we collect samples from OpenHarmony, a new ecosystem that has not been learned by general large language models, to evaluate SPLVD. SPLVD achieves the highest precision of 90.9%, demonstrating its effectiveness in real-world applications.
- 28 August 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Software Engineering and Methodology
Open AccessLeveraging Self-Paced Learning for Software Vulnerability Detection
Vulnerabilities are major risks to software systems. Researchers have proposed many deep learning approaches to detect software vulnerabilities. However, their accuracy is limited in practice. One of the main causes is low-quality samples ( i . e ., source code and its label) in the training set. To this end, we propose a new approach: SPLVD ( S elf- P aced L earning for S oftware V ulnerability D etection). SPLVD dynamically selects samples based on the training stage, which simulates the easy-to-hard human learning process. SPLVD has a data selector specifically designed for vulnerability detection. Before each training epoch, SPLVD uses the data selector to calculate the difficulty of each sample, select parts of samples for training, and update the data selector. When evaluating SPLVD, we first use three well-known datasets (BigVul, Devign, and ReVeal) containing more than 239K samples, of which 25K are vulnerable. Experimental results demonstrate that SPLVD achieves the highest F1 of 89.2% on BigVul, 68.7% on Devign, and 43.5% on ReVeal, respectively, outperforming the state-of-the-art approaches. Moreover, we collect samples from OpenHarmony, a new ecosystem that has not been learned by general large language models, to evaluate SPLVD. SPLVD achieves the highest precision of 90.9%, demonstrating its effectiveness in real-world applications.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. PP (00985589) , 1-21
Automated Localization of Affected Libraries and Versions from Vulnerability Reports
Security experts investigate vulnerability reports to analyze the impact scope of a vulnerability, specifically the affected libraries and versions. As large language models (LLMs) exhibit powerful capabilities in understanding vulnerability reports, many LLM-based approaches have been proposed to address this problem. However, existing approaches primarily focus on locating affected libraries without considering the corresponding affected versions, which results in an overestimation of vulnerability impact and leads to unnecessary protection efforts. Leveraging LLMs to locate affected libraries and versions has two challenges: inaccurate and incomplete information in the vulnerability report, which arises from the input side; LLM’s hallucinations, which arise from the output side. To do so, we propose VULLOC in this article. For the first challenge, VULLOC first performs knowledge augmentation on vulnerability reports by retrieving additional information from security vendors (e.g., Snyk) and ecosystem advisories (e.g., Jenkins). Then, VULLOC locally fine-tunes an LLM to locate affected libraries and versions based on the augmented vulnerability reports. For the second challenge, VULLOC integrates a local search algorithm to verify the validity of LLM outputs and a fixing commit detection algorithm to improve the accuracy. We evaluate VULLOC using 3,200 NVD vulnerability reports. The experimental results show that for locating affected libraries, VULLOC improves the F1@3 from 76.44% to 95.54%, and for locating affected versions, VULLOC improves the F1@3 from 39.26% to 77.35%.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. PP (00985589) , 1-21
Automated Localization of Affected Libraries and Versions from Vulnerability Reports
Security experts investigate vulnerability reports to analyze the impact scope of a vulnerability, specifically the affected libraries and versions. As large language models (LLMs) exhibit powerful capabilities in understanding vulnerability reports, many LLM-based approaches have been proposed to address this problem. However, existing approaches primarily focus on locating affected libraries without considering the corresponding affected versions, which results in an overestimation of vulnerability impact and leads to unnecessary protection efforts. Leveraging LLMs to locate affected libraries and versions has two challenges: inaccurate and incomplete information in the vulnerability report, which arises from the input side; LLM’s hallucinations, which arise from the output side. To do so, we propose VULLOC in this article. For the first challenge, VULLOC first performs knowledge augmentation on vulnerability reports by retrieving additional information from security vendors (e.g., Snyk) and ecosystem advisories (e.g., Jenkins). Then, VULLOC locally fine-tunes an LLM to locate affected libraries and versions based on the augmented vulnerability reports. For the second challenge, VULLOC integrates a local search algorithm to verify the validity of LLM outputs and a fixing commit detection algorithm to improve the accuracy. We evaluate VULLOC using 3,200 NVD vulnerability reports. The experimental results show that for locating affected libraries, VULLOC improves the F1@3 from 76.44% to 95.54%, and for locating affected versions, VULLOC improves the F1@3 from 39.26% to 77.35%.
- Top Cited PapersOpen Access
- 24 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Software Engineering and Methodology
- Vol. 35 (5) , 1-34
One Size Does Not Fit All: Investigating Efficacy of Perplexity in Detecting LLM-Generated Code
Large Language Model-Generated Code (LLMgCode) has become increasingly common in software development. So far LLMgCode has more quality issues than Human-Authored Code (HaCode). It is common for LLMgCode to mix with HaCode in a code change, while the change is signed by only human developers, without being carefully examined. Many automated methods have been proposed to detect LLMgCode from HaCode, in which the perplexity-based method ( Perplexity for short) is the state-of-the-art method. However, the efficacy evaluation of Perplexity has focused on detection accuracy. Yet it is unclear whether Perplexity is good enough in a wider range of realistic evaluation settings. To this end, we carry out a family of experiments to compare Perplexity against feature- and pre-training-based methods from three perspectives: detection accuracy , detection speed , and generalization capability . The experimental results show that Perplexity has the best generalization capability while having limited detection accuracy and detection speed. Based on that, we discuss the strengths and limitations of Perplexity , e.g., Perplexity is unsuitable for high-level programming languages. Finally, we provide recommendations to improve Perplexity and apply it in practice. As the first large-scale investigation on detecting LLMgCode from HaCode, this article provides a wide range of findings for future improvement.
- 24 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Software Engineering and Methodology
- Vol. 35 (5) , 1-34
Top Cited PapersOpen AccessOne Size Does Not Fit All: Investigating Efficacy of Perplexity in Detecting LLM-Generated Code
Large Language Model-Generated Code (LLMgCode) has become increasingly common in software development. So far LLMgCode has more quality issues than Human-Authored Code (HaCode). It is common for LLMgCode to mix with HaCode in a code change, while the change is signed by only human developers, without being carefully examined. Many automated methods have been proposed to detect LLMgCode from HaCode, in which the perplexity-based method ( Perplexity for short) is the state-of-the-art method. However, the efficacy evaluation of Perplexity has focused on detection accuracy. Yet it is unclear whether Perplexity is good enough in a wider range of realistic evaluation settings. To this end, we carry out a family of experiments to compare Perplexity against feature- and pre-training-based methods from three perspectives: detection accuracy , detection speed , and generalization capability . The experimental results show that Perplexity has the best generalization capability while having limited detection accuracy and detection speed. Based on that, we discuss the strengths and limitations of Perplexity , e.g., Perplexity is unsuitable for high-level programming languages. Finally, we provide recommendations to improve Perplexity and apply it in practice. As the first large-scale investigation on detecting LLMgCode from HaCode, this article provides a wide range of findings for future improvement.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 52 (9) , 1-18
Towards Robust Detection for Malicious Injection Variants
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 52 (9) , 1-18
Towards Robust Detection for Malicious Injection Variants
- Open Access
BacAlarm: Mining and Simulating Composite API Traffic to Prevent Broken Access Control Violations
Broken Access Control (BAC) violations, which consistently rank among the top five security risks in the OWASP API Security Top 10, refer to unauthorized access attempts arising from BAC vulnerabilities, whose successful exploitation can impose significant risks on exposed application programming interfaces (APIs). In recent years, learning-based methods have demonstrated promising prospects in detecting various types of malicious activities. However, in real-network operation and maintenance scenarios, leveraging learning-based methods for BAC detection faces two critical challenges. Firstly, under the RESTful API design principles, most systems omit recording composite traffic for performance, and together with ethical and legal bans on directly testing real-world systems, this leads to a critical shortage of training data for detecting BAC violations. Secondly, common malicious behaviors such as SQL injection typically generate individual access traffic that is inherently anomalous. In contrast, BAC is usually composed of multiple correlated access requests that appear normal when examined in isolation. To tackle these problems, we introduce \BAC, an approach for establishing a BAC violation detection model by generating and utilizing API traffic data. The \BAC consists of an API Traffic Generator and a BAC Detector. Experimental results show that \BAC outperforms current state-of-the-art invariant-based and learning-based methods with the $\text{F}_1$ and MCC improving by 21.2\% and 24.1\%.Open AccessBacAlarm: Mining and Simulating Composite API Traffic to Prevent Broken Access Control Violations
Broken Access Control (BAC) violations, which consistently rank among the top five security risks in the OWASP API Security Top 10, refer to unauthorized access attempts arising from BAC vulnerabilities, whose successful exploitation can impose significant risks on exposed application programming interfaces (APIs). In recent years, learning-based methods have demonstrated promising prospects in detecting various types of malicious activities. However, in real-network operation and maintenance scenarios, leveraging learning-based methods for BAC detection faces two critical challenges. Firstly, under the RESTful API design principles, most systems omit recording composite traffic for performance, and together with ethical and legal bans on directly testing real-world systems, this leads to a critical shortage of training data for detecting BAC violations. Secondly, common malicious behaviors such as SQL injection typically generate individual access traffic that is inherently anomalous. In contrast, BAC is usually composed of multiple correlated access requests that appear normal when examined in isolation. To tackle these problems, we introduce \BAC, an approach for establishing a BAC violation detection model by generating and utilizing API traffic data. The \BAC consists of an API Traffic Generator and a BAC Detector. Experimental results show that \BAC outperforms current state-of-the-art invariant-based and learning-based methods with the $\text{F}_1$ and MCC improving by 21.2\% and 24.1\%. - 11 August 2025
- Published by Springer Nature in Automated Software Engineering
- Vol. 32 (2) , 1-38
Automated detection of affected libraries from vulnerability reports
- 11 August 2025
- Published by Springer Nature in Automated Software Engineering
- Vol. 32 (2) , 1-38
Automated detection of affected libraries from vulnerability reports
- Open Access
Enhancing the Accuracy and Comprehensibility in Architectural Tactics Detection via Small Model-Augmented Prompt Engineering
Architectural tactics (ATs), as the concrete implementation of architectural decisions in code, address non-functional requirements of software systems. Due to the implicit nature of architectural knowledge in code implementation, developers may risk inadvertently altering or removing these tactics during code modifications or optimizations. Such unintended changes can trigger architectural erosion, gradually undermining the system's original design. While many researchers have proposed machine learning-based methods to improve the accuracy of detecting ATs in code, the black-box nature and the required architectural domain knowledge pose significant challenges for developers in verifying the results. Effective verification requires not only accurate detection results but also interpretable explanations that enhance their comprehensibility. However, this is a critical gap in current research. Large language models (LLMs) can generate easily interpretable ATs detection comments if they have domain knowledge. Fine-tuning LLMs to acquire domain knowledge faces challenges such as catastrophic forgetting and hardware constraints. Thus, we propose Prmt4TD, a small model-augmented prompting framework to enhance the accuracy and comprehensibility of ATs detection. Combining fine-tuned small models with In-Context Learning can also reduce fine-tuning costs while equipping the LLM with additional domain knowledge. Prmt4TD can leverage the remarkable processing and reasoning capabilities of LLMs to generate easily interpretable ATs detection results. Our evaluation results demonstrate that Prmt4TD achieves accuracy (\emph{F1-score}) improvement of 13\%-23\% on the ATs balanced dataset and enhances the comprehensibility of the detection results.
Open AccessEnhancing the Accuracy and Comprehensibility in Architectural Tactics Detection via Small Model-Augmented Prompt Engineering
Architectural tactics (ATs), as the concrete implementation of architectural decisions in code, address non-functional requirements of software systems. Due to the implicit nature of architectural knowledge in code implementation, developers may risk inadvertently altering or removing these tactics during code modifications or optimizations. Such unintended changes can trigger architectural erosion, gradually undermining the system's original design. While many researchers have proposed machine learning-based methods to improve the accuracy of detecting ATs in code, the black-box nature and the required architectural domain knowledge pose significant challenges for developers in verifying the results. Effective verification requires not only accurate detection results but also interpretable explanations that enhance their comprehensibility. However, this is a critical gap in current research. Large language models (LLMs) can generate easily interpretable ATs detection comments if they have domain knowledge. Fine-tuning LLMs to acquire domain knowledge faces challenges such as catastrophic forgetting and hardware constraints. Thus, we propose Prmt4TD, a small model-augmented prompting framework to enhance the accuracy and comprehensibility of ATs detection. Combining fine-tuned small models with In-Context Learning can also reduce fine-tuning costs while equipping the LLM with additional domain knowledge. Prmt4TD can leverage the remarkable processing and reasoning capabilities of LLMs to generate easily interpretable ATs detection results. Our evaluation results demonstrate that Prmt4TD achieves accuracy (\emph{F1-score}) improvement of 13\%-23\% on the ATs balanced dataset and enhances the comprehensibility of the detection results.
- 18 October 2024
- Published by Elsevier in Journal of Systems and Software
- Vol. 219
DLAP: A Deep Learning Augmented Large Language Model Prompting framework for software vulnerability detection
- 18 October 2024
- Published by Elsevier in Journal of Systems and Software
- Vol. 219
DLAP: A Deep Learning Augmented Large Language Model Prompting framework for software vulnerability detection
- 18 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
From System to Package to Interconnect: An Artificial Intelligence Powered 3D X-ray Imaging Solution for Semiconductor Package Structural Analysis and Correlative Microscopic Failure Analysis
Non-destructive 3D X-ray microscopy (XRM) has played a crucial role in fueling the advances of IC package development and failure analysis [1]-[2]. Over the past decade, the industry has increasingly focused on packaging innovations to improve device performance. The emergence of numerous new 2.5D, 3D and recent heterogenous integration packages challenges the existing X-ray imaging and analysis techniques because IC interconnects are more densely packed in larger and thicker packages. It takes several hours or longer for a 3D X-ray scanner to acquire high resolution and quality images of fine-pitch interconnects and fault regions. In this report, we will introduce a deep learning high-resolution reconstruction (DLHRR) method through the implementation of trained neutral networks capable of improving scan speed by a factor of four. To demonstrate the effectiveness of this new method applied to the packaging hierarchy, an intact smartphone, several component modules, and embedded interconnectors will be imaged and reconstructed with the DLHRR method. With the improved efficiency of the AI powered X-ray imaging technique, a correlated fs-laser/FIB-SEM workflow followed to precisely target and analyze the deeply buried defects, which has been difficult, if not impossible, for conventional package FA techniques. We will discuss the DLHRR method and applications in two following workflows: X-ray imaging workflow for package structural analysis, and correlative X-ray and fs-laser/FIB-SEM workflow for package failure analysis.
- 18 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
From System to Package to Interconnect: An Artificial Intelligence Powered 3D X-ray Imaging Solution for Semiconductor Package Structural Analysis and Correlative Microscopic Failure Analysis
Non-destructive 3D X-ray microscopy (XRM) has played a crucial role in fueling the advances of IC package development and failure analysis [1]-[2]. Over the past decade, the industry has increasingly focused on packaging innovations to improve device performance. The emergence of numerous new 2.5D, 3D and recent heterogenous integration packages challenges the existing X-ray imaging and analysis techniques because IC interconnects are more densely packed in larger and thicker packages. It takes several hours or longer for a 3D X-ray scanner to acquire high resolution and quality images of fine-pitch interconnects and fault regions. In this report, we will introduce a deep learning high-resolution reconstruction (DLHRR) method through the implementation of trained neutral networks capable of improving scan speed by a factor of four. To demonstrate the effectiveness of this new method applied to the packaging hierarchy, an intact smartphone, several component modules, and embedded interconnectors will be imaged and reconstructed with the DLHRR method. With the improved efficiency of the AI powered X-ray imaging technique, a correlated fs-laser/FIB-SEM workflow followed to precisely target and analyze the deeply buried defects, which has been difficult, if not impossible, for conventional package FA techniques. We will discuss the DLHRR method and applications in two following workflows: X-ray imaging workflow for package structural analysis, and correlative X-ray and fs-laser/FIB-SEM workflow for package failure analysis.