Yanhao Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yanhao Wang has a total of 76 co-authors. Here are the top co-authors by number of shared publications:
- Yuwei Liu
Institute of Software at Chinese Academy of Sciences, Beijing, China
6 shared publications - Zhi Xue
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China
5 shared publications - Lin Huang
Ant Group, Beijing, China
3 shared publications - Minghua Wang
Ant Group, Beijing, China
3 shared publications - Purui Su
Institute of Software, Chinese Academy of Sciences, Beijing, Beijing, China
3 shared publications - Qinsheng Hou
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China
3 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23  (6) , 13100-13116
ProteusFuzz: From Static Harness Generation to Adaptive, Agent-Guided Library Fuzzing
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23  (6) , 13100-13116
ProteusFuzz: From Static Harness Generation to Adaptive, Agent-Guided Library Fuzzing
- Open Access
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 4139-4151
KVerus: Scalable and Resilient Formal Verification Proof Generation for Rust Code
- 7 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 4139-4151
Open AccessKVerus: Scalable and Resilient Formal Verification Proof Generation for Rust Code
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 3348-3359
RusyFuzz: Unhandled Exception Guided Fuzzing for Rust OS Kernel
Rust’s strong type system and ownership model eliminate many traditional memory safety issues in OS kernels. However, logic errors—such as unchecked indexing and failed unwrap operations—can still cause panic!s that crash the entire system. Existing kernel fuzzers, designed primarily for C-based kernels and reliant on KCOV, treat all crashes uniformly and fail to account for Rust-specific failure modes. We present RusyFuzz, the first fuzzing framework tailored to Rust OS kernels that explicitly targets panic-prone code paths using an unhandled-exception-guided strategy. RusyFuzz analyzes the compiler-inserted assertions within the Rust MIR, performs backward slicing to link these assertions to system call arguments, and uses constraint solving to synthesize inputs that trigger panics. In the absence of built-in coverage support, RusyFuzz employs a lightweight, log-based instrumentation method to enable coverage-guided fuzzing. We evaluate RusyFuzz on three emerging Rust-based kernels: Asterinas, Redox OS, and RuxOS. RusyFuzz discovers 70 previously unknown and developer-confirmed vulnerabilities. Compared to Trinity, it uncovers over twice as many bugs while improving line coverage by 14.4%. These results demonstrate that unhandled-exception-guided fuzzing is critical for uncovering logic bugs and enhancing the reliability of Rust OS kernels, providing the first systematic methodology for detecting such vulnerabilities.
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 3348-3359
Open AccessRusyFuzz: Unhandled Exception Guided Fuzzing for Rust OS Kernel
Rust’s strong type system and ownership model eliminate many traditional memory safety issues in OS kernels. However, logic errors—such as unchecked indexing and failed unwrap operations—can still cause panic!s that crash the entire system. Existing kernel fuzzers, designed primarily for C-based kernels and reliant on KCOV, treat all crashes uniformly and fail to account for Rust-specific failure modes. We present RusyFuzz, the first fuzzing framework tailored to Rust OS kernels that explicitly targets panic-prone code paths using an unhandled-exception-guided strategy. RusyFuzz analyzes the compiler-inserted assertions within the Rust MIR, performs backward slicing to link these assertions to system call arguments, and uses constraint solving to synthesize inputs that trigger panics. In the absence of built-in coverage support, RusyFuzz employs a lightweight, log-based instrumentation method to enable coverage-guided fuzzing. We evaluate RusyFuzz on three emerging Rust-based kernels: Asterinas, Redox OS, and RuxOS. RusyFuzz discovers 70 previously unknown and developer-confirmed vulnerabilities. Compared to Trinity, it uncovers over twice as many bugs while improving line coverage by 14.4%. These results demonstrate that unhandled-exception-guided fuzzing is critical for uncovering logic bugs and enhancing the reliability of Rust OS kernels, providing the first systematic methodology for detecting such vulnerabilities.
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23  (4) , 7864-7877
Learning Flow Semantics for Encrypted Traffic Analysis: A Contrastive Pre-Training Approach
Encrypted traffic analysis is crucial for cyberspace security. Self-supervised learning shows great promise to enhance traffic analysis with the pre-trained traffic encoder, which is constructed using large-scale, readily available unlabeled traffic data. However, existing approaches struggle to handle the increasingly prevalent encrypted traffic, as their generative reconstruction tasks cannot process encrypted content. To this end, we propose TACO, a robust and flexible encrypted traffic analysis system based on flow semantics learning. Specifically, we first design several feasible traffic data augmentation strategies to prepare flow semantics knowledge from the unlabeled traffic. Then, our traffic encoder with a traffic partition module learns the semantics knowledge based on the contrastive pre-training paradigm. It serves as a traffic foundation encoder that can comprehend flow semantics and extract effective semantic representations. Finally, we fine-tune the traffic encoder to leverage flow semantics for various downstream encrypted traffic analysis tasks. The experimental results illustrate that TACO outperforms the optimal baseline by 7.5% in average F1 score on four traffic classification datasets and achieves an improvement of at least 11.62% in average F1 score on the three transfer tasks, while indicating superior efficiency. We will release the source code as well as the experiment data upon publication to foster future research.
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23  (4) , 7864-7877
Learning Flow Semantics for Encrypted Traffic Analysis: A Contrastive Pre-Training Approach
Encrypted traffic analysis is crucial for cyberspace security. Self-supervised learning shows great promise to enhance traffic analysis with the pre-trained traffic encoder, which is constructed using large-scale, readily available unlabeled traffic data. However, existing approaches struggle to handle the increasingly prevalent encrypted traffic, as their generative reconstruction tasks cannot process encrypted content. To this end, we propose TACO, a robust and flexible encrypted traffic analysis system based on flow semantics learning. Specifically, we first design several feasible traffic data augmentation strategies to prepare flow semantics knowledge from the unlabeled traffic. Then, our traffic encoder with a traffic partition module learns the semantics knowledge based on the contrastive pre-training paradigm. It serves as a traffic foundation encoder that can comprehend flow semantics and extract effective semantic representations. Finally, we fine-tune the traffic encoder to leverage flow semantics for various downstream encrypted traffic analysis tasks. The experimental results illustrate that TACO outperforms the optimal baseline by 7.5% in average F1 score on four traffic classification datasets and achieves an improvement of at least 11.62% in average F1 score on the three transfer tasks, while indicating superior efficiency. We will release the source code as well as the experiment data upon publication to foster future research.
- Open Access
- 9 March 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9  (1) , 133
OSmartPro: a large language model-assisted option fuzzing approach
Program options provide flexible software functionality control but complicate fuzz testing, as triggering many behaviors require specific option combinations. Although existing option-aware fuzzing approaches attempt to mutate options as inputs or leverage AI technologies to extract option relationships from documentation, these methods have limitations. Documentation is often incomplete, and some option dependencies are embedded deeply within program logic via data or control flows, making these methods challenging to detect all possible dependencies. This paper introduces OSmartPro , an advanced option-fuzzing approach that directly extracts options and infers option dependencies from source code. Given LLM’s capabilities to interpret program semantics, OSmartPro employs LLM-assisted static analysis to handle diverse option-parsing structures and extract comprehensive options. Through control and data dependency analysis, it constructs option impact graph , which it uses to guide fuzzing strategies. The tool successfully extracted complete options from all 59 programs in our test set, uncovering undocumented options in over 66% of them. Additionally, OSmartPro inferred 14,701 option combinations, identified 45.03% more execution paths compared to AFL++, and uncovered 54 zero-day vulnerabilities, of which 18 awarded CVE IDs. Lastly, in a benchmark comparison against four option-aware fuzzers, OSmartPro achieved higher line coverage in 66.7% (20 out of 30) of the programs.
- 9 March 2026
- Published by Springer Nature in Cybersecurity
- Vol. 9  (1) , 133
Open AccessOSmartPro: a large language model-assisted option fuzzing approach
Program options provide flexible software functionality control but complicate fuzz testing, as triggering many behaviors require specific option combinations. Although existing option-aware fuzzing approaches attempt to mutate options as inputs or leverage AI technologies to extract option relationships from documentation, these methods have limitations. Documentation is often incomplete, and some option dependencies are embedded deeply within program logic via data or control flows, making these methods challenging to detect all possible dependencies. This paper introduces OSmartPro , an advanced option-fuzzing approach that directly extracts options and infers option dependencies from source code. Given LLM’s capabilities to interpret program semantics, OSmartPro employs LLM-assisted static analysis to handle diverse option-parsing structures and extract comprehensive options. Through control and data dependency analysis, it constructs option impact graph , which it uses to guide fuzzing strategies. The tool successfully extracted complete options from all 59 programs in our test set, uncovering undocumented options in over 66% of them. Additionally, OSmartPro inferred 14,701 option combinations, identified 45.03% more execution paths compared to AFL++, and uncovered 54 zero-day vulnerabilities, of which 18 awarded CVE IDs. Lastly, in a benchmark comparison against four option-aware fuzzers, OSmartPro achieved higher line coverage in 66.7% (20 out of 30) of the programs.
- Open Access
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1559-1573
P rome F uzz : A Knowledge-Driven Approach to Fuzzing Harness Generation with Large Language Models
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1559-1573
Open AccessP rome F uzz : A Knowledge-Driven Approach to Fuzzing Harness Generation with Large Language Models
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 51  (12) , 3467-3485
Enhancing Real-Time Operating System Security Analysis via Slice-Based Fuzzing
Real-Time Operating System (RTOS) has become the main category of embedded systems. It is widely used to support tasks requiring real-time response such as printers and switches. The security of RTOS has been long overlooked as it was running in special environments isolated from attackers. However, with the rapid development of IoT devices, tremendous RTOS devices are connected to the public network. Due to the lack of security mechanisms, these devices are extremely vulnerable to a wide spectrum of attacks. Even worse, the monolithic design of RTOS combines various tasks and services into a single binary, which hinders the current program testing and analysis techniques working on RTOS. In this paper, we propose SFUZZ++, a novel slice-based fuzzer designed to detect security vulnerabilities in RTOS. Leveraging the insight that RTOS tasks are typically independent, single-purpose, and deterministic, SFUZZ++ extracts task-specific code slices for targeted testing. Specifically, SFUZZ++ first identifies external input points that manage user input, with assistance from LLMs, and constructs call graphs from these points. Then, it leverages forward slicing to build the sensitive call graph and prune the paths independent of sink points (e.g., memcpy). Further, it detects and handles roadblocks within the coarse-grain scope that hinder effective fuzzing, such as call sites unrelated to the user input or conditional branches unrelated to sink points. At the same time, it attempts to restore the context of the slices to recreate the actual runtime state. And then, it conducts coverage-guided fuzzing on these code snippets. Finally, SFUZZ++ leverages forward and backward slicing to track and verify each path constraint and determine whether a bug discovered in the fuzzer is a real vulnerability. SFUZZ++ successfully discovered 82 zero-day bugs on 35 RTOS samples, and 78 of them have been assigned CVE or CNVD IDs. Our empirical evaluation shows that SFUZZ++ outperforms the state-of- the-art tools (e.g., UnicornAFL) on testing RTOS.
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Software Engineering
- Vol. 51  (12) , 3467-3485
Enhancing Real-Time Operating System Security Analysis via Slice-Based Fuzzing
Real-Time Operating System (RTOS) has become the main category of embedded systems. It is widely used to support tasks requiring real-time response such as printers and switches. The security of RTOS has been long overlooked as it was running in special environments isolated from attackers. However, with the rapid development of IoT devices, tremendous RTOS devices are connected to the public network. Due to the lack of security mechanisms, these devices are extremely vulnerable to a wide spectrum of attacks. Even worse, the monolithic design of RTOS combines various tasks and services into a single binary, which hinders the current program testing and analysis techniques working on RTOS. In this paper, we propose SFUZZ++, a novel slice-based fuzzer designed to detect security vulnerabilities in RTOS. Leveraging the insight that RTOS tasks are typically independent, single-purpose, and deterministic, SFUZZ++ extracts task-specific code slices for targeted testing. Specifically, SFUZZ++ first identifies external input points that manage user input, with assistance from LLMs, and constructs call graphs from these points. Then, it leverages forward slicing to build the sensitive call graph and prune the paths independent of sink points (e.g., memcpy). Further, it detects and handles roadblocks within the coarse-grain scope that hinder effective fuzzing, such as call sites unrelated to the user input or conditional branches unrelated to sink points. At the same time, it attempts to restore the context of the slices to recreate the actual runtime state. And then, it conducts coverage-guided fuzzing on these code snippets. Finally, SFUZZ++ leverages forward and backward slicing to track and verify each path constraint and determine whether a bug discovered in the fuzzer is a real vulnerability. SFUZZ++ successfully discovered 82 zero-day bugs on 35 RTOS samples, and 78 of them have been assigned CVE or CNVD IDs. Our empirical evaluation shows that SFUZZ++ outperforms the state-of- the-art tools (e.g., UnicornAFL) on testing RTOS.
- 28 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12  (12) , 21364-21378
MODFuzz: A Multiobjective Directed Fuzzer for USB Drivers
USB interfaces have become ubiquitous in various Internet of Things (IoT) devices, all adhering to the same USB protocol. While enhancing convenience, they also widen the potential attack surface. Fuzzing is a proactive way to identify potential security threats for USB drivers. However, existing USB driver fuzzers primarily prioritize the code coverage of USB drivers, leading to a significant waste of computational resources on irrelevant code segments. To this end, we combine directed fuzzing and USB driver fuzzing for the first time, and present MODFuzz, a pioneering multi-objective directed fuzzing method for USB drivers. MODFuzz autonomously locates the most vulnerable parts within USB drivers, concentrating fuzzing efforts on these areas. Diverging from the existing directed fuzzers, MODFuzz employs a dynamic direction instead of predetermined addresses to guide the fuzzing campaign toward the triggered execution traces with a greater probability of containing vulnerabilities. MODFuzz outperforms the strong baseline in terms of execution speed (about 14% improvement) and crash generation capabilities (about 69% improvement). Meanwhile, we found six previously unknown bugs (all confirmed and assigned vulnerability IDs) in Linux kernel v6.4.10 and received acknowledgment from Red Hat.
- 28 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12  (12) , 21364-21378
MODFuzz: A Multiobjective Directed Fuzzer for USB Drivers
USB interfaces have become ubiquitous in various Internet of Things (IoT) devices, all adhering to the same USB protocol. While enhancing convenience, they also widen the potential attack surface. Fuzzing is a proactive way to identify potential security threats for USB drivers. However, existing USB driver fuzzers primarily prioritize the code coverage of USB drivers, leading to a significant waste of computational resources on irrelevant code segments. To this end, we combine directed fuzzing and USB driver fuzzing for the first time, and present MODFuzz, a pioneering multi-objective directed fuzzing method for USB drivers. MODFuzz autonomously locates the most vulnerable parts within USB drivers, concentrating fuzzing efforts on these areas. Diverging from the existing directed fuzzers, MODFuzz employs a dynamic direction instead of predetermined addresses to guide the fuzzing campaign toward the triggered execution traces with a greater probability of containing vulnerabilities. MODFuzz outperforms the strong baseline in terms of execution speed (about 14% improvement) and crash generation capabilities (about 69% improvement). Meanwhile, we found six previously unknown bugs (all confirmed and assigned vulnerability IDs) in Linux kernel v6.4.10 and received acknowledgment from Red Hat.
- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11  (15) , 25910-25924
A Semantics-Based Approach on Binary Function Similarity Detection
As a fundamental component of Internet of Things (IoT) devices, firmware plays an essential role. Nowadays, the development of IoT firmware relies extensively on third-party components and substantially enhances development efficiency. However, these components are not inherently secure, and their vulnerabilities can adversely affect the security of IoT firmware. Existing research adopts binary code similarity analysis to detect known vulnerabilities in firmware. However, it encounters significant challenges, primarily in extracting function features from the limited semantic information within binary code. Another challenge is the need for real-world datasets to assess the model’s performance in practical scenarios, such as firmware supply chain analysis. We present a detection model named PDG2VEC based on Program Dependence Graphs (PDGs) to tackle these challenges. PDG2VEC extracts function features at the variable level on PDG and assesses function similarity by evaluating whether two functions can represent each other. We conducted evaluations using three datasets, including one we created to simulate a firmware supply chain scenario. The experimental results demonstrate that PDG2VEC exhibits resilience to cross-architecture challenges and captures more precise semantics than other approaches. Furthermore, PDG2VEC outperforms state-of-the-art tools in the supply chain analysis scenario, with a 16% higher AUC value average against baseline approaches.
- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11  (15) , 25910-25924
A Semantics-Based Approach on Binary Function Similarity Detection
As a fundamental component of Internet of Things (IoT) devices, firmware plays an essential role. Nowadays, the development of IoT firmware relies extensively on third-party components and substantially enhances development efficiency. However, these components are not inherently secure, and their vulnerabilities can adversely affect the security of IoT firmware. Existing research adopts binary code similarity analysis to detect known vulnerabilities in firmware. However, it encounters significant challenges, primarily in extracting function features from the limited semantic information within binary code. Another challenge is the need for real-world datasets to assess the model’s performance in practical scenarios, such as firmware supply chain analysis. We present a detection model named PDG2VEC based on Program Dependence Graphs (PDGs) to tackle these challenges. PDG2VEC extracts function features at the variable level on PDG and assesses function similarity by evaluating whether two functions can represent each other. We conducted evaluations using three datasets, including one we created to simulate a firmware supply chain scenario. The experimental results demonstrate that PDG2VEC exhibits resilience to cross-architecture challenges and captures more precise semantics than other approaches. Furthermore, PDG2VEC outperforms state-of-the-art tools in the supply chain analysis scenario, with a 16% higher AUC value average against baseline approaches.
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 32  (3) , 2012-2025
A Novel Self-Supervised Framework Based on Masked Autoencoder for Traffic Classification
Traffic classification is a critical task in network security and management. Recent research has demonstrated the effectiveness of the deep learning-based traffic classification method. However, the following limitations remain: (1) the traffic representation is simply generated from raw packet bytes, resulting in the absence of important information; (2) the model structure of directly applying deep learning algorithms does not take traffic characteristics into account; and (3) scenario-specific classifier training usually requires a labor-intensive and time-consuming process to label data. In this paper, we introduce a masked autoencoder (MAE) based traffic transformer with multi-level flow representation to tackle these problems. To model raw traffic data, we design a formatted traffic representation matrix with hierarchical flow information. After that, we develop an efficient Traffic Transformer, in which packet-level and flow-level attention mechanisms implement more efficient feature extraction with lower complexity. At last, we utilize MAE paradigm to pre-train our classifier with a large amount of unlabeled data, and perform fine-tuning with a few labeled data for a series of traffic classification tasks. Experiment findings reveal that our method outperforms state-of-the-art methods on five real-world traffic datasets by a large margin. The code is available at https://github.com/NSSL-SJTU/YaTC.
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 32  (3) , 2012-2025
A Novel Self-Supervised Framework Based on Masked Autoencoder for Traffic Classification
Traffic classification is a critical task in network security and management. Recent research has demonstrated the effectiveness of the deep learning-based traffic classification method. However, the following limitations remain: (1) the traffic representation is simply generated from raw packet bytes, resulting in the absence of important information; (2) the model structure of directly applying deep learning algorithms does not take traffic characteristics into account; and (3) scenario-specific classifier training usually requires a labor-intensive and time-consuming process to label data. In this paper, we introduce a masked autoencoder (MAE) based traffic transformer with multi-level flow representation to tackle these problems. To model raw traffic data, we design a formatted traffic representation matrix with hierarchical flow information. After that, we develop an efficient Traffic Transformer, in which packet-level and flow-level attention mechanisms implement more efficient feature extraction with lower complexity. At last, we utilize MAE paradigm to pre-train our classifier with a large amount of unlabeled data, and perform fine-tuning with a few labeled data for a series of traffic classification tasks. Experiment findings reveal that our method outperforms state-of-the-art methods on five real-world traffic datasets by a large margin. The code is available at https://github.com/NSSL-SJTU/YaTC.