Tao Wei
Research Subjects
Tao Wei has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Chao Zhang
Tsinghua University, Beijing, China and Zhongguancun Laboratory, Beijing, China
1 shared publication - Cheng Hong
Ant Group
1 shared publication - Derun Zhao
Ant Group Ant Group Ant Group Ant Group Ant Group
1 shared publication - Han Qiu
Tsinghua University, Beijing, China
1 shared publication - Haoqi Wu1 shared publication
- Hengqi Guo
Fudan University
1 shared publication
Latest Publications
- Open Access
Prefix Probing: Lightweight Harmful Content Detection for Large Language Models
Large language models often face a three-way trade-off among detection accuracy, inference latency, and deployment cost when used in real-world safety-sensitive applications. This paper introduces Prefix Probing, a black-box harmful content detection method that compares the conditional log-probabilities of "agreement/execution" versus "refusal/safety" opening prefixes and leverages prefix caching to reduce detection overhead to near first-token latency. During inference, the method requires only a single log-probability computation over the probe prefixes to produce a harmfulness score and apply a threshold, without invoking any additional models or multi-stage inference. To further enhance the discriminative power of the prefixes, we design an efficient prefix construction algorithm that automatically discovers highly informative prefixes, substantially improving detection performance. Extensive experiments demonstrate that Prefix Probing achieves detection effectiveness comparable to mainstream external safety models while incurring only minimal computational cost and requiring no extra model deployment, highlighting its strong practicality and efficiency.
Open AccessPrefix Probing: Lightweight Harmful Content Detection for Large Language Models
Large language models often face a three-way trade-off among detection accuracy, inference latency, and deployment cost when used in real-world safety-sensitive applications. This paper introduces Prefix Probing, a black-box harmful content detection method that compares the conditional log-probabilities of "agreement/execution" versus "refusal/safety" opening prefixes and leverages prefix caching to reduce detection overhead to near first-token latency. During inference, the method requires only a single log-probability computation over the probe prefixes to produce a harmfulness score and apply a threshold, without invoking any additional models or multi-stage inference. To further enhance the discriminative power of the prefixes, we design an efficient prefix construction algorithm that automatically discovers highly informative prefixes, substantially improving detection performance. Extensive experiments demonstrate that Prefix Probing achieves detection effectiveness comparable to mainstream external safety models while incurring only minimal computational cost and requiring no extra model deployment, highlighting its strong practicality and efficiency.
- Open Access
- 13 September 2024
- Published by Springer Nature
- p. 307-324
UnsafeCop: Towards Memory Safety for Real-World Unsafe Rust Code with Practical Bounded Model Checking
Rust has gained popularity as a safer alternative to C/C++ for low-level programming due to its memory-safety features and minimal runtime overhead. However, the use of the “unsafe” keyword allows developers to bypass safety guarantees, posing memory-safety risks. Bounded Model Checking (BMC) is commonly used to detect memory-safety problems, but it has limitations for large-scale programs, as it can only detect bugs within a bounded number of executions.In this paper, we introduce UnsafeCop that utilizes and enhances BMC for analyzing memory safety in real-world unsafe Rust code. Our methodology incorporates harness design, loop bound inference, and both loop and function stubbing for comprehensive analysis. We optimize verification efficiency through a strategic function verification order, leveraging both types of stubbing. We conducted a case study on TECC (Trusted-Environment-based Cryptographic Computing), a proprietary framework consisting of 30,174 lines of Rust code, including 3,019 lines of unsafe Rust code, developed by Ant Group. Experimental results demonstrate that UnsafeCop effectively detects and verifies dozens of memory safety issues, reducing verification time by 73.71% compared to the traditional non-stubbing approach, highlighting its practical effectiveness.
- 13 September 2024
- Published by Springer Nature
- p. 307-324
Open AccessUnsafeCop: Towards Memory Safety for Real-World Unsafe Rust Code with Practical Bounded Model Checking
Rust has gained popularity as a safer alternative to C/C++ for low-level programming due to its memory-safety features and minimal runtime overhead. However, the use of the “unsafe” keyword allows developers to bypass safety guarantees, posing memory-safety risks. Bounded Model Checking (BMC) is commonly used to detect memory-safety problems, but it has limitations for large-scale programs, as it can only detect bugs within a bounded number of executions.In this paper, we introduce UnsafeCop that utilizes and enhances BMC for analyzing memory safety in real-world unsafe Rust code. Our methodology incorporates harness design, loop bound inference, and both loop and function stubbing for comprehensive analysis. We optimize verification efficiency through a strategic function verification order, leveraging both types of stubbing. We conducted a case study on TECC (Trusted-Environment-based Cryptographic Computing), a proprietary framework consisting of 30,174 lines of Rust code, including 3,019 lines of unsafe Rust code, developed by Ant Group. Experimental results demonstrate that UnsafeCop effectively detects and verifies dozens of memory safety issues, reducing verification time by 73.71% compared to the traditional non-stubbing approach, highlighting its practical effectiveness.
- Open Access
PUMA: Secure Inference of LLaMA-7B in Five Minutes
With ChatGPT as a representative, tons of companies have began to provide services based on large Transformers models. However, using such a service inevitably leak users' prompts to the model provider. Previous studies have studied secure inference for Transformer models using secure multiparty computation (MPC), where model parameters and clients' prompts are kept secret. Despite this, these frameworks are still limited in terms of model performance, efficiency, and deployment. To address these limitations, we propose framework PUMA to enable fast and secure Transformer model inference. Our framework designs high quality approximations for expensive functions such as GeLU and softmax, and significantly reduce the cost of secure inference while preserving the model performance. Additionally, we design secure Embedding and LayerNorm procedures that faithfully implement the desired functionality without undermining the Transformer architecture. PUMA is about $2\times$ faster than the state-of-the-art framework MPCFORMER(ICLR 2023) and has similar accuracy as plaintext models without fine-tuning (which the previous works failed to achieve). PUMA can even evaluate LLaMA-7B in around 5 minutes to generate 1 token. To our best knowledge, this is the first time that a model with such a parameter size is able to be evaluated under MPC. PUMA has been open-sourced in the Github repository of SecretFlow-SPU.Open AccessPUMA: Secure Inference of LLaMA-7B in Five Minutes
With ChatGPT as a representative, tons of companies have began to provide services based on large Transformers models. However, using such a service inevitably leak users' prompts to the model provider. Previous studies have studied secure inference for Transformer models using secure multiparty computation (MPC), where model parameters and clients' prompts are kept secret. Despite this, these frameworks are still limited in terms of model performance, efficiency, and deployment. To address these limitations, we propose framework PUMA to enable fast and secure Transformer model inference. Our framework designs high quality approximations for expensive functions such as GeLU and softmax, and significantly reduce the cost of secure inference while preserving the model performance. Additionally, we design secure Embedding and LayerNorm procedures that faithfully implement the desired functionality without undermining the Transformer architecture. PUMA is about $2\times$ faster than the state-of-the-art framework MPCFORMER(ICLR 2023) and has similar accuracy as plaintext models without fine-tuning (which the previous works failed to achieve). PUMA can even evaluate LLaMA-7B in around 5 minutes to generate 1 token. To our best knowledge, this is the first time that a model with such a parameter size is able to be evaluated under MPC. PUMA has been open-sourced in the Github repository of SecretFlow-SPU. - Open Access
Black-box Dataset Ownership Verification via Backdoor Watermarking
Deep learning, especially deep neural networks (DNNs), has been widely and successfully adopted in many critical applications for its high effectiveness and efficiency. The rapid development of DNNs has benefited from the existence of some high-quality datasets ($e.g.$, ImageNet), which allow researchers and developers to easily verify the performance of their methods. Currently, almost all existing released datasets require that they can only be adopted for academic or educational purposes rather than commercial purposes without permission. However, there is still no good way to ensure that. In this paper, we formulate the protection of released datasets as verifying whether they are adopted for training a (suspicious) third-party model, where defenders can only query the model while having no information about its parameters and training details. Based on this formulation, we propose to embed external patterns via backdoor watermarking for the ownership verification to protect them. Our method contains two main parts, including dataset watermarking and dataset verification. Specifically, we exploit poison-only backdoor attacks ($e.g.$, BadNets) for dataset watermarking and design a hypothesis-test-guided method for dataset verification. We also provide some theoretical analyses of our methods. Experiments on multiple benchmark datasets of different tasks are conducted, which verify the effectiveness of our method. The code for reproducing main experiments is available at \url{https://github.com/THUYimingLi/DVBW}.
Open AccessBlack-box Dataset Ownership Verification via Backdoor Watermarking
Deep learning, especially deep neural networks (DNNs), has been widely and successfully adopted in many critical applications for its high effectiveness and efficiency. The rapid development of DNNs has benefited from the existence of some high-quality datasets ($e.g.$, ImageNet), which allow researchers and developers to easily verify the performance of their methods. Currently, almost all existing released datasets require that they can only be adopted for academic or educational purposes rather than commercial purposes without permission. However, there is still no good way to ensure that. In this paper, we formulate the protection of released datasets as verifying whether they are adopted for training a (suspicious) third-party model, where defenders can only query the model while having no information about its parameters and training details. Based on this formulation, we propose to embed external patterns via backdoor watermarking for the ownership verification to protect them. Our method contains two main parts, including dataset watermarking and dataset verification. Specifically, we exploit poison-only backdoor attacks ($e.g.$, BadNets) for dataset watermarking and design a hypothesis-test-guided method for dataset verification. We also provide some theoretical analyses of our methods. Experiments on multiple benchmark datasets of different tasks are conducted, which verify the effectiveness of our method. The code for reproducing main experiments is available at \url{https://github.com/THUYimingLi/DVBW}.
- Open Access
Aegis: Mitigating Targeted Bit-flip Attacks against Deep Neural Networks
Bit-flip attacks (BFAs) have attracted substantial attention recently, in which an adversary could tamper with a small number of model parameter bits to break the integrity of DNNs. To mitigate such threats, a batch of defense methods are proposed, focusing on the untargeted scenarios. Unfortunately, they either require extra trustworthy applications or make models more vulnerable to targeted BFAs. Countermeasures against targeted BFAs, stealthier and more purposeful by nature, are far from well established. In this work, we propose Aegis, a novel defense method to mitigate targeted BFAs. The core observation is that existing targeted attacks focus on flipping critical bits in certain important layers. Thus, we design a dynamic-exit mechanism to attach extra internal classifiers (ICs) to hidden layers. This mechanism enables input samples to early-exit from different layers, which effectively upsets the adversary's attack plans. Moreover, the dynamic-exit mechanism randomly selects ICs for predictions during each inference to significantly increase the attack cost for the adaptive attacks where all defense mechanisms are transparent to the adversary. We further propose a robustness training strategy to adapt ICs to the attack scenarios by simulating BFAs during the IC training phase, to increase model robustness. Extensive evaluations over four well-known datasets and two popular DNN structures reveal that Aegis could effectively mitigate different state-of-the-art targeted attacks, reducing attack success rate by 5-10$\times$, significantly outperforming existing defense methods.
Open AccessAegis: Mitigating Targeted Bit-flip Attacks against Deep Neural Networks
Bit-flip attacks (BFAs) have attracted substantial attention recently, in which an adversary could tamper with a small number of model parameter bits to break the integrity of DNNs. To mitigate such threats, a batch of defense methods are proposed, focusing on the untargeted scenarios. Unfortunately, they either require extra trustworthy applications or make models more vulnerable to targeted BFAs. Countermeasures against targeted BFAs, stealthier and more purposeful by nature, are far from well established. In this work, we propose Aegis, a novel defense method to mitigate targeted BFAs. The core observation is that existing targeted attacks focus on flipping critical bits in certain important layers. Thus, we design a dynamic-exit mechanism to attach extra internal classifiers (ICs) to hidden layers. This mechanism enables input samples to early-exit from different layers, which effectively upsets the adversary's attack plans. Moreover, the dynamic-exit mechanism randomly selects ICs for predictions during each inference to significantly increase the attack cost for the adaptive attacks where all defense mechanisms are transparent to the adversary. We further propose a robustness training strategy to adapt ICs to the attack scenarios by simulating BFAs during the IC training phase, to increase model robustness. Extensive evaluations over four well-known datasets and two popular DNN structures reveal that Aegis could effectively mitigate different state-of-the-art targeted attacks, reducing attack success rate by 5-10$\times$, significantly outperforming existing defense methods.
- 10 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (1) , 377-391
CVTEE: A Compatible Verified TEE Architecture With Enhanced Security
Sensitive resources in Trusted Execution Environment (TEE) have suffered serious security threats in recent years. Previous protection approaches either lack a strong assurance of TEE security properties or are limited to a single platform. We propose a compatible verified TEE architecture, called CVTEE, which delegates a security monitor to manage TEE resources securely. This architecture has two key advantages: i) its functional correctness and security are guaranteed by a machine-checkable proof of security objectives of Trusted Application (TA) isolation, runtime confidentiality, and runtime integrity, and ii) it is applicable to different TEE platforms and implementation-independent due to its high level of abstraction and non-determinism of data types. Note that access control policy and information flow control policy are the core for security management of resources. After formally specifying the security attributes of TEE resources, we develop these policies based on CC in the security monitor and provide atomic interfaces. CVTEE is formally verified with 386 lemmas/theorems and ~ 10,000 LOC of Isabelle/HOL. In addition, we implement a proof of concept for the access control module of Teaclave, and prove that the constructed access control model meets the security requirements through 5 theorems.
- 10 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (1) , 377-391
CVTEE: A Compatible Verified TEE Architecture With Enhanced Security
Sensitive resources in Trusted Execution Environment (TEE) have suffered serious security threats in recent years. Previous protection approaches either lack a strong assurance of TEE security properties or are limited to a single platform. We propose a compatible verified TEE architecture, called CVTEE, which delegates a security monitor to manage TEE resources securely. This architecture has two key advantages: i) its functional correctness and security are guaranteed by a machine-checkable proof of security objectives of Trusted Application (TA) isolation, runtime confidentiality, and runtime integrity, and ii) it is applicable to different TEE platforms and implementation-independent due to its high level of abstraction and non-determinism of data types. Note that access control policy and information flow control policy are the core for security management of resources. After formally specifying the security attributes of TEE resources, we develop these policies based on CC in the security monitor and provide atomic interfaces. CVTEE is formally verified with 386 lemmas/theorems and ~ 10,000 LOC of Isabelle/HOL. In addition, we implement a proof of concept for the access control module of Teaclave, and prove that the constructed access control model meets the security requirements through 5 theorems.