Junzhou Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Junzhou Zhao has a total of 82 co-authors. Here are the top co-authors by number of shared publications:
- Pinghui Wang
Xi'an Jiaotong University, Xi'an, China
30 shared publications - Xiaohong Guan
Xi'an Jiaotong University, Xi'an, China
18 shared publications - Jing Tao
Xi'an Jiaotong University, Xi'an, Shaanxi, China
7 shared publications - John Chi-Shing Lui
The Chinese University of Hong Kong (CUHK)
7 shared publications - Don Towsley
University of Massachusetts Amherst
5 shared publications Jing Tao
MOE Key Laboratory for Intelligent Networks and Network Security, Xian Jiaotong University, Xian, China
5 shared publications
Latest Publications
- 23 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 8604-8621
Flexible Privacy-Preserving Cardinality Query for Massive Distributed Datasets
Counting distinct elements (cardinality) across multiple data holders (DHs) privately is fundamental with broad applications, ranging from crowd counting to network monitoring. It is referred to as private distributed cardinality estimation (PDCE). Similarly, determining the intersection cardinality between the unions of two DH groups holds practical significance, a problem we call private distributed intersection cardinality estimation (PDICE). While many efficient methods (e.g., FM and LL sketches) exist, their differential privacy relies on secret hash functions. In PDCE, DHs must share the functions to enable sketch merging, which breaks this assumption and invalidates such guarantees. Although a recent protocol implements the FM sketch on a secret-sharing-based multiparty computation (MPC) framework for PDCE, we observe that it lacks differential privacy guarantees and is computationally expensive. To address these limitations, we propose DP-DICE-Bino, a novel protocol that is computationally efficient and differentially private for PDCE. DP-DICE-Bino is flexible in that it can compute the cardinality of any group of DHs without repeatedly interacting with the DHs. Furthermore, DP-DICE-Bino can also handle PDICE. Experiments show that DP-DICE-Bino achieves orders-of-magnitude speedups and reduces the estimation error by several times compared with the state of the art under the same security requirements.
- 23 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 8604-8621
Flexible Privacy-Preserving Cardinality Query for Massive Distributed Datasets
Counting distinct elements (cardinality) across multiple data holders (DHs) privately is fundamental with broad applications, ranging from crowd counting to network monitoring. It is referred to as private distributed cardinality estimation (PDCE). Similarly, determining the intersection cardinality between the unions of two DH groups holds practical significance, a problem we call private distributed intersection cardinality estimation (PDICE). While many efficient methods (e.g., FM and LL sketches) exist, their differential privacy relies on secret hash functions. In PDCE, DHs must share the functions to enable sketch merging, which breaks this assumption and invalidates such guarantees. Although a recent protocol implements the FM sketch on a secret-sharing-based multiparty computation (MPC) framework for PDCE, we observe that it lacks differential privacy guarantees and is computationally expensive. To address these limitations, we propose DP-DICE-Bino, a novel protocol that is computationally efficient and differentially private for PDCE. DP-DICE-Bino is flexible in that it can compute the cardinality of any group of DHs without repeatedly interacting with the DHs. Furthermore, DP-DICE-Bino can also handle PDICE. Experiments show that DP-DICE-Bino achieves orders-of-magnitude speedups and reduces the estimation error by several times compared with the state of the art under the same security requirements.
- Open Access
- 2 April 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (1) , 1-27
MTSP-LDP: A Framework for Multi-Task Streaming Data Publication under Local Differential Privacy
The proliferation of streaming data analytics in data-driven applications raises critical privacy concerns, as directly collecting user data may compromise personal privacy. Although existing w -event local differential privacy (LDP) mechanisms provide formal guarantees without relying on trusted third parties, their practical deployment is hindered by two key limitations. First, these methods are designed primarily for publishing simple statistics at each timestamp, making them inherently unsuitable for complex queries. Second, they handle data at each timestamp independently, failing to capture temporal correlations and consequently degrading the overall utility. To address these issues, we propose MTSP-LDP, a novel framework for M ulti- T ask S treaming data P ublication under w -event LDP. MTSP-LDP adopts an Optimal Privacy Budget Allocation algorithm to dynamically allocate privacy budgets by analyzing temporal correlations within each window. It then constructs a data-adaptive private binary tree structure to support complex queries, which is further refined by cross-timestamp grouping and smoothing operations to enhance estimation accuracy. Furthermore, a unified Budget-Free Multi-Task Processing mechanism is introduced to support a variety of streaming queries without consuming additional privacy budget. Extensive experiments on real-world datasets demonstrate that MTSP-LDP consistently achieves high utility across various streaming tasks, significantly outperforming existing methods.
- 2 April 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (1) , 1-27
Open AccessMTSP-LDP: A Framework for Multi-Task Streaming Data Publication under Local Differential Privacy
The proliferation of streaming data analytics in data-driven applications raises critical privacy concerns, as directly collecting user data may compromise personal privacy. Although existing w -event local differential privacy (LDP) mechanisms provide formal guarantees without relying on trusted third parties, their practical deployment is hindered by two key limitations. First, these methods are designed primarily for publishing simple statistics at each timestamp, making them inherently unsuitable for complex queries. Second, they handle data at each timestamp independently, failing to capture temporal correlations and consequently degrading the overall utility. To address these issues, we propose MTSP-LDP, a novel framework for M ulti- T ask S treaming data P ublication under w -event LDP. MTSP-LDP adopts an Optimal Privacy Budget Allocation algorithm to dynamically allocate privacy budgets by analyzing temporal correlations within each window. It then constructs a data-adaptive private binary tree structure to support complex queries, which is further refined by cross-timestamp grouping and smoothing operations to enhance estimation accuracy. Furthermore, a unified Budget-Free Multi-Task Processing mechanism is introduced to support a variety of streaming queries without consuming additional privacy budget. Extensive experiments on real-world datasets demonstrate that MTSP-LDP consistently achieves high utility across various streaming tasks, significantly outperforming existing methods.
- 5 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 21 (15566013) , 2873-2888
Extracting Training Dialogue Data From Large Language Model-Based Task Bots
Large Language Models (LLMs) have been widely adopted to enhance Task-Oriented Dialogue Systems (TODS) by modeling complex language patterns and delivering contextually appropriate responses. However, this integration introduces significant privacy risks, as LLMs, functioning as soft knowledge bases that compress extensive training data into rich knowledge representations, can inadvertently memorize training dialogue data containing not only identifiable information such as phone numbers but also entire dialogue-level events like complete travel schedules. Despite the critical nature of this privacy concern, how LLM memorization is inherited in developing task bots remains unexplored. In this work, we address this gap through a systematic quantitative study that involves evaluating existing training data extraction attacks, analyzing key characteristics of task-oriented dialogue modeling that render existing methods ineffective, and proposing novel attack techniques tailored for LLM-based TODS that enhance both response sampling and membership inference. Experimental results demonstrate the effectiveness of our proposed data extraction attack. Our method can extract thousands of training labels of dialogue states with best-case precision exceeding 70%. Furthermore, we provide an in-depth analysis of training data memorization in LLM-based TODS by identifying and quantifying key influencing factors and discussing targeted mitigation strategies.
- 5 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 21 (15566013) , 2873-2888
Extracting Training Dialogue Data From Large Language Model-Based Task Bots
Large Language Models (LLMs) have been widely adopted to enhance Task-Oriented Dialogue Systems (TODS) by modeling complex language patterns and delivering contextually appropriate responses. However, this integration introduces significant privacy risks, as LLMs, functioning as soft knowledge bases that compress extensive training data into rich knowledge representations, can inadvertently memorize training dialogue data containing not only identifiable information such as phone numbers but also entire dialogue-level events like complete travel schedules. Despite the critical nature of this privacy concern, how LLM memorization is inherited in developing task bots remains unexplored. In this work, we address this gap through a systematic quantitative study that involves evaluating existing training data extraction attacks, analyzing key characteristics of task-oriented dialogue modeling that render existing methods ineffective, and proposing novel attack techniques tailored for LLM-based TODS that enhance both response sampling and membership inference. Experimental results demonstrate the effectiveness of our proposed data extraction attack. Our method can extract thousands of training labels of dialogue states with best-case precision exceeding 70%. Furthermore, we provide an in-depth analysis of training data memorization in LLM-based TODS by identifying and quantifying key influencing factors and discussing targeted mitigation strategies.
- Open Access
FloCA: Towards Faithful and Logically Consistent Flowchart Reasoning
Flowchart-oriented dialogue (FOD) systems aim to guide users through multi-turn decision-making or operational procedures by following a domain-specific flowchart to achieve a task goal. In this work, we formalize flowchart reasoning in FOD as grounding user input to flowchart nodes at each dialogue turn while ensuring node transition is consistent with the correct flowchart path. Despite recent advances of LLMs in task-oriented dialogue systems, adapting them to FOD still faces two limitations: (1) LLMs lack an explicit mechanism to represent and reason over flowchart topology, and (2) they are prone to hallucinations, leading to unfaithful flowchart reasoning. To address these limitations, we propose FloCA, a zero-shot flowchart-oriented conversational agent. FloCA uses an LLM for intent understanding and response generation while delegating flowchart reasoning to an external tool that performs topology-constrained graph execution, ensuring faithful and logically consistent node transitions across dialogue turns. We further introduce an evaluation framework with an LLM-based user simulator and five new metrics covering reasoning accuracy and interaction efficiency. Extensive experiments on FLODIAL and PFDial datasets highlight the bottlenecks of existing LLM-based methods and demonstrate the superiority of FloCA. Our codes are available at https://github.com/Jinzi-Zou/FloCA-flowchart-reasoning.
Open AccessFloCA: Towards Faithful and Logically Consistent Flowchart Reasoning
Flowchart-oriented dialogue (FOD) systems aim to guide users through multi-turn decision-making or operational procedures by following a domain-specific flowchart to achieve a task goal. In this work, we formalize flowchart reasoning in FOD as grounding user input to flowchart nodes at each dialogue turn while ensuring node transition is consistent with the correct flowchart path. Despite recent advances of LLMs in task-oriented dialogue systems, adapting them to FOD still faces two limitations: (1) LLMs lack an explicit mechanism to represent and reason over flowchart topology, and (2) they are prone to hallucinations, leading to unfaithful flowchart reasoning. To address these limitations, we propose FloCA, a zero-shot flowchart-oriented conversational agent. FloCA uses an LLM for intent understanding and response generation while delegating flowchart reasoning to an external tool that performs topology-constrained graph execution, ensuring faithful and logically consistent node transitions across dialogue turns. We further introduce an evaluation framework with an LLM-based user simulator and five new metrics covering reasoning accuracy and interaction efficiency. Extensive experiments on FLODIAL and PFDial datasets highlight the bottlenecks of existing LLM-based methods and demonstrate the superiority of FloCA. Our codes are available at https://github.com/Jinzi-Zou/FloCA-flowchart-reasoning.
- Open Access
A Theoretical Analysis of Detecting Large Model-Generated Time Series
Motivated by the increasing risks of data misuse and fabrication, we investigate the problem of identifying synthetic time series generated by Time-Series Large Models (TSLMs) in this work. While there are extensive researches on detecting model generated text, we find that these existing methods are not applicable to time series data due to the fundamental modality difference, as time series usually have lower information density and smoother probability distributions than text data, which limit the discriminative power of token-based detectors. To address this issue, we examine the subtle distributional differences between real and model-generated time series and propose the contraction hypothesis, which states that model-generated time series, unlike real ones, exhibit progressively decreasing uncertainty under recursive forecasting. We formally prove this hypothesis under theoretical assumptions on model behavior and time series structure. Model-generated time series exhibit progressively concentrated distributions under recursive forecasting, leading to uncertainty contraction. We provide empirical validation of the hypothesis across diverse datasets. Building on this insight, we introduce the Uncertainty Contraction Estimator (UCE), a white-box detector that aggregates uncertainty metrics over successive prefixes to identify TSLM-generated time series. Extensive experiments on 32 datasets show that UCE consistently outperforms state-of-the-art baselines, offering a reliable and generalizable solution for detecting model-generated time series.
Open AccessA Theoretical Analysis of Detecting Large Model-Generated Time Series
Motivated by the increasing risks of data misuse and fabrication, we investigate the problem of identifying synthetic time series generated by Time-Series Large Models (TSLMs) in this work. While there are extensive researches on detecting model generated text, we find that these existing methods are not applicable to time series data due to the fundamental modality difference, as time series usually have lower information density and smoother probability distributions than text data, which limit the discriminative power of token-based detectors. To address this issue, we examine the subtle distributional differences between real and model-generated time series and propose the contraction hypothesis, which states that model-generated time series, unlike real ones, exhibit progressively decreasing uncertainty under recursive forecasting. We formally prove this hypothesis under theoretical assumptions on model behavior and time series structure. Model-generated time series exhibit progressively concentrated distributions under recursive forecasting, leading to uncertainty contraction. We provide empirical validation of the hypothesis across diverse datasets. Building on this insight, we introduce the Uncertainty Contraction Estimator (UCE), a white-box detector that aggregates uncertainty metrics over successive prefixes to identify TSLM-generated time series. Extensive experiments on 32 datasets show that UCE consistently outperforms state-of-the-art baselines, offering a reliable and generalizable solution for detecting model-generated time series.
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (2) , 2501-2518
MPC-FH: Privately Estimating Frequency Histogram for Advertising Measurement
An advertiser typically conducts a campaign across several publishers. Different publishers often reach different sets of persons, and some people may be exposed to the campaign via multiple publishers. Due to privacy restrictions, it is prohibited to collect all audience activities from multiple publishers in a centralized manner, therefore, we desire secure protocols for mining audience activities. In this paper, we propose a secure protocol for efficiently estimating the frequency histogram, i.e. the fraction of persons (or users) appearing a given number of times across all publishers. Our protocol MPC-FH is designed based on a novel sketch method MPS that uniformly samples a fraction of users across multiple parties and can be efficiently implemented on SPDZ, which is a popular secure multiparty computation (MPC) scheme. Our protocol provides more strict security guarantees than the state-of-the-art. In addition, our experimental results on a variety of large datasets demonstrate that our protocol can also accelerate the computation speed of existing protocols by orders of magnitudes.
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (2) , 2501-2518
MPC-FH: Privately Estimating Frequency Histogram for Advertising Measurement
An advertiser typically conducts a campaign across several publishers. Different publishers often reach different sets of persons, and some people may be exposed to the campaign via multiple publishers. Due to privacy restrictions, it is prohibited to collect all audience activities from multiple publishers in a centralized manner, therefore, we desire secure protocols for mining audience activities. In this paper, we propose a secure protocol for efficiently estimating the frequency histogram, i.e. the fraction of persons (or users) appearing a given number of times across all publishers. Our protocol MPC-FH is designed based on a novel sketch method MPS that uniformly samples a fraction of users across multiple parties and can be efficiently implemented on SPDZ, which is a popular secure multiparty computation (MPC) scheme. Our protocol provides more strict security guarantees than the state-of-the-art. In addition, our experimental results on a variety of large datasets demonstrate that our protocol can also accelerate the computation speed of existing protocols by orders of magnitudes.
- Open Access
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 2883-2892
Task Negative Sampling Enhanced Graph Few-shot Learning
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 2883-2892
Open AccessTask Negative Sampling Enhanced Graph Few-shot Learning
- Open Access
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1296-1307
PARSIFAL: Private and Robust Sign Federated Learning
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1296-1307
Open AccessPARSIFAL: Private and Robust Sign Federated Learning
- 13 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 36 (8) , 1779-1796
Building Accurate and Interpretable Online Classifiers on Edge Devices
By integrating machine learning with edge devices, we can augment the capabilities of edge devices, such as IoT devices, household appliances, and wearable technologies. These edge devices generally operate on microcontrollers with inherently limited resources, such as constrained RAM capacity and limited computational power. Nonetheless, they often process data in a high-velocity stream fashion, exemplified by sequences of activities and statuses monitored by advanced industrial sensors. In practical scenarios, models must be interpretable to facilitate troubleshooting and behavior understanding. Implementing machine learning models on edge devices is valuable and challenging, striking a balance between model efficacy and resource constraint. To address this challenge, we introduce our novel Onfesk, which combines online learning algorithms with an innovative interpretable kernel. Specifically, our Onfesk trains an online classifier over the kernel's feature sketches. Benefiting from our specially designed modules, the kernel's feature sketches can be efficiently produced, and the memory requirements of the classifier can be significantly reduced. As a result, Onfesk delivers effective and efficient performance in environments with limited resources without compromising on model interpretability. Extensive experiments with diverse real-world datasets have shown that Onfesk outperforms state-of-the-art methods, achieving up to a 7.4% improvement in accuracy within identical memory constraints.
- 13 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 36 (8) , 1779-1796
Building Accurate and Interpretable Online Classifiers on Edge Devices
By integrating machine learning with edge devices, we can augment the capabilities of edge devices, such as IoT devices, household appliances, and wearable technologies. These edge devices generally operate on microcontrollers with inherently limited resources, such as constrained RAM capacity and limited computational power. Nonetheless, they often process data in a high-velocity stream fashion, exemplified by sequences of activities and statuses monitored by advanced industrial sensors. In practical scenarios, models must be interpretable to facilitate troubleshooting and behavior understanding. Implementing machine learning models on edge devices is valuable and challenging, striking a balance between model efficacy and resource constraint. To address this challenge, we introduce our novel Onfesk, which combines online learning algorithms with an innovative interpretable kernel. Specifically, our Onfesk trains an online classifier over the kernel's feature sketches. Benefiting from our specially designed modules, the kernel's feature sketches can be efficiently produced, and the memory requirements of the classifier can be significantly reduced. As a result, Onfesk delivers effective and efficient performance in environments with limited resources without compromising on model interpretability. Extensive experiments with diverse real-world datasets have shown that Onfesk outperforms state-of-the-art methods, achieving up to a 7.4% improvement in accuracy within identical memory constraints.
- Open Access
- 9 May 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Information Systems
- Vol. 43 (4) , 1-32
How Vital Is the Jurisprudential Relevance: Law Article-Intervened Legal Case Retrieval and Matching
Legal case retrieval aims to automatically scour comparable legal cases based on a given query, which is crucial for offering relevant precedents to support the judgment in intelligent legal systems. Due to similar goals, it is often associated with a similar case matching task. To address them, a daunting challenge is assessing the uniquely defined legal-rational similarity within the judicial domain, which distinctly deviates from the semantic similarities in general text retrieval. Past works either tagged domain-specific factors or incorporated reference laws to capture legal-rational information. However, their heavy reliance on expert or unrealistic assumptions restricts their practical applicability in real-world scenarios. In this article, we propose an end-to-end model named LCM-LAI to solve the above challenges. Through meticulous theoretical analysis, LCM-LAI employs a dependent multi-task learning framework to capture legal-rational information within legal cases by a law article prediction sub-task, without any additional assumptions in inference. In addition, LCM-LAI proposes an article-aware attention mechanism to evaluate the legal-rational similarity between across-case sentences based on the law distribution, which is more effective than semantic similarity. We perform a series of exhaustive experiments that include two different tasks that involving four real-world datasets. The results demonstrate that LCM-LAI achieves state-of-the-art performance.
- 9 May 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Information Systems
- Vol. 43 (4) , 1-32
Open AccessHow Vital Is the Jurisprudential Relevance: Law Article-Intervened Legal Case Retrieval and Matching
Legal case retrieval aims to automatically scour comparable legal cases based on a given query, which is crucial for offering relevant precedents to support the judgment in intelligent legal systems. Due to similar goals, it is often associated with a similar case matching task. To address them, a daunting challenge is assessing the uniquely defined legal-rational similarity within the judicial domain, which distinctly deviates from the semantic similarities in general text retrieval. Past works either tagged domain-specific factors or incorporated reference laws to capture legal-rational information. However, their heavy reliance on expert or unrealistic assumptions restricts their practical applicability in real-world scenarios. In this article, we propose an end-to-end model named LCM-LAI to solve the above challenges. Through meticulous theoretical analysis, LCM-LAI employs a dependent multi-task learning framework to capture legal-rational information within legal cases by a law article prediction sub-task, without any additional assumptions in inference. In addition, LCM-LAI proposes an article-aware attention mechanism to evaluate the legal-rational similarity between across-case sentences based on the law distribution, which is more effective than semantic similarity. We perform a series of exhaustive experiments that include two different tasks that involving four real-world datasets. The results demonstrate that LCM-LAI achieves state-of-the-art performance.