Zhuotao Liu · InspiringGroup
News
Research updates, awards, talks, and academic service.
2026
Two papers are accepted by NDSS 2027. One paper addresses the incremental learning problem in network traffic analysis, and the other paper addresses secure LLM inferences via function secret sharing.
Our work on trust-native AI inference infrastructure using legacy GPUs was accepted to ACM CCS 2026.
Our work on meta-learning for continual traffic analysis was accepted to USENIX Security 2026.
2025
RingSG received a Distinguished Paper Award at ACM CCS 2025.
Our paper on confidential/private inference over Transformer-based large models via novel cryptography constructions is accepted by IEEE S&P 2026.
Another three papers on RL-driven traffic analysis, website fingerprinting, and web application firewall are accepted by ISOC NDSS 2026.
Our paper on understanding BGP hijacking attacks given partially deployed ROV is accepted by ISOC NDSS 2026.
Another edition of USENIX Security. Looking forward to your submissions.
Our paper on addressing the robustness of AI-driven traffic analysis in dynamic networking conditions is accepted by ACM CCS 2025.
Our paper on certified robustness of AI-driven traffic analysis is accepted by USENIX Security 2025. This work features the most advanced provable robustness of AI models, in the domain of network traffic analysis!
Another edition of IEEE S&P. Looking forward to your submissions.
Our paper on scaling neural network inference on the intelligent network dataplane is accepted by ACM SIGCOMM 2025. The era of Network AI Agent arrives!
Our paper (RingSG) on optimal secure vertex-centric computation for collaborative graph processing is accepted by ACM CCS 2025. This work features the optimal communication and computation complexity within the MPC-based vertex-centric abstraction for collaborative graph processing.
Our paper (MM4flow) on constructing multi-modal pre-trained large models for traffic analysis is accepted by ACM CCS 2025.
Another year of CCS cycle. Your contribution is welcome.
2024
Our paper on pre-training transformer models for analyzing network-traffic data is accepted by IEEE S&P 2025.
Our paper on Secure Internet Routing won USENIX Security 2024 Distinguished Paper Award and the Internet Defense Prize, simultaneously!
Our paper on rearchitecting the Random Early Detection for high performance transport in DCN is accepted by NSDI 2025.
Our paper on performing fine-grained and large-scale webpage fingerprinting via encrypted traffic analysis is accepted by ACM CCS 2024.
Our paper on protecting LLM models by identifying key parameters is accepted by ICML 2024 NextGenAISafety workshop.
Invited to serve on the TPC of IEEE S&P 2025. Please consider to submit.
CoGNN is awarded all three badges from ACM CCS 2024 artifact evaluation committee.
Our paper (CoGNN) on training graph neural networks (GNNs) over distributed and private graph data across multiple data providers is accepted by ACM CCS 2024. This work features multiple novel crypto constructions to realize fully-distributed and scalable GNN training/inference over distributed private graph data.
Our paper on understanding the data privacy in FL is accepted by USENIX Security 2024.
Our paper on detecting and interpreting BGP anomalies at scale is accepted by USENIX Security 2024.
2023
Invited to serve on the TPC of ACM CCS 2024 (Machine Learning Security Track), Usenix Security 2024, SIGMETRICS 2024 and IEEE ICDCS 2024. Please consider to submit.
Our paper on intelligent network dataplane (INDP) is accepted by USENIX NSDI 2024, This work features the most advanced IDNP design that enables neural network driven traffic analysis at line-speed.
Our paper on privacy-preserving machine learning is accepted by NDSS 2024. This work features multiple novel cryptography constructions to accelerate AI training / inference over encrypted data.
Our paper on zero-day Web attack detection is accepted by ACM CCS 2023.
Our paper on building a utility-driven data marketplace is accepted by ACM CCS 2023. This work features a robust and verifiable Federated Learning architecture, underpinned by a secure model evaluation protocol to remove malicious model updates and a zero-knowledge proving system that ensures fair data trading.
Our paper on Intelligent Network Data Plane wins the USENIX Security 2023 Distinguished Paper Award!
One paper on encrypted traffic analysis is accepted by NDSS 2024. This work handles the challenging problem of obtaining robust models with low-quality training data.
One paper on container security is accepted by USENIX Security 2023. This work reveals several key security issues for Cloud containers.
One paper on transforming the legal sector towards Web3.0-friendly by (semi-)automatically representing legal agreements as smart contracts on Blockchain is accepted by ACM ISSTA 2023.
One paper on the management plane of Google’s datacenter networking architecture is accepted by USENIX ATC 2023. This work features years of deployment and operational experiences at Google!
One paper on optical datacenter networking design is accepted by IEEE ToN 2023.
One paper on multi-tab website fingerprinting attack (essentially analysis over encrypted network traffic) is accepted by IEEE S&P (Oakland) 2023.
One paper on certifying the ownership of DNN models is accepted by WWW 2023.
2022
Invited to serve as the PC member for ACM CCS 2023. Welcome your contributions.
Invited to give a talk about “Smart Contract and Zero-Knowledge Proofs” on a Web3.0 course offered by Graduate School of People’s Bank of China (Tsinghua Wudaokou).
Invited to give a talk about “Intelligent Network Data Plane” on Huawei Datacenter Networking Forum.
One work accepted by ACM CoNEXT'22, on enabling content providers to retain complete control over user experiences and economic/social fairness policies on the public Internet.
Invited to serve as the PC member for IEEE MetaCom 2023, a fresh venue on Metaverse and Web3.0. Welcome your contributions.
Invited to give a talk on the 5th Huawei Forum on Trust and Privacy for the Future Digital World. Will give a talk titled “A Data-Centric View on Web3.0”.
Invited to serve as the PC member for IEEE ICDCS 2023, a well-established venue on parallel and distributed systems. Welcome your contributions.
One work accepted by USENIX Security'23, on an efficient design of Intelligent Network Dataplane (IDP) that natively enables ML-driven traffic analysis at line-speed.
Welcome to Haotian Xu and Chenyang Li to join InspiringGroup as graduate students.
One work accepted by HotNets 2022, on unifying L3-layer Internet innovations in a more deployable manner.
One work accepted by ESORICS 2022, on verifiable outsourced ML training.
Two works accepted by IEEE TIFS'22, on Network Virtual Functions and Learning-based Congestion Control, respectively.
Invited to serve as the PC member of ACM CCS 2022.
2021
Our work on detecting anomaly accesses to database systems is accepted by ACM SIGMOD'22.
Our work on performing practical differentially private online advertising is accepted by Computers & Security 2021.
Our work on hot-patching IoT devices to fix vulnerabilities based on eBPF is accepted by USENIX Security'22.
Our work on detecting malicious accounts in WeChat is accepted by ACSAC'21.
Served in the Technical Interview Committee for graduate student admission in Department of Computer Science and Institute of Cyberspace. Congrats to all incoming graduates of class 2022. Welcome Jinzhu and Xuanqi to join my group.
Our work on investigating Advanced Persistent Threat (APT) via causality analysis is accepted by USENIX Security'22. Source code is released.
Our work on security of Graph Neural Networks is accepted by ACM CCS'21. Source code is released.
Our work on de-anonymizing encrypted networking traffic is accepted by IEEE TDSC'21.
Visit Prof. Xinbing Wang (my first-ever research advisor) from Shanghai Jiao Tong University. Give a talk on datacenter networking and next-generation Internet architecture.
Our work on Web3.0 (the so-called blockchain-powered next-generation Internet or Metaverse) is accepted by IEEE TDSC'21. Source code is released.
Our work on verifying software-defined networking is accepted by IEEE TPDS'21. Source code is released.
Welcome Dai Zhang, Jiang Li & Zhenhua Zou. My research group at Tsinghua officially started!