新闻发布
近期，“信息与智能大讲堂青年论坛”系列讲座在腾讯会议平台共成功举办6场。
2024年5月21日下午，讲座特邀北京理工大学左聪教授担任主讲嘉宾，讲座题目为“Result-pattern-hiding Conjunctive Searchable Symmetric Encryption with Forward and Backward Privacy”。
2024年5月23日下午，讲座特邀陕西师范大学来齐齐副教授担任主讲嘉宾，讲座题目为“Ring/Module Learning with Errors under Linear Leakage -- Hardness and Applications”。特邀北京大学李萌助理教授担任主讲嘉宾，讲座题目为“Efficient Private Transformer Inference through Network/Protocol Co-optimization”。
2024年5月24日上午，讲座特邀北京航空航天大学边松副教授担任主讲嘉宾，讲座题目为“面向隐私图神经网络推理的算术-逻辑混合的全同态加密加速器”。特邀电子科技大学李雄教授担任主讲嘉宾，讲座题目为“高效的隐私保护多方多数据排序”。
2024年5月27日上午，讲座特邀复旦大学黄橙青年研究员担任主讲嘉宾，讲座题目为“Multi-Client Secure and Efficient DPF-based Keyword Search for Cloud Storage”。
在这6场讲座中，各位老师都进行了精彩的内容分享，每场讲座结束后，老师们与青年学者们亲切互动，对不同问题进行详细解答，线上讨论热烈。
讲座在腾讯会议平台线上进行的同时，通过“信息与智能学报-英文”视频号、“西安电子科技大学学报”视频号同步直播，受到了国内外学者、研究生的广泛关注、讨论与好评。
“信息与智能大讲堂”系列讲座由《信息与智能学报（英文）》、《西安电子科技大学学报》、《西安电子科技大学学报（社会科学版）》、《电子科技》主办，西安电子科技大学通信工程学院协办。大讲堂围绕信息及智能领域，包括信息与通信工程、电子科学与技术、计算机科学与技术、网络空间安全等学科，探讨最新的科学与技术研究进展、发展趋势及未来挑战，搭建学术交流平台，促进相关领域的学科建设、科学研究和人才培养。西安电子科技大学期刊中心将持续主办“信息与智能大讲堂”系列讲座，为高校师生、科研人员答疑解惑，敬请期待。
嘉宾介绍
左聪：Cong Zuo received a B.S. degree from the School of Computer Engineering, Nanjing Institute of Technology, and an M.S. degree from the School of Computer and Information Engineering, Zhejiang Gongshang University, China. Recently, he got his Ph.D. degree from Monash University. He is currently a professor at the school of cyberspace science and technology, Beijing Institute of Technology (BIT). Before joining BIT, he was a research fellow at Nanyang Technological University. His main research interest is on cybersecurity, in particular, Database Security and Applied Cryptography.
Dynamic searchable symmetric encryption (DSSE) enables the data owner to outsource its database (document sets) to an untrusted server and make searches and updates securely and efficiently. Conjunctive DSSE can process conjunctive queries that return the documents containing multiple keywords. However, a conjunctive search could leak the keyword pair result pattern (KPRP), where attackers can learn which documents contain any two keywords involved in the query. File-injection attack shows that KPRP can be utilized to recover searched keywords. To protect data effectively, DSSE should also achieve forward privacy, i.e., hides the link between updates to previous searches, and backward privacy, i.e., prevents deleted entries being accessed by subsequent searches. Otherwise, the attacker could recover updated/searched keywords and records. However, no conjunctive DSSE scheme in the literature can hide KPRP in sub-linear search efficiency while guaranteeing forward and backward privacy.
In this work, we propose the first sub-linear KPRP-hiding conjunctive DSSE scheme (named HDXT) with both forward and backward privacy guarantees. To achieve these three security properties, we introduce a new cryptographic primitive: Attribute-updatable Hidden Map Encryption (AUHME). AUHME enables HDXT to efficiently and securely perform conjunctive queries and update the database in an oblivious way. In comparison with previous work that has weaker security guarantees, HDXT shows comparable, and in some cases, even better performance.
来齐齐：陕西师范大学计算机科学学院副教授、硕士生导师。研究方向为后量子安全的公钥密码方案的设计与分析。2015年获得西安电子科技大学密码学专业博士学位。目前，在国际密码学会顶级会议Eurocrypt，PKC等发表多篇论文。主持国家自然科学基金面上项目、青年项目各一项、ISN重点实验室开放课题一项、密码科学技术国家重点实验室开放课题一项、河南省网络密码技术重点实验室研究课题一项。获党政机要密码科学技术奖三等奖一项。
内容提要：
This work studies the hardness of decision Module Learning with Errors (MLWE) under linear leakage, which has been used as a foundation to derive more efficient lattice-based zero-knowledge proofs in a recent paradigm of Lyubashevsky, Nguyen, and Seiler (PKC 21), Lyubashevsky, Nguyen, and Plancon (CRYPTO 22). Unlike in the plain LWE setting, it was unknown whether this problem remains provably hard in the module/ring setting.
This work shows a reduction from the search MLWE to decision MLWE with linear leakage. Thus, the main problem remains hard asymptotically as long as the non-leakage version of MLWE is hard. Additionally, we also refine the paradigm of Lyubashevsky, Nguyen, and Seiler (PKC 21), Lyubashevsky, Nguyen, and Plancon (CRYPTO 22) by showing a more fine-grained tradeoff between efficiency and leakage. This can lead to further optimizations of lattice proofs under the paradigm.
李萌：2022年加入北京大学集成电路学院和人工智能研究院，任助理教授、博士生导师，获国家青年高层次人才计划（海外）。加入北京大学前，他曾任职于全球最大社交媒体公司Facebook的虚拟现实和增强现实实验室，作为技术主管从事高效人工智能加速算法和系统的研究。他于2018年和2013年分别在美国德州大学奥斯汀分校和北京大学获得博士和学士学位。他的研究兴趣集中于高效、安全的多模态人工智能加速算法和系统。他在国际顶级会议、期刊发表文章70余篇，并于2017年和2018年获得IEEE HOST和ACM GLSVLSI的会议最佳论文。此外，他还获得了欧洲设计自动化协会最佳博士论文、ACM学生科研竞赛总决赛第一名等奖项。
内容提要：
Recent years have witnessed the fast evolution of AI and deep learning (DL). However, with the wide application of DL comes ever-increasing privacy concerns with data and neural network (NN) models. Private NN inference based on two-party computation (2PC) is proposed, which provides cryptographically strong privacy protection. However, such a strong privacy guarantee is achieved at significant communication and computation overhead, leading to several orders of magnitude latency degradation. In this talk, I will discuss some of my recent works on NN/protocol co-optimization. Specifically, I will highlight the importance of 2PC-aware NN optimization and NN-aware protocol design in mitigating the latency gap. I will also discuss the future directions to further improve the efficiency of private inference for modern networks.
边松：北京航空航天大学网络空间安全学院副教授。他的主要研究方向是同态加密、隐私保护计算与密码软硬件协同加速，于CCS、NDSS、USENIX Security、VLDB及DAC等多个领域的CCF-A类期刊及会议以第一或通讯作者发表论文十余篇，获CCS 2023杰出论文奖。主持国家重点研发计划青年科学家、国家自然基金委青年基金等项目，入选第九届中国科协青年人才托举工程。他担任多个跨领域的国际会议与期刊的程序委员与审稿人, 包括CVPR、TIFS及TCAD。他是IEEE与CCF的会员，CCF杰出演讲者。
内容提要：
图神经网络（GNNs）在社交媒体和生物信息学等领域的应用日益广泛，推动了基于云的GNN推理服务的发展。然而在云上处理敏感数据时，如何保护数据隐私成为一个关键问题。全同态加密（FHE）能在加密的数据上进行GNN计算。但现有的基于FHE的GNN推理方案常常因为计算开销大、精度下降或数据保护不完整而在实际应用中受到限制。
本文提出了PPGNN，提出了一种基于算术与逻辑混合FHE方案的隐私保护GNN推理算法，该算法在维持高精度的同时，仅需较小的FHE参数，并将计算负载集中在适合硬件加速的并行计算部分。在此基础上，我们设计了一种专用硬件架构来加速该算法，该架构包含用于分别加速算术FHE和逻辑FHE算子的专用硬件单元，并实现了两部分计算单元的流水线处理。实验结果表明，PPGNN相较于最先进的算术FHE和逻辑FHE加速器，分别实现了2.7倍和1.5倍的速度提升，并且平均减少了约18倍的能耗。
李雄：电子科技大学教授、博导、国家级和四川省青年人才。主要研究方向为密码协议、数据安全与隐私计算。主持国家级和省部级项目10余项。曾入选全球高被引科学家, 中国高被引学者, 全球前2%顶尖科学家榜单。发表论文200余篇, 其中在IEEE系列期刊、CCF AB类期刊、SCI 一二区期刊论文100余篇, Google Scholar被引11000余次, H指数63。荣获2023 IEEE MASS Best Paper Runner-Up Award, 荣获2020年IEEE Systems Journal最佳论文奖, 荣获2020年和2015年Journal of Network and Computer Applications最佳研究论文奖。担任中文信息学会大数据安全与隐私计算专委会常务委员、中国计算机学会区块链专委会执行委员。
内容提要：
隐私保护排序允许多个参与方在不泄露各自数据隐私的前提下计算多个数据集的排序结果, 是一类基础的安全多方计算问题。现有协议大多只支持两个参与方且均未考虑恶意参与者的穷举攻击。本工作提出一个高效的隐私保护多方多数据排序协议。多个参与方仅需O(1)轮交互即可以隐私保护的方式获得其持有的多个数据的排序结果。具体来讲, 设计了基于多项式的编码方法、多项式加密、聚合多项式生成和解密多项式生成算法, 通过多项式加法实现隐私保护的多方多数据排序, 各参与方通过不经意传输获得排序结果。安全性分析表明该协议不仅实现了半诚实安全性, 而且达到了不合谋恶意用户穷举攻击的恶意安全性。此外, 大量实验表明协议是通信和计算高效的。
黄橙：Dr. Cheng Huang (IEEE Member '20) received his B.Eng. and M.S. degrees in information security from Xidian University, China, in 2013 and 2016, respectively, and his Ph.D. degree in Electrical and Computer Engineering from the University of Waterloo, ON, Canada, in 2020. He is currently an Associate Professor with the School of Computer Science at Fudan University. Before joining Fudan University, he was a Research Fellow in the Department of Electrical and Computer Engineering at the University of Waterloo from 2020 to 2023. His research interests lie in the areas of security and privacy in vehicular networks, data security, and secure computation. He has published over 60 papers in prestigious journals and conferences, including IEEE TDSC, IEEE JSAC, IEEE TVT, and IEEE TII, and has received Best Paper Awards from ICCC '15, ICC '18, GLOBECOM '22, and ICCC '23. He serves as the Associate Editor for Peer-to-Peer Networking and Applications (Springer), the Symposium Co-Chair of IEEE GLOBECOM '24, and has served as the publicity chair of ICA3PP '22, PST '23, SustainCom '23, and as the TPC member of many international conferences.
内容提要
In this talk, we present a multi-client secure and efficient keyword search scheme for cloud storage, which is built upon distributed point function (DPF). Specifically, in the proposed scheme, outsourced keyword indexes are encoded by using garbled bloom filter and cuckoo filter, instead of bloom filter adopted by most of the state-of-the-art DPF-based schemes. In this way, clients can apply cuckoo hashing into DPF and utilize a segmentation method to interact with cloud servers for keyword search, and servers can obliviously aggregate DPF evaluation results to perform the search. Accordingly, the computational complexity at server side can be significantly reduced. Furthermore, the proposed scheme preserves constant downlink overheads, which is more communication-efficient for multi-keyword conjunctive search. To achieve privacy preservation and access control for multiple clients, we propose a double encryption method to encrypt outsourced indexes and correspondingly put forward an authorization algorithm from set-constrained pseudorandom functions by which fine-grained search-authorized keys can be generated, and collusion attacks among clients are addressed by integrating Wegman-Carter message authentication codes and cover-free systems. Our scheme is designed under both semi-honest and malicious models (i.e., malicious servers may return incorrect query results), and its security is proved in a simulation-based paradigm. We also develop a proof-of-concept prototype and perform extensive experiments to show our scheme's practicality and efficiency in terms of computation, communication, and storage overheads.
