Innovative Cryptographic Technologies and Applications Forum – Session Summaries and Speaker Information

The forum, organized by the China Cryptology Society's Big Data and AI Security Committee and hosted by Ant Technology Research Institute, will be held at the INCLUSION·Bund Conference on September 7, showcasing leading experts from academia and industry.

01. "Hardware Security and Data Factor Market" Speaker: Fan Junfeng, CEO of Shenzhen New Creation Information Security Technology Co., Ltd. Abstract: Hardware security, a mature technology widely used in finance and authentication, is re‑examined from the perspective of building a data‑factor market, focusing on security requirements for data generation, transmission, storage, and computation, and discussing a regulatable, easily deployable hardware‑security platform.

02. "Secure Two‑Party GPT Inference: Techniques and Outlook" Speaker: Liu Jian, Researcher at Zhejiang University. Abstract: ChatGPT raises serious privacy concerns; existing secure inference solutions cannot handle large GPT models. The talk introduces CipherGPT, the first secure two‑party GPT inference framework, highlighting novel cryptographic protocols that improve efficiency over state‑of‑the‑art, while acknowledging current impracticality and future research directions.

03. "Performance‑Optimized Result‑Hiding Searchable Encryption" Speaker: Wang Jianfeng, Professor at Xidian University. Abstract: Encrypted Multi‑Map (EMM) constructions often leak the number of query results, leading to leakage‑abuse attacks. The presentation describes the first communication‑optimal and storage‑optimal volume‑hiding EMM protocol using XOR filters, a novel ciphertext index, and an XOR‑based homomorphic PRF for integrity verification, achieving optimal retrieval and storage costs.

04. "End‑to‑Same‑End Encryption: Modularly Augmenting an App with Efficient, Portable, Blind Cloud Storage" Speaker: Chen Long, Associate Researcher at the Institute of Software, Chinese Academy of Sciences. Abstract: The talk proposes a system that protects cloud‑stored user data from the cloud itself without requiring new programmable operations on the storage service. Using a key server and existing app password login, the solution offers end‑to‑same‑end encryption, proven security, and demonstrated efficiency on Amazon S3, with a prototype deployed in Snapchat's "My Eyes Only".

05. "Fully Homomorphic Encryption Database Design" Speaker: Bian Song, Associate Professor at Beihang University. Abstract: Fully homomorphic encryption (FHE) enables arbitrary computation on ciphertexts. The report explores how to build a fully encrypted database, covering homomorphic implementations of filtering, linear and non‑linear aggregation, sorting, and grouping operators, and discusses handling algebraic values versus logical states for OLAP queries.

06. "SUPERPACK: Dishonest Majority MPC with Constant Online Communication" Speaker: Song Yifan, Assistant Professor at Tsinghua University. Abstract: The presentation addresses efficiency improvements for multi‑party computation (MPC) under the dishonest‑majority model, extending beyond the worst‑case single‑honest‑party scenario of SPDZ, and shows how multiple honest parties can be leveraged to reduce online communication complexity.

07. "Active Privacy: Edge‑Cloud Integrated Privacy Computing" Speaker: Wang Jun, Senior Expert at OPPO Research Institute. Abstract: With growing data generation, protecting user privacy while enabling personalized services is critical. The talk surveys active privacy from an edge‑cloud perspective, presenting recent research and practical explorations at OPPO on privacy‑preserving computation across user devices and cloud services.

08. "Bicoptor: Two‑Round Secure Three‑party Non‑linear Computation without Preprocessing for Privacy‑preserving Machine Learning" Speaker: Zhou Lijing, Privacy Computing Expert at Huawei. Abstract: Bicoptor introduces a new 3‑party secure computation protocol that achieves two‑round communication without any preprocessing, enabling efficient evaluation of non‑linear functions (e.g., ReLU, MaxPool) for privacy‑preserving machine learning, achieving orders‑of‑magnitude speedups over prior works.