中央研究院 資訊科學研究所

Fully homomorphic encryption (FHE) offers a powerful paradigm for privacy-preserving computing, enabling third parties to process data without ever decrypting it. However, massive ciphertext expansion, costly bootstrapping processes for noise reduction, and severe computational overhead have rendered FHE largely impractical for decades.

Fortunately, recent advancements in approximate arithmetic, hardware-accelerated bootstrapping, and parallelized HE algorithms are steadily closing the performance gap between FHE and plaintext computation. While much of the field currently focuses on local optimizations—such as accelerating specific operations or addressing isolated bottlenecks—FHE applications can achieve significantly higher efficiency if computation chains are optimized at a broader, system-level scope.

In this talk, Hung-Wei will present three concepts that make FHE more feasible for real-world applications:

1. Reducing storage and data-exchange overhead via word-level encryption.
2. Minimizing transcoding demands by substituting logical operations with approximate arithmetic operations.
3. Decreasing bootstrapping frequency through novel code optimization strategies.

To conclude, Hung-Wei will demonstrate how these proposed concepts successfully improve the state-of-the-art performance of FHE-based database queries by at least 20x.

Hung-Wei Tseng is currently an associate professor in the Department of Electrical and Computer Engineering and a cooperating faculty of the Department of Computer Science and Engineering at University of California, Riverside.

Hung-Wei is interested in designing architecture, programming language frameworks, and system infrastructures that allow applications and programmers to use modern heterogeneous hardware components more efficiently. Hung-Wei's recent focus is on using hardware accelerators (e.g., TPU, Ray Tracing) to improve application performance and privacy-perserving computing through more efficient homomorphic encryptions.

主要學術榮譽與獎項說明(崇高學術榮譽)如下：

• IEEE Micro Top Picks（2026、2024、2020、2012）：IEEE Micro Top Picks為電腦架構（Computer Architecture）領域最具代表性的年度榮譽之一。其研究成果多次獲選 IEEE Micro 年度代表性論文，顯示其在電腦架構領域具有長期且穩定的學術影響力。
• ACM/IEEE MICRO Best Paper（2019、2021）：MICRO 為電腦架構領域頂尖國際會議，其研究曾獲 MICRO Best Paper Honorable Mention 與 Best Paper Nomination，代表其研究具高度創新性與競爭力。
• IEEE RTAS Outstanding Paper Award（2021）：其研究於 IEEE RTAS 獲傑出論文獎肯定，顯示該成果在相關研究社群中具有相當的創新性與代表性。
• Facebook Research Award（2018）：由國際科技公司 Facebook Research 頒發之研究獎助，顯示其研究方向受到國際科技產業界關注。
• 多項 NSF 研究計畫補助：曾獲美國國家科學基金會（NSF）多項研究補助，顯示其研究長期受到美國重要研究機構支持。