Yancheng Zhang

Yancheng Zhang

Auto-generated profile, based on publicly available information

Yancheng Zhang

University of Central Florida, Orlando, FL, USA

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 November 2023
Verified
Conference Paper
Open access
JXJiaqi Xue
Jiaqi Xue
YZYancheng Zhang
Yancheng Zhang
YWYanshan Wang
Yanshan Wang
XWXueqiang Wang
Xueqiang Wang
HZHao Zheng
Hao Zheng
QLQian Lou
Qian Lou
Secure training, while protecting the confidentiality of both data and model weights, typically incurs significant training overhead. Traditional Fully Homomorphic Encryption (FHE)-based non-inter-active training models are heavily burdened by computationally demanding bootstrapping. To develop an efficient secure training system, we established a foundational framework, CryptoTrain-B, utilizing a hybrid cryptographic protocol that merges FHE with Oblivious Transfer (OT) for handling linear and non-linear operations, respectively. This integration eliminates the need for costly bootstrapping. Although CryptoTrain-B sets a new baseline in performance, reducing its training overhead remains essential. We found that ciphertext-ciphertext multiplication (CCMul) is a critical bottleneck in operations involving encrypted inputs and models. Our solution, the CCMul-Precompute technique, involves precomputing CCMul offline and resorting to the less resource-intensive ciphertext-plaintext multiplication (CPMul) during private training. Furthermore, conventional polynomial convolution in FHE systems tends to encode irrelevant and redundant values into polynomial slots, necessitating additional polynomials and ciphertexts for input representation and leading to extra multiplications. Addressing this, we introduce correlated polynomial convolution, which encodes only related input values into polynomials, thus drastically reducing the number of computations and overheads. By integrating CCMul-Precompute and correlated polynomial convolution into CryptoTrain-B, we facilitate a rapid and efficient secure training framework, CryptoTrain. Extensive experiments demonstrate that CryptoTrain achieves a ~5.3× training time reduction compared to prior methods.

19 November 2023
Verified
Conference Paper
Open access
YZYancheng Zhang
Yancheng Zhang
XCXun Chen
Xun Chen
QLQian Lou
Qian Lou
Fully homomorphic encryption (FHE) allows computation over encrypted data without decryption and is considered one of the most essential primitives for privacy-preserving applications. However, there are still no universal FHE schemes that can support efficient and precise evaluation of both arithmetic and logic operations. Many endeavors have been made to enhance the capability of FHE for general computation, such as scheme switching and polynomial approximation. However, the overhead of scheme switching remains prohibitive for real applications with large bit-width inputs. On the other hand, the approximation methods have large errors around the pivotal point, limiting their application in precision-sensitive tasks. Recent studies show that FV-style HE schemes can support efficient and precise logic operations via polynomial interpolation. However, these methods cannot be seamlessly incorporated with arithmetic operations. In this work, we introduce HEBridge to connect the arithmetic and logic operations in the FV-style HE schemes. We first demonstrate that the arithmetic and logic operations operate over different underlying plaintext spaces. To enable continuous arithmetic and logic operations, we propose a reduction function and a lifting function to switch between these plaintext spaces. With HEBridge, we can exploit fast arithmetic and precise logic operations simultaneously in FV-style HE schemes. Experimental results show that the proposed HEBridge is 32.9× faster than direct interpolation methods and 1 to 3 orders of magnitude more efficient than scheme switching on large bit-width inputs.