| Details |
| Project Title | cuHEDL: Privacy Preserving Deep Learning Through Fully Homomorphic Encryption and Advanced GPU Acceleration Techniques
|
|
|
| Location | Sg Long Campus |
| Stipend | RM2,000.00 (T & C applies)
|
| Education Level | Minimum Bachelor degree with CGPA of 2.75 and above |
| Specific Skills / Knowledge Required | C/C++ programming, knowledge related to mathematics/cryptography and graphical processing unit (GPU) are added advantage
|
| Job Description | - Deep Learning applications are commonly hosted on the cloud servers where users submit their own data and obtain the classification/inference results from the pre-trained models. Privacy preservation becomes an important issue because the users may want to protect their sensitive data from being known to the cloud. To provide privacy preservation, Fully Homomorphic Encryption can be used to encrypt the user’s sensitive data. The cloud servers can process encrypted data and produce the results without knowing the actual content; only the users with the encryption key can decrypt the results and consume it. However, the performance of such solution may not be practical due to the expensive computation. Existing schemes represent the plaintext in polynomial form, which is not easy to work with Deep Learning models (floating points).
- This project aims to develop advanced packing and encoding techniques for Fully Homomorphic Encryption to enable parallel execution. The proposed techniques aim to improve efficiency of homomorphic operations and control the growth of noise after every multiplication. It also aims to handle approximation of activation functions, so that the homomorphic operations can be performed without losing too much accuracy. The developed techniques will be implemented on GPU platform with massively parallel architecture. The main focus will be on improving the speed of large integer multiplication, which is the main bottleneck in homomorphic operations.
- The outcome of this project (cuHEDL) are techniques to allow practical use of FHE on Deep Learning to provide privacy preservation, which is not currently available. It provides advancement to Fully Homomorphic Encryption in handling Deep Learning applications and allow practical performance on GPU platform.
- The successful candidate will be given training on cryptography and GPU programming. The candidate has to register for a Master by Research at UTAR and will be paid a monthly stipend of RM 2000 in the first year and renewable in the second year, depending on the performance. Salary adjustment depends on the performance.
|
| How to Apply | If you are interested, please contact Ts. Dr. Yap Wun She at yapws@utar.edu.my
|
