A new model obfuscation approach GroupCover, which uses sufficient randomization and mutual covering obfuscation to protect model weights.

In this paper, we aim to further improve. TEE-based solutions. Specifically, we focus on the model obfuscation technique that can utilize both the security of ...

May 1, 2024 · A new model obfuscation approach GroupCover, which uses sufficient randomization and mutual covering obfuscation to protect model weights.

Code for GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs.

“GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs” (CCF-A, 2609/9653=27.03%). [CCS'23a] ...

GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs. Z Zhang, N Wang, Z Zhang, Y Zhang, T Zhang, J Liu ...

GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs. Z Zhang, N Wang, Z Zhang, Y Zhang, T Zhang, J Liu ...

Grokking Group Multiplication with Cosets · GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs ...

Oct 11, 2023 · We propose TEESLICE, a novel TSDP solution for. DNN inference that isolates privacy from offloaded model parts to provide a strong security ...

GroupCover: A Secure, Efficient and Scalable Inference Framework for On-device Model Protection based on TEEs · Vision Transformers as Probabilistic Expansion ...