What is it about?
SecureFrameNet partitions neural graphs into secure and host subgraphs, utilizing Protected Virtual Machines for secure execution, ensuring model protection and performance on edge systems
Featured Image
Photo by Growtika on Unsplash
Why is it important?
The paper proposes a method for securing frozen neural graphs by identifying critical layers and separating them into secure sub-graphs, executed in a secure container like pKVM.SecureFrameNet partitions neural graphs into secure and host subgraphs, utilizing Protected Virtual Machines (pKVM) for secure execution.The framework allows developers to securely execute models without compromising performance, offering end-to-end tools and APIs for easy adoption in edge computing environments.
Perspectives
The paper addresses adversarial attacks, emphasizing the importance of protecting neural network models from extraction and potential misuse. It introduces SecureFrameNet, a secure framework that partitions neural graphs into secure subgraphs for execution on edge systems, enhancing security against RAM attacks and reverse engineering. The framework utilizes Protected Virtual Machines (pKVM) for secure model execution, offering a more effective solution compared to Trusted Execution Environments (TEE) .
Renju Nair
International Institute of Information Technology
Read the Original
This page is a summary of: SecureFrameNet:A rich computing capable secure framework for deploying neural network models on edge protected system., October 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3639856.3639864.
You can read the full text:
Contributors
The following have contributed to this page
