A trusted and collaborative framework for deep learning in IoT

Abstract More and more Internet of Things (IoT) applications provide intelligent services, with the development of artificial intelligence algorithms, such as deep reinforcement learning. However, along with the trend of utilizing a large model with high accuracy in AI-enabled IoT, resource-limited IoT devices are difficult to handle these large-scale models with high response latency. By collaborating with edge nodes, the devices could respond quickly. However, IoT applications contain a large amount of user privacy, and pushing data to others might lead to privacy leakage. Inspired by the trusted execution environment technology, we propose a framework that enables trusted collaboration for future AI-enabled IoTs, in terms of computation security and transmission security, where the data could be processed in an isolated environment, and two approaches are proposed to ensure the security in data transmission. Experimental results show that our framework provides flexible and dynamic collaboration with low overhead and can effectively support collaborative edge intelligence.