Trust model and privacy protection in edge computing environment

With the rapid increase in the number of terminals accessing Internet and the corresponding generated data, edge computing decreases the data sent to cloud center and reduces the pressure to bandwidth by processing data at the “edge” of network, either by the device or a local server. Meanwhile, the continuous evolution of Internet applications makes the content-centric mode be the trend of future Internet. Although the “content-centric” mode and in-network caching make content-centric model more appropriate for solving the issues of resource discovery, compute reuse, and mobility management than the current IP address centric mode, some important security and privacy problems are still needed to be solved. Therefore, a hierarchical generalized trust model and corresponding privacy protection strategy are proposed for the edge computing environment. In the trust model, the direct trust between nodes for different interaction relationships within the intra-layers and across different layers are calculated, the indirect trust of the non-direct interaction node is obtained by the recommendation trust from the direct interaction nodes, and then the comprehensive trust is acquired by combining direct trust and indirect trust. Based on the proposed trust model, privacy is protected by encrypting transmission content between trusted interaction nodes. A symmetric encryption algorithm is applied between the interaction nodes in the intra-layer. For both of interaction nodes in cloud/edge layers, the key length is set based on their environment trust. Meanwhile, for terminals, the key length is set based on their environment trust and available resource. An asymmetric encryption is applied between the inter-layer interaction nodes. For an interaction from one node in higher layer to the other in lower layer, the key is calculated based on the environment trust of the node in lower layer and the similarity of the request content from nodes in lower layer. For the opposite interaction, the key is calculated based on the environment trust of the node in higher layer and the available resource of the node in lower layer. While the consideration of environment trust improves the interaction security and helps prevent privacy leakage, the consideration of available resource adjusts the key length adaptively, which reduces overhead and improves calculation efficiency. Experimental results show that the trust model has superior performance, in terms of interaction success rate, accuracy rate, recall rate, etc., than other trust models and the privacy protection policy has less overhead than fixed key length settings.