Security and Privacy Preservation Method Based on Data Disturbance for Collaborative Edge Computing for Social-Internet-of-Things (S-IoT)

Socia1 Internet of Things (SIoT) is new paradigm that integrate human social with physical Internet of Things (IoT) systems, social IoT is ray of hope in providing ubiquitous connectivity among users. The collaborative edge computing for SIoT is significant because it ensures human social features and connections among IoT devices and users. But it is formidable to perform CEC for social IoT due to user privacy, data security and network heterogeneity and so on. The problem of computing and storage in IoT devices is solved by Collaborative edge computing (CEC) with deep learning (DL). To prevent privacy breach and security threats at edge network and edge devices in SIoT system. This research work discusses the data protection method based on data disturbance approach. To ensure correctness of model adversarial training viewpoint is introduced. The proposed method is capable of predicting contextual filtering words in adversarial text examples, and adaptively determine word replacement order. The proposed method BESAW, a BERT Based Simulated Annealing Word algorithm is used. BESAW compare with five existing methods includes TextFooler, BAE, BERT-Attack, PWWS, and PSO in experimental analysis.