Privacy preservation in IoT: Deep learning tuned key generation-assisted data sanitization and restoration process

Authors:
Publisher website
Google Scholar
Abstract
In the age of the Internet of Things (IoT), an immense volume of data is rapidly generated and entrusted to external units, including cloud centers for storage and analysis. This implies that the data produced by IoT devices is conveyed and stored in a distant location, where it can undergo processing, examination, and utilization to extract valuable insights and facilitate informed decision-making. Privacy protection methods can be applied in IoT settings to safeguard sensitive and private information from potential security risks. Within the scope of IoT privacy, data sanitization involves the permanent and irreversible concealment of all sensitive data from extensive streams of information. Conversely, data restoration is the reverse process of data sanitization. Considering privacy preservation in IoT for data streaming, this research paper implements a Deep Learning (DL) tuned key-based data sanitization and restoration approach. The entire process of the implemented approach is handled by two significant steps such as data sanitization and data restoration. In the data sanitization step, three processes take place such as retrieval of features, key generation via Self-Adaptive Osprey Optimization (SAOO) and key tuning via Enhanced Convolutional Neural Network (ECNN). Initially, features such as raw features, statistical features and improved entropy-based features are retrieved (considering this as sensitive information. To sanitize this, key generation is the subsequent phase, where optimal keys are generated by SAOO and its security and confidentiality are ensured by the Khatri–Rao product. Then, these optimal keys are tuned by ECNN to ensure the efficacy of privacy preservation in IoT without compromising the performance of the model. Finally, the data restoration process is to recover the previously sanitized data in an IoT environment and reconstitute the original data from their sanitized form while maintaining the security and confidentiality of raw sensitive data. With these steps, the privacy of an IoT environment is protected from any form of issues. Additionally, this approach is experimentally explored to determine its efficacy using various analyses.

This publication has 38 references indexed in Scilit: