An Unstructured IoT Data Conversion and Secured Data Deduplication System in Cloud Network using Adaptive Deep Learning and Optimal Key-aided Cryptography

Abstract

Internet of Things (IoT) sensors continuously creates unstructured data, which needs to be transformed into a structured format in order to extract information from it. The designed data is helpful for other kinds of analysis, such as forecasting data that may be analyzed to make predictions about the future. More and more people are adopting cloud computing, due to more convenient, cost-effective, and time-efficient services than traditional architecture. Because cloud storage systems provide consumers with easy and affordable network storage, they are becoming more and more popular. The same data with multiple copies are frequently stored by cloud systems, particularly in backup conditions. In order to ensure that only one unique instance is stored data deduplication finds and eliminates these duplicates, thereby reducing the total memory space. Hence, in this proposed model, a new deep learning approach is proposed to address the issues present in the existing research. Initially, the essential data are gathered from the IoT sensors, where the data is in the unstructured format. At first, the gathered unstructured data is cleaned and given for the entity extraction process with the aid of Adaptive Dilated Conv-Cascaded Long Short-Term Memory (LSTM) with Attention Mechanism (ADCL-AM). Here the parameters present in the model are optimized using the Enhanced Position Updating-based Flamingo Search Algorithm (EPU-FSA) optimization. After extracting the entity, the pattern generation is carried out. Thus, the unstructured data is converted into a structured form, which is further used for the data deduplication process. In this phase, the structured data is subjected to ADCL-AM for data deduplication. Here also, the parameters present in the model are optimized using the same IFSA optimization. Subsequently, the deduplicated data are securely managed by using Hyper-Elliptic Curve Cryptography (HECC), in which the keys are optimally tuned by the developed IFSA algorithm. The competence of the proposed model is analyzed based on an experimental outcome. The result demonstrated that the proposed approach outperformed the standard techniques.