IoT DATA STREAM HANDLING AND ANALYSIS

Abstract

Internet of Things (IoT) is an internet connectivity extension to the physical devices on heterogeneous networks to meet application requirements of IoT data on various grounds. IoT devices produce a significant volume of sensitive data, which they transfer to the cloud for processing and decision-making. IoT devices generate enormous amounts of streamed data, necessitating real-time responses. Numerous devices in the IoT enable continuous information generation and communication. When IoT devices generate a large amount of stream data and transmit it through the network, which relies on the internet and low-speed IoT infrastructure, real-time data transmission becomes a challenge. IoT devices face crucial challenges in processing, analyzing, and handling such data, as well as ensuring data security across multiple connected devices. We divided this research work into two subdomains: IoT data stream handling and IoT data stream analysis and we presented their significant contributions to the IoT data stream handling and analysis domain. Out of them, three belong to IoT data stream handling and three to the IoT data stream analysis subdomain. First, we have conducted a systematic literature review of IoT data stream handling and analysis issues to highlight the challenges of the existing work and identify the various application areas as well. Second, we proposed IoT data stream handling models. We developed a secure and energy efficient hybrid model to address identified challenges and provide an effective solution that uses compression and encryption to improve energy and security efficiency in IoT devices. This hybrid architecture uses ChaCha12-Poly1305 for authenticated encryption with associated data (AEAD) and SZ 2.1 compression. Using the proposed model the device’s battery life is also enhanced by 10% while ensuring the security of data. It also reduces both encryption and overall processing time by 95% and 98% respectively for two datasets used in the experiment. Another proposed delta encoding model compresses and encrypts the data stream during IoT network transmission, enabling compression in the form of deltas. We employed a lightweight stream cipher encryption technique to meet security requirements. The proposed delta encoding model is compared with the baseline LDPC encoding on temperature sensor datasets. As an result, the data are compressed up to 37.59%, and the average transmission time and the average power consumption are reduced by 72.57% and 68.86%, respectively. v Third, we worked on IoT data stream analysis, which involved analysing and forecasting the data generated by the IoT system. During pandemic time, we proposed a model for prediction of the transmission process of COVID-19 cases that uses the technique of long-short-term memory (LSTM), a deep learning technique suited for real data predictions based on time series stream data. The results shows that predicted model returns efficient results with minimal error for May, 4-15, 2020. This analysis helps the relevant nation to take decisions in that time. Later, we proposed the grape, apple and potato leaf disease detection network (GAP-LDDN), which uses dual attention techniques to extract features, identify crop diseases, and classify diseases. As a result, model achieves 99.98%,97.88% and 99.88% accuracy for grapes, apple and potato datasets respectively. Then we experimented with machine learning techniques to analyse biofuel-producing plants. Lastly, as an application of our research work, we invented a product as a solution to a problem in the agriculture sector, that was patented and is in the process of being translated into a commercialized product as an IoT and AI-based smart autonomous weed detection and removal system that removes unwanted weed from the crop field as well as collects and processes real-time data to make future predictions that will be used by farmers. This research successfully provides a more reliable, effective, efficient, optimal, and secure data stream handling and analysis system using Internet of Things technology. The results of the experiments, analysis, and performance evaluation confirm that the provided work creates a practical and dependable IoT environment. Furthermore, the comparison study demonstrates that the proposed method is better than the existing methods that are already in use. Consequently, the results of this research successfully provide an IoT stream data handling and analysis system that is efficient, optimal, and secure by utilizing IoT technology.