Employing Hybrid Serial Cascaded Adaptive Network for Anomaly Detection and Prevention in IoT Time Series Data With Optimal Interdomain Routing

Abstract

Nowadays, emerging Internet of Things (IoT) in data processing becomes interested research topic and is used for various applications. This varies from big data processing and analytics for accumulating entire sensor data to detect and generate long-term trends. Nevertheless, this results in the requirement for resources like the execution of power, memory, and bandwidth, in which resource-constrained IoT tools suffers during the data transmission in order to improve effective operations. Hence, these results in risk and poor availability of data, and accepted by an insider like a malignant system administrator, without leaving any hints of their activities. Henceforth, a novel anomaly detection and prevention model over IoT time series data using Deep Learning (DL) is presented to identify and mitigate the attacks. Primitively, the required data is assembled from the accurate websites. The anomalies are detected from the obtained data utilizing the developed Hybrid Serial Cascaded Adaptive Network (HSCAN). This method is constructed by the concatenation of the Spatial-Temporal Attention (STA-AE) based Autoencoder and Long Short-Term Memory (LSTM). Once the detection of anomalies tends to be completed, it is eliminated for the superior operation of the data. Furthermore, the efficacy of the prevention and the detection phase is improved by fine-optimizing the parameters from the offered model through the Iterative Concept of Peregrine Falcon Optimization (ICPFO) algorithm. After processing the detection and the prevention stage, the routing stage is taken over by the developed ICPFO by assuming the energy, latency, and Packet Delivery Ratio (PDR). Hence, the efficacy of the recommended approach in anomaly detection and processing the routing is estimated and compared to the classical methods.