Research and Application of Network Traffic Anomaly Detection Algorithm Based on Deep Learning

Abstract

Traffic anomaly detection is the process of spotting odd trends or departures from typical network traffic behavior. One drawback of traffic anomaly detection systems is their susceptibility to false positives. These systems may sometimes incorrectly flag normal variations in traffic patterns as anomalies, leading to unnecessary alerts and potentially diverting resources towards investigating non-existent issues. In this manuscript Research and Application of Network Traffic Anomaly Detection Algorithm Based on Deep Learning (RA-NTADA-EPTANN) is proposed. Initially, the data are collected from DS2OS Dataset. The collected data are fed to Pre-processing segment. In pre-processing segment, Confidence Partitioning Sampling Filtering (CPSF) is used to data cleaning, handling the missing values and noisy data. Then, the pre-processed data are given to feature selection process. Feature selection is done by Humboldt squid optimization algorithm (HSOA). In feature selection technique, seven features are selected. Finally the selected feature attributes are given to efficient predefined time adaptive neural network (EPTANN)for Anomaly Detection classifying such as the DOS attack, data probing, malicious control, malicious operation, scan, spying, and wrong setup. In general, Efficient Predefined Time Adaptive Neural Network(ANN) does not express some adaption of optimization strategies for determining optimal parameters to ensure accurate Anomaly Detection. Hence, Multi-Agent Cubature Kalman Optimizer (MACKO)is to optimize to Efficient Predefined Time Adaptive Neural Network which accurately Anomaly Detection. The proposed technique implemented in python and efficacy of RA- NTADA-EPTANN technique is assessed with support of numerous performances like Accuracy, Computational Time, F1-Score, Precision, Recall and ROC is analysed. Proposed RA- NTADA-EPTANN method attains 15.12%, 22.23% and 35.32% higher computational time analysed with the existing for Deep neural network based anomaly detection in Internet of Things network traffic tracking for the applications of future smart cities (AD-IOTNTT-DNN),Network traffic anomaly detection method based on chaotic neural network (NTAD-CNN) and Deep learning-based network anomaly detection and classification in an imbalanced cloud environment (NAD-ICE-DCNN),respectively.