OPTIMIZING INTRUSION DETECTION PERFORMANCE THROUGH MLP NEURAL NETWORKS

In the technological environment, cyber-attacks are popular and become more sophisticated, and existing intrusion detection models are not being adequate in hard threat escapes. The reason for increasing demands in the network systems, malicious attackers are increasing their involvement in these areas. Numerous organizations or network based environments have tackling hundreds of attacks on network frequently. Conventionally, various manual methods have used for intrusion detection like traffic log reviews, flow and packet analysis and security monitoring. Nonetheless, these techniques are found to be time-consuming and inefficient for obtaining better results and also manual methods are unable to detect all types of attacks for network security. To overcome this problem, several researches have focused on intrusion detection systems to provide effective security to the networks. Conversely, it results with speed and accuracy lacks. For improving the intrusion detection, the proposed research employs a Deep Learning (DL) approach Encoded Multi-Layer Perceptron (Encoded MLP) with attention mechanism. For the process of feature optimization, the proposed model uses Adaptive Beetle Algorithm which is used to provide accurate results for the detection model. It uses RT-IoT dataset for evaluation. Correspondingly, the efficacy of the projected model is calculated utilising various evaluation metrics like F1 Score, recall, accuracy, precision and accuracy to estimate the performance of the proposed DL research. Furthermore, the comparison of existing models discloses the efficiency of the proposed method. The present research is envisioned to contribute to the emerging intrusion detection methods for network and cyber security among hackers and malicious attacks.