Optimal Trained Deep Maxout Model for Intrusion Detection in Cloud

Abstract

The decentralized and distributed nature of cloud computing makes it easier for diverse society sectors, including environment, business, and government, to accept and use it. As it permeates every element of society, this computing paradigm is open to attacks and intrusions. Both security and privacy are seriously jeopardized by the enormous amount of data stored on the cloud. Along with this increase in usage, security risks to networks, the Internet, websites, and businesses are also rising. In a massive data environment like this, it can be challenging to find intrusions. Numerous artificial intelligence (AI) or machine learning-based intrusion-detection systems (IDSs) have been proposed for various forms of network assaults, but the majority of these systems are either unable to recognize unknown attacks or are unable to react to such attacks in real time. We suggested a new Deep Maxout algorithm-based intrusion detection system. Preprocessing, feature extraction, and detection are the three processes used in this instance of intrusion detection. The z-score normalization method is applied to normalize the input data during the preprocessing stage. Following that, the feature extraction stage will use the normalized data. There are two features extracted that are statistical features and higher-order statistical features (HOS). The statistical features are defined in terms of the median, standard deviation, minimum, maximum, entropy, and variance. HOS characteristics are skewness and kurtosis. An optimized Deep Maxout model is used for the detection process, where the training will be carried out by the COOT optimization algorithm. This assures the improvement of detection accuracy.