Outlier detection with optimal hybrid deep learning enabled intrusion detection system for ubiquitous and smart environment

Ubiquitous system returns to making pervasive computing in daily lives, the objects of smart environment becomes intelligent and interconnect without anyone being conscious of the communication process. It includes the concept of mobility to the perception of omnipresence; thus, it makes reference to moving intelligent objects. To handle the new security requirements of ubiquitous computing, intrusion detection systems (IDSs) need to be designed using artificial intelligence (AI) techniques. With this motivation, this paper presents an outlier detection with optimal deep learning enabled IDS (ODODL-IDS) model for ubiquitous and smart environments. The major intention of the ODODL-IDS model is to determine the outliers and then classify the presence of intrusions. For outlier removal process, Cluster-based Local Outlier Factor (CBLOF) is applied. In addition, the hybrid convolutional neural network with attention long short term memory (CNN-ALSTM) model is employed for intrusion detection and classification. Moreover, the hyperparameter tuning of the CNN-ALSTM model can be performed using the poor and rich optimization algorithm (PROA). The experimental result analysis of the ODLDL-IDS model is validated using distinct benchmark intrusion dataset and the comparative result analysis pointed out the supremacy of the ODODL-IDS technique over the recent approaches with maximum accuracy of 98.73%.