Optimal weighted fusion based insider data leakage detection and classification model for Ubiquitous computing systems

Abstract

With the rapid growth and evolving advancement in artificial technology, Ubiquitous computing methods plays a vital role in day to day lives. Insider threats were malicious activities which are executed by an official worker within an institution. Insider threats indicate cybersecurity challenges for public and private companies, as an insider attack could cause widespread damages to company properties than exterior attacks. Many prevailing methods in the domain of insider threat concentrated on identifying general insider attack situations. But insider attacks are takes place in several ways, and one of the dangerous attack is a data leakage attack which is accomplished by a malicious insider before one leaves a company. This article develops a Metaheuristic with Weighted Fusion Based Insider Data Leakage Detection and Classification (MWF-IDLDC) Model for Ubiquitous Computing Systems. The presented MWF-IDLDC technique primarily performs pre-processing and feature extraction at the initial stage. In addition, the MWF-IDLDC technique derives a weighted fusion-based feature extraction approach comprising three DL methods namely long short-term memory (LSTM), gated recurrent unit (GRU), and stacked autoencoder (SAE). For optimally tuning the hyperparameters related to the DL models, the ant lion optimizer (ALO) algorithm is utilized in this study. The design of fusion process with ALO hyperparameter optimizer demonstrate the novelty of the work. The experimental validation of the MWF-IDLDC approach can be tested using a series of simulations and the outcomes were scrutinized under distinct aspects. The comparative analysis highlighted the improvements of the MWF-IDLDC method compared to recent approaches with maximum accuracy of 99.40%.