The Gorilla Troops Optimizer-Based Ensemble Deep Learning Model for Real-Time Zero-Day Attack Detection and Classification

Securing computer networks is becoming increasingly important and gaining significant attention. Security attacks, especially zero-day attacks, pose significant risks to enterprise and critical networks due to their unknown vulnerabilities and detection challenges. To ensure cybersecurity in networks, Intrusion Detection Systems (IDSs) observe network traffic for malicious actions and related attacks. Deep Learning (DL) and Machine Learning (ML)-based IDS are widely adopted for their adaptability and robust detection capabilities, particularly against zero-day attacks. This study presents the Gorilla Troops Optimizer-based Ensemble DL Model for Zero-Day Attack Detection (GTOEDLM-ZDAD) technique, aimed at classifying and detecting zero-day attacks using ensemble and advanced optimization algorithms. Initially, Linear Scaling Normalization (LSN) is used, and the Chimpanzee Optimization Algorithm (ChoA) is utilized for feature subset selection. An ensemble DL model uses Deep Q-Network (DQN), Bidirectional Gated Recurrent Unit (BiGRU), and Deep Belief Network (DBN) for classification. Finally, Gorilla Troops Optimizer (GTO)-based hyperparameter tuning is performed. A wide range of experimentation of the GTOEDLM-ZDAD technique on the ToN-IoT dataset achieved a superior accuracy of 98.33% over existing approaches and baseline models.