Defence Mechanism against Advanced Persistent Threat Attack Using Significant Features based Deep Learning Model

Abstract

Translate Article

The advanced persistent threat (APT) attack is the most frequent and seriously damaging one that can be launched against the target system. This type of malicious activity is deliberate and focused, with specific objectives in mind. The rise of this attack technique poses a significant challenge to the information security systems of various businesses, governments, and organizations. The techniques of utilizing deep learning (DL) or machine learning (ML) methods to evaluate abnormal behaviours and network traffic signs for preventing and detecting APT attacks are becoming popular in recent times. Though, the APT attack detection technique utilizes behaviour evaluation and analysis methods for confronting various problems owing to the absence of distinctive data of attack drives. So, the experimental outcomes display better recognition, it does not bring higher effectiveness in practice. Therefore, this study develops a novel Defence against Advanced Persistent Threat Attack Detection Feature Selection with Deep Learning (DAPTAD-FSDL) Method. The key objective of the DAPTAD-FSDL approach lies in the robust defence against the detection of APT attacks. In the presented DAPTAD-FSDL model, Z-score normalization is performed to change the raw data into a compatible format. Then, the presented DAPTAD-FSDL technique accomplishes the feature selection process using the grey wolf optimizer (GWO) algorithm. Besides, the attention with CNN-BiLSTM model is employed for the defence against the detection of APT attacks. Finally, a multi-objective improved golden jackal optimization algorithm (MIGJOA) is exploited for the optimal hyperparameter adjustment of the A-CNN-BiLSTM model. The experimental analysis of the DAPTAD-FSDL algorithm is tested employing the APT NSL-KDD dataset. The simulation outcomes described the supremacy of the DAPTAD-FSDL technique under different measures.