Threat Modelling Based on NIDS Attack Emulation with Deep Learning & Optimization Techniques
Main Article Content
Abstract
The prevalence of cyber-attacks has become a part of every software organization today which necessitates the need of protection so that user data, policies and procedures remain intact. Attacks and threats tend to create an imbalance in the confidentiality and integrity of the system. The infrastructure of the software is thus at risk and can be exposed to various malicious attacks. In such a scenario, a threat modelling procedure is needed so that user data remains secured. A service model that caters to the discerning issues of cyber-attacks is an intrusion detection technique that provides a coherent view and makes the detection system more robust in nature. The associated security concerns are therefore identified using the IDS technique and respective configurations are thus made. Therefore, the study thus proposed focuses to implement a network based intrusion detection system that inclines to identify cyber-attacks in conjunction with deep learning based strategies. Additionally the study makes use of Long Short-Term Memory (LSTM), Bi-directional Long Short-Term Memory (Bi-LSTM), and Bi-directional Long Short-Term Memory with Particle Swarm Optimization (BILSTM with PSO) in order to determine the excruciating patterns within a networking traffic. The entire process of implementation is done using the three algorithms based on DL. For this purpose a dataset consisting of various threat and attack scenarios is used. UNSW-NB15 is acquired from the Kaggle repository and further trained and tested on. However, evaluation is done on the basis of accuracy and precision factors and the model is rigorously run to generate improved results. The study on evaluation highlights that the Bi-LSTM model along with PSO generates highest optimization accuracy of 99percent and is therefore considered to inhibit tremendously potential in terms of generating insights to detect cyber threats and attacks