Developing a Hybrid Machine Learning Model for Enhanced Cyber-Security Threat Identification
Keywords:
Intrusion Detection System (IDS), Deep Learning, CNN–LSTM Hybrid Model, Cybersecurity Threat Classification, CSE-CIC-IDS2018 DatasetAbstract
Malicious use of technology is still a public threat and protecting our digital infrastructure across the globe requires solid intrusion detection solutions. We examine the establishment of a machine learning-based model for improved cyber threat analysis utilizing the open-source dataset presented by the CSE-CIC-IDS2018. It uses Feedforward Neural Network (FNN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) networks to test these models. Moreover, a CNN-RNN (LSTM) hybrid model is proposed to combine the spatial characteristic extraction feature of CNN and the sequential learning ability of LSTM.
The experimental results show that the LSTM and CNN models perform best with an accuracy index of 97.00% each, and the ROC curve showed very strong classification capability. Up next were the FNN model with an accuracy of 96.98%, and CNN-RNN hybrid model achieving an accuracy of 96.94% and an F1-score of 95.96%. The findings highlight the potential that deep learning techniques hold when it comes to identifying cyber threats, with LSTM and hybrid models outperforming conventional means. This research has implications for the field of AI and cybersecurity by contributing to the development of effective intrusion detection systems and highlighting the importance of optimizing deep learning architectures for threat analysis. In future research, real-time deployment strategies and testing the proposed model on different cybersecurity datasets could be explored for improved applicability in practice.