Detecting Network Security Bypass Threats Using Machine Learning Methods: Detecting Intruders on the Network

The problem of cybersecurity has grown in importance. Machine learning (ML) systems can detect network penetration. Imbalanced data sets have a detrimental impact on typical network intrusion detection. To be more precise, seven traditional ML algorithms were tested against two versions of a fully connected neural network, one with and one without an autoencoder. Additionally, an electing classifier is suggested as a means to integrate the outcomes of these nine ML algorithms. The majority electing classifier allows for the combination of several weak classifiers into a strong classifier. The number and type of weak classifiers used will have an impact on the final ensemble classifier's performance Three distinct resampling methods oversampling, undersampling, and hybrid sampling are used to evaluate each model. Next, we will go over the specifics of the trials and how we analyzed the data. The comparison results show that the performance of the classifiers on balanced data outperforms those on ( https://www.sciencedirect.com/topics/computer-science/imbalanced-data) imbalanced data, and the electing classifier outperforms the nine algorithms. A weighted F1 score is a good performance metric to evaluate solutions in intrusion detection systems. Due to the importance of the F1 score parameter, the proposed method has reached a predict of 80%, which is a significant improvement compared to related works.