Securing 5G Networks through Machine Learning – A Comparative Analysis

5G Networks have evolved to support higher speeds and bandwidths, ultra-density, lower latencies, and low TCO. These features are achieved through many new technologies and applications that make the network vulnerable and prone to cyber-attacks. Security being an important criteria for public networks, there is a requirement to update the security mechanisms to detect and prevent these cyber attacks. Detection of network anomalies can always be done through identifying the signature of the attack. However, this requires knowledge of all existing attacks which is a formidable task. The second option is to study the normal behaviour of the network and identify any anomalies as a possible attack that should be further explored. All these options are computation intensive and require time which can leave the network exposed. In order to speed up the identification of anomalies, Machine Learning is a viable option. This paper reviews and explores application of various Machine Learning models for Intrusion Detection. A new dataset – CSE-CIC IDS2018 has been used to model and compare the effectiveness of several machine learning techniques in identification of cyber attacks. The subset of dataset selected has balanced number of flows for normal traffic as well as attacks so binary classification has been applied on it. The dataset was preprocessed and cleaned and then feature selection was carried out to identify 20 relevant features. A number of supervised learning classifiers including Gaussian Naive Bayes, Gradient Boosting, Logistic Regression, Decision Tree, Random Forest and Support Vector Machine were applied on the dataset to check their effectiveness. This paper also presents a comparison of performance of these classifiers for Intrusion Detection in the network using CSE CIC IDS2018 dataset.