Cyber-Attack Monitoring and Detection using Machine Learning Techniques

In today's globalized world, cybersecurity stands as a critical pillar safeguarding the integrity of digital infrastructure against an ever-evolving landscape of cyber threats. This paper focuses on cyberattack detection utilizing machine learning algorithms applied to network traffic data. With a network traffic dataset containing essential features extracted from network packets, including packet length, source and destination IP addresses, protocol information, and TCP/UDP port numbers, we explore the effectiveness of seven classifiers for the binary classification tasks, distinguishing between cyber-attack instances and regular network traffic behavior. The classifiers used in this study were selected based on their suitability for the task and their prevalence in the field. They include Logistic Regression, Stochastic Gradient Descent (SGD) Classifier, Nearest Centroid Classifier, Ridge Classifier, Passive Aggressive Classifier, Decision Tree Classifier, and Bernoulli Naive Bayes. We evaluate the performance of each classifier in terms of accuracy, precision, recall, F1-score, area under the receiver operating characteristic (ROC AUC), and confusion matrix. The study's findings on the effectiveness and shortcomings of different machine learning methods in identifying cyberattacks can be instrumental in creating strong network security defenses.