Real Time Network Attack Detection Using Machine Learning Techniques
Snehal Sutar, Piyusha Khune, Hrittika Gandhi, Kumari Snehal Pattebahadur, Chetan Nimba Aher · 2023
The threat of cybercrime is rapidly increasing on a global scale, exploiting vulnerabilities found in cloud computing infrastructure. To combat these cybersecurity issues, machine learning has emerged as a powerful solution, capable of addressing challenges such as intrusion detection, malware classification, spam detection, and phishing detection. Compared to traditional methods, machine learning algorithms offer more effective identification of cyber risks, thereby alleviating the burden on security analysts. Among these algorithms, deep learning has demonstrated superior performance, enhancing the cost-effectiveness of cybersecurity measures. Developing an efficient network attack detection system using machine learning involves several stages, including data collection, preprocessing, algorithm selection, model training, evaluation, and deployment. Continuous monitoring and regular updates further enhance the system's ability to counter various network attacks. In our study, we achieved high levels of accuracy, precision, and recall in detecting specific types of attacks, namely SQL injection (Accuracy: 0.928, R2 score: 0.659, Precision: 0.985, Recall: 0.774) and phishing attacks (Accuracy: 0.841, R2 score: 0.364, Precision: 0.919, Recall: 0.760). Additionally, our Intrusion Detection System demonstrated accuracy levels of 0.928 for K-Nearest Neighbors (KNN), 0.659 for Bernoulli Naive Bayes (BNB), and 0.985 for Decision Tree Classifier (DTC). Furthermore, the model exhibited an accuracy of 0.985 in effectively detecting cross-site scripting attacks. These outcomes underscore the model's effectiveness in accurately identifying various types of attacks, providing robust cybersecurity measures.