JavaScript-based attacks have become one of the most common and successful types of attack. Existing techniques for detecting malicious JavaScripts could fail for different reasons. In this project, we propose an efficient method of detecting previously unknown malicious activity using an interceptor at the client-side. In this, we are using a machine-learning algorithm to detect JavaScript attacks. The interceptor monitors the traffic exchanged between the browser and server and extracts the real-time traffic features needed for XSS detection. Supervised machine learning classifiers such as k- NN, Support Vector Machines, Gaussian Naïve Bayes, and Logistic Regression are used on a dataset gathered for achieving high accuracy in webpage classification through learning from patterns. The experimental results obtained, show that this method can efficiently classify malicious code from benign code with promising results.