A machine learning approach for web intrusion detection: MAMLS perspective

Abstract

Open Web Applications Security Project (OWASP), an open-source community committed to serve application developers and security professionals has always accentuated on the dire consequences of web application vulnerabilities like SQLI, XSS, LDAP, and Buffer overflow attacks frequently occurring on the web application threat landscape. Since these attacks are difficult to comprehend, machine learning algorithms are often applied to this problem context for decoding anomalous patterns. This work explores the performance of algorithms like decision forest, neural networks, support vector machine, and logistic regression. Their performance has been evaluated using standard performance metrics. HTTP CSIC 2010, a web intrusion detection dataset is used in this study. Experimental results indicate that SVM and LR have been superior in their performance than their counterparts. Predictive workflows have been created using Microsoft Azure Machine Learning Studio (MAMLS), a scalable machine learning platform which facilitates an integrated development environment to data scientists.