Using Multivariate Heuristic Analysis for Detecting Attacks in Website Log Files: A Formulaic Approach

As cyberattacks on websites evolve and become more sophisticated, there is a pressing need for detection methodologies that can adapt to this ever-changing landscape. This pilot study evaluates current methodologies in order to identify gaps in current literature and assesses their ability to be deployed in a real-world scenario. In order to do this, we propose a shift towards a multivariate framework, which measures the influence of several key factors. It was hypothesised that historic data is useful in predicting attacks. The study was given access to real website data in order to verify the efficacy of a multivariate approach on finding a variety of attacks. Results indicated a significant improvement in accuracy, specificity and sensitivity in attack detection in comparison to previous methods. This empirical evidence highlights the importance of using real-world data in cyber security and takes an essential preliminary step to be expanded by future research.