In response to the increased challenges the modern network & web application security landscape arises, the development of Firewalls and Web Application Firewalls systems taking advantage of the Machine Learning technology and the intelligence it provides has already been initiated by multiple vendors investing money and manpower on it. Motivated by the above statement, a security solution making use of machine learning technology will be examined. Main objective of this thesis, except from a holistic analysis of concepts such as network, web application, security and machine learning, will also attempt the development, implementation and evaluation of a machine learning – integrated WAF. More specifically, it will begin from the phase of parsing HTTP Requests, extract the characteristics that will assist on their classification, import them on a proposed classifier and build/train the corresponding model which will properly evaluated. Finally, the resulting model will be integrated with a reverse proxy and through the classification process will detect and mitigate malicious-identified requests.