The malware detection challenge of accuracy

Real time Malware detection is still a big challenge; although considerable research showed advances of design and build systems that can automatically predicate the maliciousness of specific file, program, or website, Malware is continuously growing in terms of numbers and maliciousness. Web-based Malware detection is also growing with the expansion of the Internet and the availability of higher speeds and bandwidths. In this paper, we design, develop and evaluate an application that able to determine whether targeted website is malicious or not by utilizing available detection APIs. These APIs are able to communicate with several public scanners and Malware repositories. While the availability of many public scanners can help utilize those public services, however due to the fact that in most cases, they produce conflicting decisions, the process to make a final detection inference is not a trivial task. We conducted experiments to evaluate the different decision outcomes that come from the different scanners that utilized machine learning, data mining and other techniques. We also evaluated the issue of "unrated" decision based on the different Malware scanners.