Real time prediction and detection of malicious activities

The PhD project deals with the problem of hacking the infrastructure of a company or the everyday activities of individuals. The increased sophisticationand targeted nature of security threats, coupled with their increasing frequency, has ensured that security breaches are now the top issue affecting allusers and organizations today. Regarding the economic effect, security researches predicted that the global cost of data breaches would increase from $3trillion in 2015 to $6 trillion by 2021.Traditional security approaches are good practice, but they are no longer enough. Hackers increasingly bypasses perimeter security, enabling cyberthieves to pose as authorized users with access to corporate networks for unlimited periods. Organizational threats manifest themselves through changingand complex signals that are difficult to detect with traditional signature-based and rule-based monitoring solutions. Instead of signature and reputation-based detection methods, artificial intelligence innovations, use machine-learning algorithms to drive from post-incident to pre-incident threat intelligence.With machine learning, the computer is trained to find differences to distinguish normal behaviors from malicious activities but much faster than signature-based techniques. On the other hand, machine learning has a major disadvantage; most of the times predict a normal behavior as a threat. Hence, whilethe analysts are trying to classify a false predict activity as normal the true attacks are already in the system and exploit it.In our approach, we implement a new robust technique to reduce the ratio of these false predictions. POST Luxembourg is the industrial partner and ontop of its framework, we apply algorithms to predict and identify true attacks accurately. The outcome of the project will benefit the Luxembourgishindividuals and increase the reputation of provided services