Modern critical infrastructures have to process significantly large data sets. Intrusion-detection systems and unified threat management systems have the role of keeping critical infrastructures secure against cyber-attacks. However, in the world of big data, these systems are struggling to cope with overload and often become the bottle neck in the data network. To overcome this, our research investigates the use of deploying intrusion-detection and intrusion-prediction techniques in a cloud environment. Consequently, in this paper, a survey of existing intrusion-detection systems is presented and a discussion on how their deployment can enhance current security techniques in a cloud computing environment is put forward. A novel technique for intrusion prediction system is also put forward in this paper. Predictive statistical methods are used for proving the concepts put forward. The initial results show the necessity for using evolving statistical methods in prediction solutions; and the insufficiency of 'single-technique models' for building general solutions to predict intrusions. Furthermore, as this research shows, the concept of integrating multiple methods, such as game theory concepts and risk assessment methods, facilitates the development of a more efficient prediction model.