Threat Modeling and Recovery from attacks in Cloud Computing

Abstract

In the 1950s, large organisations used mainframes accessed through dumb terminals, allowing multiple users to share data and computing power. By the 1970s, systems like IBM System/370 introduced virtual machines, enabling multiple OS instances on a single node. In the 1990s, virtualisation and shared hosting gained popularity, paving the way for cloud computing though this growth also brought increasing challenges from malware attacks on cloud networks. On this foundation, this thesis presents an attempt to address four areas of concern. The first problem relates to threat modelling, for which the thesis presents an algorithm for performing goal-based threat modelling for clouds. After this, a modified digital forensic process model is presented in this thesis to cater to the requirements of handling incidents like launching multiple attacks on a single cloud node from different sources. Next, we consider epidemic modelling and the percolation phenomenon that is seen in complex networks and use these two concepts to explain the spread of botnets in a Peer-to-Peer (P2P) Cloud. Finally, an alternate optimal path needs to be found in the network for propagating a recovery image. Towards this end, Strength Pareto Evolutionary Algorithm (SPEA) is used. The possible future directions that the collective work in this thesis can lead to are also discussed. newline