A risk assessment and optimisation model for minimising network security risk and cost

Abstract

Network security risk analysis has received great attention within the scientific community, due to the current proliferation of network attacks and threats. Although, considerable effort has been placed on improving security best practices, insufficient effort has been expanded on seeking to understand the relationship between risk-related variables and objectives related to cost-effective network security decisions. This thesis seeks to improve the body of knowledge focusing on the trade-offs between financial costs and risk while analysing the impact an identified vulnerability may have on confidentiality, integrity and availability (CIA). Both security best practices and risk assessment methodologies have been extensively investigated to give a clear picture of the main limitations in the area of risk analysis. The work begins by analysing information visualisation techniques, which are used to build attack scenarios and identify additional threats and vulnerabilities. Special attention is paid to attack graphs, which have been used as a base to design a novel visualisation technique, referred to as an Onion Skin Layered Technique (OSLT), used to improve system knowledge as well as for threat identification. By analysing a list of threats and vulnerabilities during the first risk assessment stages, the work focuses on the development of a novel Risk Assessment and Optimisation Model (RAOM), which expands the knowledge of risk analysis by formulating a multi-objective optimisation problem, where objectives such as cost and risk are to be minimised. The optimisation routine is developed so as to accommodate conflicting objectives and to provide the human decision maker with an optimum solution set. The aim is to minimise the cost of security countermeasures without increasing the risk of a vulnerability being exploited by a threat and resulting in some impact on CIA. Due to the multi-objective nature of the problem a performance comparison between multi-objective Tabu Search (MOTS) Methods, Exhaustive Search and a multi-objective Genetic Algorithm (MOGA) has been also carried out. Finally, extensive experimentation has been carried out with both artificial and real world problem data (taken from the case study) to show that the method is capable of delivering solutions for real world problem data sets.