| In Cyber Physical Systems (CPSs), traditional security mechanisms such as cryptography and access control are not enough to ensure the security of the system since complex interactions between the cyber portion and physical portion happen frequently. In particular, the physical infrastructure is inherently observable; aggregated physical observations can lead to unintended cyber information leakage. Information flow analysis, which aims to control the way information flows among different entities, is better suited for CPSs than the access control security mechanism. However, quantifying information leakage in CPSs can be challenging due to the flow of implicit information between the cyber portion, the physical portion, and the outside world. Within algorithmic theory, the online problem considers inputs that arrive one by one and deals with extracting the algorithmic solution through an advice tape without knowing some parts of the input. This dissertation focuses on statistical methods to quantify information leakage in CPSs due to algorithmic leakages, especially CPSs that allocate constrained resources. The proposed framework is based on the advice tape concept of algorithmically quantifying information leakage and statistical analysis. With aggregated physical observations, the amount of information leakage of the constrained resource due to the cyber algorithm can be quantified through the proposed algorithms. An electric smart grid has been used as an example to develop confidence intervals of information leakage within a real CPS. The characteristic of the physical system, which is represented as an invariant, is also considered and influences the information quantification results. The impact of this work is that it allows the user to express an observer's uncertainty about a secret as a function of the revealed part. Thus, it can be used as an algorithmic design in a CPS to allocate resources while maximizing the uncertainty of the information flow to an observer. |
