Abstract
Voting systems have played an important role in human democracy for thousands of years. In traditional voting systems, all received votes are tallied manually. For large scale elections, this method is not only inefficient, but also error prone making it hard to provide a completely accurate result. An initial motivation for introducing mechanical or electronic support has been to provide efficient tallying and cost reduction. Although voting equipment, e.g. lever machines or DRE machines, can be designed under very strict standards, or independently verified by third parties, their internal workings are still hidden when they are used in an election. Thus, voters have to trust that the system will correctly tally the election. However, recent high-profile reports have exposed that some such equipment in fact suffers from a variety of security flaws. In recent years, thanks to the improvement of cryptographic techniques, researchers have found some mathematical solutions to design secure voting systems, in which security is the key feature; the correct behaviour of these systems can be verified publicly, without the loss of voter privacy. As a result, instead of trusting the provided equipment or election officials, voters can themselves verify that their votes have been correctly counted. This thesis has contributed to the research of secure voting systems in two aspects; First, it has analysed two existing secure voting systems, the Voting Ducks scheme by Kutylowski et al. and the Pret a Voter with Paillier encryption scheme by Ryan. The thesis has identified a number of security flaws within these two systems which were not previously known. Second, the thesis has introduced a number of contributions extending the design of the Pret a Voter protocols. Not only are the extended systems better equipped to handle different election methods, but also they enjoy more security features.