M.Sc. Research Seminar
Doron Oded. Supervisor: Dr. Lior Rokach, Prof. Noam Tractinsky
Keystrokes dynamics based verification systems analyze and process biometric data. Data that contains certain amount of noise, caused by imperfect measuring instruments, constantly changing human patterns, and more. In this thesis, the question of the effect of cognitive load condition on keystroke dynamics is tested, and found significant. Di-graph latency was found affected by cognitive load condition while feature latency was not. A cognitive load awareness method is developed in order to attempt to neutralize the noise caused by the cognitive load effect, and by that improve verification accuracy. Mixed results are presented regarding the effectiveness of the cognitive load awareness method to neutralize the cognitive load effect.
Anonymization of Sequential Releases of Databases
Raz Wasserstein. Supervisors: Prof. Bracha Shapira ,Dr. Lior Rokach
An enormous amount of information, some of it very sensitive, is collected every day about individuals. In today’s global network of organizational connections, the demand is growing to disseminate and share this information. This, however, could violate the privacy of those from whom the information has been gathered. In dealing with this dilemma of publishing useful information while preserving data privacy, a new research field has emerged -- privacy preserving data publishing. Sequential release is a data publishing scenario where multiple releases of the same underlying table are published over a period of time. In this thesis we create a framework that offers a solution to the problem of maintaining privacy in a sequential release scenario. This framework is based on an anonymization algorithm that is applied to each new release prior to it being published. The anonymization algorithm joins all the information published in all previous releases and, if necessary, generalizes/anonymizes the new release in order to maintain the individual's privacy considering all published releases. This framework enhances well-known existing frameworks, while focusing on the quality of the published releases and their level of privacy maintained. Our theoretical study is followed by extensive experimentation that demonstrates a significant improvement in terms of the utility of the published releases.