|M.Sc Student||Tamari Natan|
|Subject||Experimental Semiquantum Key Distribution: Classical Alice|
|Department||Department of Electrical Engineering||Supervisors||Professor Meir Orenstein|
|Professor Tal Mor|
|Full Thesis text|
One of the oldest and most serious problems in cryptography is that of key distribution. Two parties, Alice and Bob, want to share sensitive information, secure in their knowledge that Eve, an eavesdropper, cannot read their correspondence. In order to do so, they must share some prior information, which is known as the key.
Many key distribution schemes making use of quantum physics have been put forward, the most famous being BB84. Collectively, they are known as quantum key distribution (QKD) protocols. These protocols can be proven informationally secure. This means that any attempt by an eavesdropper, to gain information on the final key necessarily exposes her to a possibility that she will be noticed by the legitimate parties.
In 2007, Boyer, Kenigsberg and Mor showed that it is possible to achieve such security even when one of the parties is classical, meaning it can only perform measurements in a single quantum basis. In this thesis, we describe an attempt to experimentally realize the Classical Alice with Mirror protocol. Our goal is that this realization be performed in clear, pre-defined steps, and is thus repeatable.
The experimental system is an instrumentation-operated optical system that is software-controlled. We focused on several aspects of that system - the key exchange’s bit-rate, which was limited by communication with the controlling computer, the ability to reliably detect the presence of an eavesdropper, and the error rate on the key bits. In addition, the security proof is expanded, allowing the removal of a significant experimental constraint.
Our results are as follows: The error rate on the raw key bits is about 11%. The threshold error rate for a given QKD protocol depends on the exact model used for the communication and for Eve’s information. For BB84, as an example, this rate would be good enough for Alice and Bob to perform key distillation. However, for Classical Alice with Mirror, the situation is considerably more complex, as Eve’s information depends on two separate error rates - that of the key bits and that of the eavesdropping check bits. Our current method of eavesdropper detection manages to achieve an error rate of approximately 15%. At the moment, a typical number of raw key bits per second is 6.5, which is better than 200 times the rate achieved in previous work.