Channel Input Design for Secure Transmission over Fast Fading Channels with Statistical-CSIT

• Applicant: Dr.-Ing. Pin-Hsun Lin, Ph.D.
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
• Term: from 2017 to 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 329361154

In this project fundamental problems of the holistic system design of secure communication systems will be studied using fast fading channels with statistic CSI at the transmitter. Particular focus is placed on the channel input. At first, new criteria will be designed to identify and characterize the degradedness of wiretap channels to avoid complicated channel prefixing. Secondly, optimized signal processing will be developed regarding the channel input distribution and channel prefixing to achieve higher secrecy rates in the non degraded case.In this project, the widespread, but in many cases unrealistic assumption of perfect channel knowledge is dropped. Instead, statistical channel knowledge will be assumed, which can be obtained a priori. When Alice transmits messages to her legitimate communication partner, she can treat the closest but unknown user as an eavesdropper and use the statistical channel knowledge to generate her wiretap codebook. Apart from a few special cases, generally the channel input distribution is unknown in the case of imperfect channel knowledge. These will be examined in the course of the project. In case of imperfect knowledge of the eavesdropper's channel, artificial noise is added to the signal to reduce the signal to noise ratio at the eavesdropper. However, there are several myths about artificial noise. Among others there are the mistaken belief that if the transmitter has not perfect channel knowledge of the eavesdropping channel in case of fast fading channels, artificial noise has to be used, and that the signal and the artificial noise must be Gaussian. For these statements, there are no analytical proofs. On the contrary, they may be incorrect under certain conditions. We examine these myths and derive necessary and sufficient conditions for the use of artificial noise, with respect to the degradedness of the wiretap channel. As long as the wiretap channel is degraded, a signal processing in the form of artificial noise is not necessary; therefore, the system design can be considerably simplified. In the case of non-degraded wiretap channels other signal processing capabilities should be examined instead of the Gaussian distribution of the signal and the artificial noise. In this case, both the input distribution, and the channel prefixing will be considered. The project will therefore develop a theoretically based, systematic signal processing for general wiretap channels with statistical channel knowledge at the transmitter, which will allow efficient system design and efficient resource allocation using physical layer security algorithms.

• DFG Programme: Research Grants