Optimal Power Allocation in Block Fading Gaussian Channels with Secrecy Constraints

Printer-friendly versionSend by email

Physical layer security (PLS) investigates the potential of exploiting impairments of real communication channels, such as fading and noise, in order to achieve confidentiality in information exchange. PLS was pioneered by Wyner, who introduced the wiretap channel and established the possibility of creating perfectly secure communication links without relying on private (secret) keys. Recently, considerable efforts have been channelled to generalizing this result to the wireless fading channel and to multi-user scenarios. In the present talk we investigate optimal power allocation policies in block fading Gaussian (BF-Gaussian) wireless networks with secrecy and delay constraints. We study networks with both acausal and causal channel state information (CSI) feedback, M-block delay tolerance and frame based power and secrecy constraints.

We begin with the acausal scenario; first, we derive secure waterfilling algorithms and discuss the feasibility of physical layer techniques in cooperative networks. Interestingly, we show that as few as twelve legitimate users suffice to provide enough multi-user diversity in order to transmit securely 1 bit/sec/Hz with a probability of secrecy outage less than 1% in Rayleigh conditions. Then, we formulate the optimal power allocation problem with causal CSI feedback using dynamic programming and derive transmission policies for the low and the high SNR regimes. We demonstrate that when the available power resources are very low the optimal strategy is a “threshold policy”, while when the available power budget is very large a “constant power policy” maximizes the frame secrecy capacity. Finally, we introduce a novel universal approximation in the resource allocation problem - the “blind horizon approximation” (BHA). In the secrecy framework, the novel BHA policy is shown through simulations to outperform all other policies as long as the mean channel gain of the legitimate user is distinctively greater than the mean channel gain of the eavesdropper. Interestingly, the secrecy rates achieved by the BHA policy compare well with the secrecy rates of the secure waterfilling policy in the case of acausal CSI feedback to the transmitter. These results demonstrate that PLS techniques can be used in realistic scenarios with negligible computational overheads.

Mon, 19/05/2014 - 14:00 - 16:00
Main Lecture Room (IIT)

© 2018 - Institute of Informatics and Telecommunications | National Centre for Scientific Research "Demokritos"