Optimal Power Allocation in Block Fading Gaussian Channels with Causal CSI and Secrecy Constraints

The optimal power allocation that maximizes the secrecy capacity of block fading Gaussian (BF-Gaussian) networks with causal channel state information (CSI), M-block delay tolerance and a frame based power constraint is examined. In particular, we formulate the secrecy capacity maximization as a dynamic program. We propose suitable linear approximations of the secrecy capacity density in the low SNR, the high SNR and the intermediate SNR regimes, according to the overall available power budget. Our findings indicate that when the available power resources are very low (low SNR case) the optimal strategy is a threshold policy. On the other hand when the available power budget is infinite (high SNR case) a constant power policy maximizes the frame secrecy capacity. Finally, when the power budget is finite (medium SNR case), an approximate tractable power allocation policy is derived.