Abstract
In wireless networks, communication links may be subject to random fatal attacks: for example, sensor networks under sudden power losses or cognitive radio networks with unpredictable primary user spectrum occupancy. Under such circumstances, it is critical to quantify how fast and reliably information can be collected over attacked links. In our previous work, we studied such channels by considering the single point-to-point case and introduced the notion of dying channels, where a single dying channel is modeled as a random delay-limited channel. In this paper, we extend the single point-to-point dying channel case to the parallel multi-channel case where each sub-channel is a dying channel. According to different attack models, we investigate the outage performance for the parallel dying channels with two setups: (1) the independent-attack case where the attacks on different sub-channels are independent of each other; (2) the m-dependent-attack case where the random attacks on different sub-channels are correlated in such a manner that only the random attacks within m adjacent sub-channels are correlated. For a target sum rate, we further characterize the asymptotic behavior of the outage probabilities for the above two cases. By the central limit theorems for independent and m-dependent sequences, we show that the outage probability diminishes to zero for both cases when the number of sub-channels increases, where the outage exponents are studied to reveal how fast the outage probability improves.