Alternating CSIT: A Framework for Heterogenous Channel Knowledge.
In my recent work, I have studied the impact of spatio-temporal variations in channel knowledge and how such variations can be harnessed to increase spectral efficiency. In currently deployed systems (such as FDD LTE), channels are estimated at the receivers and are periodically fed back to the transmitters. It is natural to expect that the nature of channel feedback from users may not be uniform (i.e., quality of feedback can vary in the time/frequency grid and also across users). Such spatio-temporal variations are often overlooked in practice and the current designs are based on a myopic approach. Recent work on alternating CSIT develops a framework for modeling this heterogeneity and shows that such variations can be synergistically utilized to increase the network degrees of freedom.
Secure Distributed Cloud Storage with Exact Repair.
I have recently investigated security issues in distributed cloud storage systems. A distributed storage system (DSS) is parameterized by three parameters (n, k, d), in which data is coded across n distributed nodes, and it must be recoverable from any k out of n nodes. This is called as the reconstruction property of a DSS. If a node fails, any d out of (n-1) nodes help in the repair of the failed node so that the regeneration property of the DSS is preserved. From a practically relevant perspective, it is often desirable that that repair process must replace a failed node by its exact replica (referred to as the exact repair property). In this context, I have investigated the fundamental tradeoff between storage and repair bandwidth if the repair process must be information theoretically secure. We have made new progress on the simplest non-trivial instances of this problem which prior to this work have been open, and establish the optimal storage-vs-repair bandwidth tradeoff for several classes of this problem.
Interference Management via Feedback
In my recent work, I have explored the usefulness of feedback for interference management. This aspect is studied for several models, including the two-user multiple-input multiple-output (MIMO) interference channel with feedback and delayed channel state information at the transmitters (CSIT). Rather surprisingly, it turns out that feedback can increase the degrees of freedom of the interference channel even with delayed CSIT. On the contrary, for the MIMO broadcast channel, feedback if available in addition to delayed CSI is not useful. The key to this result is to exploit the fact that feedback induces user cooperation, which leads to an increase in the degrees of freedom for the network.
The benefit of feedback for a K-user cyclic interference network is studied, and it is shown that feedback capacity is a function of K, the number of users. The capacity gain (per-user) obtained via feedback is shown to decrease as K increases, and vanishes in the limit of large K. Feedback capacity is also studied for the fully connected K-user interference channel. Most surprisingly, for this network, it turns out that the feedback gain (per-user) is independent of K, the number of users. The per-user (approximate) feedback capacity for the K-user system is the same as the feedback capacity of a 2-user interference channel. The key idea behind this result is a novel coding scheme, which we call as cooperative interference alignment, which is a combination of interference alignment and interference decoding.
Cooperative Communication and Relaying
In several practical scenarios, the broadcast nature of wireless medium provides rich signal interactions which can enhance throughput of a communication network.
For instance, wireless transceivers can overhear each other signals. Such overheard information can be beneficial in increasing transmission rates. In other instances, noisy feedback
can be present from the receivers to the transmitters. Such models of user cooperation and noisy feedback are studied under a common framework of generalized feedback. Outer bounds are obtained for the capacity regions of the two-user interference and multi-access channels with generalized feedback.
The key ideas behind these results is a generalization of the dependence balance bound, which was initially proposed for the
two-way channel. The results show strict improvements over vanilla cut-set bounds for several models, which include the multiple-access channel (MAC)
with noiseless feedback, the Gaussian MAC with noisy feedback, and also for the interference channel with user-cooperation.
Information Theoretic Security
I have also focused on secure source coding problems, and the usefulness of interaction between terminals to enhance secrecy. In one of my papers, this aspect is explored, where a sender is interested in transmission of a memoryless source X to a legitimate receiver in presence of an eavesdropper. A helper, which has access to a correlated source Y can broadcast information through a secure, rate-limited link to the sender and the legitimate receiver. It turns out that such a secure link solves a dual purpose: a) it lowers the rate required for source transmission, and b) reduces the information leakage about X at the eavesdropper. The interesting aspect of the result is to show that classical source-coding schemes do not achieve the optimal rate-leakage tradeoff. In another setting, I have looked at a situation in which there are two legitimate receivers, one with and the second without correlated side information Y. The sender is interested in describing a source X in a lossy fashion to both receivers while keeping the source Y secret from the second receiver. It turns out that the availability of side-information Y at the sender significantly reduces the leakage of Y at terminal 2. The common availability of side-information Y at the sender and terminal 1 can also be regarded as a form of backward channel between these two terminals.