Wireless devices are ubiquitous nowadays and, since most of them use the same unlicensed frequency bands, the high number of packet losses due to interference and collisions degrade performance. Reliability, energy consumption, and latency are key challenges for future dense networks. Allowing the transmitter to take action, i.e., vacating the channel, as soon as a collision or interference is detected is crucial in improving these metrics. In-band full duplex radios enable the transmitter to simultaneously transmit packets and sense the spectrum for collisions and interference. This paper studies two important questions regarding transmitter-based collision and interference detection: (1) from an overall system perspective, does such detection outperform receiver-based detection and (2) which test statistic is the most accurate and sensitive at detecting collisions and interference. First, NS-3 simulations are used to show that transmitter-based detection reduces the energy consumption while improving the throughput in a typical star topology network. Next, we present a measurement-based study of four different techniques for transmitter-based collision and interference detection. In particular, we compare the energy detector with three goodness-of-fit tests in terms of probability of detection and false alarm. Our analysis shows that transmitter-based detection can detect between 80% to 100% of the collisions and interference occurring at the receiver, depending on the distance between the transmitter and the receiver. Of those detectable by the transmitter, our measurement results show that goodness-of-fit tests can detect nearly 100% of the collisions and have at least 10 dB better sensitivity as compared to the commonly proposed energy detection test. In general, the proposed techniques can detect interfering signals that are up to 25 dB below the remaining self-interference power.

In this work, we consider the wideband sensing problem of using a single antenna receiver for sensing an unknown number of incumbent signals occupying frequency bands that may be partially overlapping. Many communications standards today, such as the IEEE 802.11 and LTE-Advanced, use frequency bands that overlap amongst themselves or do not have a frequency guard band. Current sensing methods for a single antenna receiver sense for signal energy at individual frequencies and do not differentiate between signals. However, sensing each signal separately is necessary to make inferences about the higher layers of the incumbent communication system. Hence, in this work, signals are sensed as signal energy in a set of frequency bins instead of individual frequency bins. The number of signals and the frequency bands occupied by each are estimated. To achieve this, the temporal dimension is exploited by using periodic power measurements as input and the fact that radios transmit intermittently and not continually. Two algorithms are proposed. First, repeated non- negative matrix factorization is used to estimate the number of signals and the frequency support of each. Secondly, energy detection results on adaptively chosen frequency bins are used to detect the start and end of the frequency band occupied by each signal. Novel performance metrics are proposed since this is simultaneously a model size and parameter estimation problem. Systems with signals occupying adjacent partially overlapping frequency bands are simulated to evaluate the performance of both algorithms.

While cooperative spectrum sensing improves sensing reliability when secondary users (SUs) have independent measurements, these gains are limited in the presence of correlated or malicious SUs. In this paper, we propose algorithms to improve spectrum sensing performance when the system contains both honest and malicious SUs all of whom may be experiencing correlated fading channels. We show that, when the SUs' reports at each time slot are independent and identically distributed, it is impossible to distinguish between malicious collaboration, i.e., collusion, and correlations caused by the environment. Thus, the optimal test statistic for cooperative spectrum sensing depends on the correlations in the SU reports but not on the source of these correlations. We propose two algorithms: one to identify SUs whose reports are correlated and another to infer the spectrum occupancy by using the learned structure of correlations. Groups of SUs whose reports are correlated are identified by learning the structure of the underlying Bayesian network model. This structure is implemented as a factor graph to infer the spectrum occupancy using a loopy belief propagation algorithm. We derive an upper bound of the error probability of the proposed structure learning algorithm and prove convergence of our loopy belief propagation algorithm.

Consider spectrum sensing systems where the presence of the primary user of the spectrum is detected by secondary users (SUs) in a centralized cooperative fashion. The sensing results can be correlated due to environmental reasons. SUs may or may not be honest. If dishonest, they could be colluding. Existing methods to fuse the SU reports either ignore the correlation in the SU reports, or they need to know the source of correlation. In this paper, we propose a belief propagation based fusion algorithm to exploit the correlations in reports of groups of SUs irrespective of cause. We show that identifying the groups of SUs having correlated reports reduces the probability of error of spectrum sensing. Our method is based on modeling the probability distribution underlying the SU reports as a Bayesian network. The process of learning the Bayesian network also shows that it is theoretically impossible to identify collusion.