The inherent vulnerability of wireless communication necessitates strategies to enhance its security, particularly in the face of jamming attacks. This paper uses the collaborations of multiple sensing nodes (SNs) in the wireless network to present a cooperative anti-jamming approach (CAJ) designed to neutralize the impact of jamming attacks. We propose an eigenvector (EV) method to estimate the direction of the channel vector from pilot symbols. Through our analysis, we demonstrate that with an adequate number of pilot symbols, the performance of the proposed EV method is comparable to the scenario where the perfect channel state information (CSI) is utilized. Both analytical formulas and simulations illustrate the excellent performance of the proposed EV-CAJ under strong jamming signals. Considering severe jamming, the proposed EV-CAJ method exhibits only a 0.7 dB degradation compared to the case without jamming especially when the number of SNs is significantly larger than the number of jamming nodes (JNs). Moreover, the extension of the proposed method can handle multiple jammers at the expense of degrees of freedom (DoF). We also investigate the method’s ability to remain robust in fast-fading channels with different coherence times. Our proposed approach demonstrates good resilience, particularly when the ratio of the channel’s coherence time to the time frame is small. This is especially important in the case of mobile jammers with large Doppler shifts.

Wireless communication can be simply subjected to malicious attacks due to its open nature and shared medium. Detecting jamming attacks is the first and necessary step to adopt the anti-jamming strategies. This paper presents novel cooperative jamming detection methods that use the low-rank structure of the received signal matrix. We employed the likelihood ratio test to propose detectors for various scenarios. We regarded several scenarios with different numbers of friendly and jamming nodes and different levels of available statistical information on noise. We also provided an analytical examination of the false alarm performance of one of the proposed detectors, which can be used to adjust the detection threshold. We discussed the synthetic signal generation and the Monte Carlo (MC)-based threshold setting method, where knowledge of the distribution of the jamming-free signal, as well as several parameters such as noise variance and channel state information (CSI), is required to accurately generate synthetic signals for threshold estimation. Extensive simulations reveal that the proposed detectors outperform several existing methods, offering robust and accurate jamming detection in a collaborative network of sensing nodes.

Selection of hyperparameters in deep neural networks is a challenging problem due to the wide search space and emergence of various layers with specific hyperparameters. There exists an absence of consideration for the neural architecture selection of convolutional neural networks (CNNs) for spectrum sensing. Here, we develop a method using reinforcement learning and Q-learning to systematically search and evaluate various architectures for generated datasets including different signals and channels in the spectrum sensing problem. We show by extensive simulations that CNN-based detectors proposed by our developed method outperform several detectors in the literature. For the most complex dataset, the proposed approach provides 9% enhancement in accuracy at the cost of higher computational complexity. Furthermore, a novel method using multi-armed bandit model for selection of the sensing time is proposed to achieve higher throughput and accuracy while minimizing the consumed energy. The method dynamically adjusts the sensing time under the time-varying condition of the channel without prior information. We demonstrate through a simulated scenario that the proposed method improves the achieved reward by about 20% compared to the conventional policies. Consequently, this study effectively manages the selection of important hyperparameters for CNN-based detectors offering superior performance of cognitive radio network.

In this paper, we consider spectrum sensing (SS) problems with various general noise models such as Middleton class A (MCA), isometric complex symmetric$\alpha $-stable ($\text{S}\alpha \text{S}$), and isometric complex generalized Gaussian distribution (CGGD). This approach enables us to examine the effect of practical phenomena such as impulsive noise on SS problems. In this general framework, we propose a detector based on convolutional neural networks (CNNs) with favorable performance under various noise models. The proposed model-free and data-driven CNN offers robustness in diverse noise scenarios. Thus, it can be utilized in environments with different physical behaviors. We demonstrate this method outperforms the highly regarded likelihood ratio test (LRT) in most cases. For all impulsive cases, the proposed CNN is the superior detector, providing a near-optimum performance for the conventional Gaussian noise. We indicate the proposed data-driven CNN offers an appropriate alternative solution to LRT. However, it requires more computational operations, a rich training dataset, and a training process, instead. Furthermore, the main rationale for proposing this CNN is that it enables the network to generalize its effective performance to various noise models and cases. To this end, quantitative simulations confirm superiority of the proposed CNN compared to other recent deep-learning methods.

The key role of spectrum sensing in cognitive radios attracted substantial research attention to improve the performance of detectors. We consider a general model for the receiver noise with the potential of generalization towards modeling noise in various environments. This general noise model describes accurately noise characteristics ranging from the Gaussian noise to the severe impulsive noise. However, many previous studies are based on the ideal Gaussian noise, and they cannot capture the non-Gaussian models. To provide a robust detector against different behaviors of the noise in various environments, we employ a convolutional neural network (CNN) compatible with different noise models. The proposed CNN detector is data-driven, and due to its single-dimensional input layer, it is consistent with the received signal and requires no pre-processing. The likelihood ratio test (LRT), the Wald, and the Rao tests for this problem are derived to enrich the paper with comparative evaluations of the proposed CNN and conventional model-based approaches and other neural networks. Although various simulated scenarios substantiate the general superiority and robustness of the CNN-based against impulsive noise and mismatch of parameters, it requires higher computational complexity than other discussed detectors.

We investigate the spectrum sensing problem in a single-input-multiple-output cognitive radio system under in-phase and quadrature imbalance (IQI). To do this, the received signal model is obtained in the presence of IQI at the analog front-ends of both the primary and secondary users. We first use a preprocessing stage to mitigate the effect of receiver IQI (RX-IQI). It is found that the received matrix signal is composed of an unknown rank-2 matrix contaminated by the receiver noise matrix. Then, we apply the likelihood ratio test principle to develop three eigenvalue-based detectors. Analytical expressions for the false alarm probabilities of the proposed detectors are derived to adjust their detection thresholds, especially under noise variance uncertainty. Simulation results are provided to illustrate the spectrum sensing performance of the proposed detectors in comparison with existing detectors in the literature also to verify the analytical expressions derived for the false alarm probability. Our simulation results show that the proposed spectrum sensing algorithms are not susceptible to the RX-IQI effects. In addition, we analytically show that one of the proposed spectrum sensing algorithms has constant false alarm rate property with respect to the noise variance uncertainty.