Detection of traffic patterns in the radio spectrum for cognitive wireless network management

Abstract

Dynamic Spectrum Access allows using the spectrum opportunistically by identifying wireless technologies sharing the same medium. However, detecting a given technology is, most of the time, not enough to increase spectrum efficiency and mitigate coexistence problems due to radio interference. As a solution, recognizing traffic patterns may lead to select the best time to access the shared spectrum optimally. To this extent, we present a traffic recognition approach that, to the best of our knowledge, is the first non-intrusive method to detect traffic patterns directly from the radio spectrum, contrary to traditional packet-based analysis methods. In particular, we designed a Deep Learning (DL) architecture that differentiates between Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) traffic, burst traffic with different duty cycles, and traffic with varying rates of transmission. As input to these models, we explore the use of images representing the spectrum in time and time-frequency. Furthermore, we present a novel data randomization approach to generate realistic synthetic data that combines two state-of-the-art simulators. Finally, we show that after training and testing our models in the generated dataset, we achieve an accuracy of ≥ 96 % and outperform state-of-the-art methods based on IP-packets with DL.

A General Approach for Traffic Classification in Wireless Networks Using Deep Learning

“…Moreover, these results provide initial insights about the feasibility of this approach for real-time classification. These evaluations complement and extend previous results where the performance of DL approaches solving the TC problem have been evaluated and compared using byte [11] and spectrum representation of the packets [8], [10]. 16-Quadrature Amplitude Modulation (QAM), 64-QAM for 802.11g/n with coding rates of 1/2, 3/4, and 5/6 according to the standard and modulation selected).…”

“…As a solution, a few spectrum-based traffic classifiers have been proposed in recent years to perform TC on raw spectrum data. Authors in [10] present a DL-based algorithm that can classify traffic patterns of different types of applications directly from the radio spectrum with accuracy ≥ 96% and outperform state-of-the-art methods based on IPpackets with DL. They use images representing the spectrum in time and time-frequency as input data for their CNN-based DL architecture.…”

“…Compared to similar works like [8], the proposed procedure includes not only the traffic classifier designing but also the complete chain to achieve spectrum-based TC. Moreover, the proposed approach removes the need for specialized algorithms to separate/aggregate users' traffic flows (e.g., a radio identification procedure [9]), as the one required on recent approaches like [10] , as it uses as classification object single L1 packets (as a sequence of raw IQ samples). • We designed and evaluated a DL architecture based on CNNs to solve the task of classifying packets directly on spectrum data.…”

Artificial Intelligence for Radio Communication Context-Awareness

“…A schematic grouping of various machine learning methods is shown in Figure 4. Below, we outline the basics of these methods and their application to Single node spectrum sensing [68], modulation recognition [69], multi-carrier modulation recognition [70] Neural Networks: CNN Single node spectrum sensing [71], combining sensing results in cooperative spectrum sensing [72], modulation recognition [73], traffic pattern recognition [74] Neural Networks: LSTM Prediction of channel state [75], modulation recognition [76], prediction of PU's next spectrum state [77] Unsupervised learning K-means clustering Single node spectrum sensing [78], cooperative spectrum sensing [64] PCA Modulation recognition [79], multi-carrier modulation recognition [80] Reinforcement learning Q-learning and others Learning spectrum sensing policy [81], cooperative sensing [82] build and enrich context awareness in radio communication systems. The possible applications of different methods for context awareness are displayed in Table 2.…”

Evaluation of the performance of a collaborative proposal of multiple access in cognitive radio networks

“…The related literature uses mathematical traffic models based on random estimations (Camelo et al, 2020;Han et al, 2020) however, although they are detailed traffic models, they do not present evidence of their performance under actual PU behaviors. The model proposed in this research incorporates actual PU behavior within the simulation environment through solid spectral occupancy measurements in the GSM frequency band .…”