Energy detection and machine learning for the identification of wireless MAC technologies

ISM spectrum is becoming increasingly populated with various wireless technologies, rendering it a scarce resource. Consequently, wireless coexistence is increasingly vulnerable to new wireless devices attempting to access the same spectrum. This paper presents a novel method for identifying wireless technologies through the use of simple energy detection techniques. Energy detection is used to measure the channel statistical temporal characteristics including activity and inactivity probability distributions. Features uniquely belonging to specific wireless technologies are extracted from the probability distributions and fed into a machine-learning algorithm to identify the technologies under evaluation. Wireless technology identification enables situational awareness to improve coexistence and reduce interference among the devices. An intelligent wireless device is capable of detecting wireless technologies operating within same vicinity. This can be performed by scanning energy levels without the need for signal demodulation and decoding. In this work, a wireless technology identification algorithm was assessed experimentally. Temporal traffic pattern for 802.11b/g/n homogeneous and heterogeneous networks were measured and used as algorithm input. Identification accuracies of up to 96.83% and 85.9% were achieved for homogeneous and heterogeneous networks, respectively.