Using Deep Learning to Combat Pilot Jamming in SISO OFDM-based Wireless Networks

Wireless networks are integral to society but remain vulnerable to attacks that disrupt legitimate communication channels. This work presents a Deep Learning (DL)-Modulator/Demodulator (MODEM) that effectively overcomes pilot jamming in Single-Input Single-Output (SISO) Orthogonal Frequency Division Multiplexing (OFDM) wireless networks. The DL-MODEM’s anti-jam and data recovery performance is assessed using an adversary that employs partial and full pilot jamming under noisy Rayleigh fading conditions. The DL-MODEM achieves a superior Bit Error Rate (BER) versus other anti-pilot jamming countermeasures, such as randomizing pilot locations between transmissions. The DL-MODEM’s superiority is pronounced when all pilots are jammed. This emphasizes its efficacy, which is attributed to its adaptability under diverse and time-varying channels. Its advantage is furthered by the fact that it does not require online over-the-air re-training because multiple versions can be trained offline to learn various environmental and jamming conditions sufficiently.