DeepWiFi: Cognitive WiFi with Deep Learning

We present the DeepWiFi protocol, which hardens baseline WiFi (IEEE 802.11ac) with deep learning and sustains high throughput by mitigating out-of-network interference. is interoperable builds upon existing WiFi's PHY transceiver chain without changing MAC frame format. Users run for: i) RF front end processing; ii) spectrum sensing signal classification; iii) authentication; iv) channel selection access; v) power control; vi) modulation coding scheme (MCS) adaptation; vii) routing. mitigates effects of probabilistic, sensing-based, adaptive jammers. processing applies a learning-based autoencoder to extract spectrum-representative features. Then neural network trained classify waveforms reliably as idle, WiFi, or jammer. Utilizing labels, users effectively access idle jammed channels, while avoiding interference legitimate transmissions (authenticated machine fingerprinting) resulting in higher throughput. optimize their transmit for low probability intercept/detection MCS maximize link rates used backpressure algorithm Supported embedded platform implementation, provides major gains compared another jamming-resistant especially when channels are likely be signal-to-interference-plus-noise-ratio low.