Wireless communication systems are increasingly vulnerable to jamming attacks due to the open nature of the medium. Traditional anti-jamming techniques such as Frequency Hopping Spread Spectrum (FHSS) and Direct Sequence Spread Spectrum (DSSS) are resource-intensive and fail when a jammer disrupts the entire frequency band. This paper presents DeNoise, a lightweight, protocol-agnostic anti-jamming approach that leverages message repetition and ensemble averaging to mitigate any zero-mean random interference, with Additive White Gaussian Noise (AWGN) being the most common and most practically relevant instance. Unlike conventional methods, DeNoise enables successful message decoding even under full-spectrum jamming, at the cost of a reduced bit rate. We evaluate DeNoise through extensive real-world Software-Defined Radio experiments spanning over 720 million I/Q samples across 96 hardware configurations – six jammer waveforms (AWGN, Uniform, Laplacian, Impulse, Narrowband, single-Tone), four spatial scenarios, two jamming powers, and BPSK/QPSK modulations – and complement these with a hybrid Rayleigh/Jakes fading sweep over fd∈{0,5,20,50,200,500,1000} Hz. In the worst-case spatial configuration (5m jammer-Rx, 10m Tx-Rx) under jamming at ＋20dB above the legitimate signal, BPSK BER drops from 31.6% at a single transmission to 2.1% with 20 repetitions, and QPSK from 38.0% to 4.98%, with consistent improvements across all six jammer waveforms. DeNoise thus offers a robust, tunable solution for restoring communication reliability under aggressive zero-mean jamming.
