Drift-Resilient LO Leakage Detection in Cognitive Radio: Maximum Likelihood Modeling and Real-World Validation

Local oscillator (LO) leakage detection is pivotal for identifying covert receivers in cognitive radio networks, yet existing methods suffer from performance degradation under frequency drift caused by temperature variations, aging, and manufacturing tolerances. To address this challenge, we propose a drift-resilient LO leakage detector based on maximum likelihood (ML) statistical modeling. Unlike conventional energy detection assuming fixed LO frequencies, our approach explicitly models the time-varying LO leakage as a non-central chi-square distributed process under frequency drift, enabling robust detection across dynamic spectral environments. Experimental validation is conducted on commercial RF systems-on-chip (SoCs) and software-defined radio (SDR) platforms, emulating real-world drift scenarios. Results show a higher detection probability under ± 0.5 kHz drift, outperforming state-of-the-art methods while maintaining a false alarm rate below 10%. With low computation complexity, this practical RF impairment mitigation technique offers a deployable solution for next-generation spectrum-agile networks.