Lucid dreaming (LD) is a state of conscious awareness of the current dream state, predominantly associated with REM sleep. Research progress in uncovering the neurobiological basis of LD has been hindered by low sample sizes, diverse EEG setups, and specific artifact issues like saccadic eye movements and signal non-stationarity. To address these matters, we developed a multi-stage preprocessing pipeline that integrates standardized early-stage preprocessing, artifact subspace reconstruction, and signal-space projection. This approach enhanced data quality by precisely removing saccadic potential effects even in setups with minimal channels. To uncover the electrophysiological correlates of LD, we applied this methodology to LD data collected across laboratories and explored sensor- and source-level markers hypothesized to underlie LD. Compared to non-lucid REM sleep, in line with recent findings we observed few robust differences on the EEG sensor level. In contrast, on the source level, beta power (12-30 Hz) was reduced during LD in right central and parietal areas including the temporo-parietal junction, potentially associated with a conscious reassessment of the veridicality of the currently perceived reality. Gamma1 power (30-36 Hz) around the onset of LD eye signaling increased in right temporo-occipital regions including the right precuneus, in line with its involvement in self-referential thinking. Source-level connectivity analyses revealed alpha (8-12 Hz) mediated communication between anterior frontal and posterior areas, which are usually functionally disconnected during non-lucid REM sleep. Taken together, these findings illuminate the electrophysiological correlates of LD, laying further groundwork for decoding the mechanisms of this intriguing state of consciousness.