In the controlled source electromagnetic (CSEM) signal processing, random noise introduced by measurement devices and human activities can significantly impact data process and interpretation. To extract effective signals, algorithmic noise suppression is more efficient than hardware-based solutions. Wavelet thresholding algorithms are favored by geophysicists for their superior performance. The selection of the wavelet basis, thresholding parameters, and decomposition level plays a critical role, but it often requires prior knowledge or extensive tuning. To address this, an improved wavelet thresholding approach for random noise suppression is proposed: sparrow search algorithm-based improved wavelet thresholding (SSA-IWT). This method: (1) aims to maximize the normalized cross-correlation (NCC) between the denoised signal and the transmitted signal, using SSA to automatically optimize the best [decomposition level, wavelet basis, threshold]；(2) adopts signal decomposition and uses the NCC between the denoised and transmitted signals as a measurement indicator, which eliminates the need for extensive field measurements from remote, low-interference areas. The results of simulations and measured data processing demonstrate that SSA-IWT outperforms other conventional approaches.