As the navigation downlink requires higher and higher information transmission rates, the use of code shift keying (CSK) modulation to increase the transmission rate of navigation messages instead of traditional BPSK and BOC modulation has gradually attracted attention. Non-binary low density parity check (NLDPC) codes can effectively improve the reliability of CSK, and the introduction of an iterative detection mechanism can achieve a compromise between reliability and complexity of the NLDPC-CSK scheme. However, the complexity of the detection mechanism based on symbol probability is still very high. Therefore, we propose a low-complexity iterative detection algorithm with probabilistic information truncation, which adaptively adjusts the decoder and demodulator by reducing the dimension of the matrix that transmits external information. The decoding selects the extended minimum sum (EMS) algorithm that shortens the probability information to replace Q-ary sum product algorithm (QSPA), and CSK uses the external information output by the decoding and the index value for demodulation. Theoretical analysis and simulation results show that selecting a proper truncation length for the EMS algorithm can greatly reduce the decoding complexity. The improved CSK (ICSK) demodulation indexes the corresponding spreading code sequence through the external information, and does not need to perform correlation operations with all shifted versions of the basic spreading code. The demodulation complexity decreases linearly with the decrease of the external information dimension. Finally, the proposed iterative detection algorithm greatly reduces the complexity of the algorithm under the transmission with little loss of reliability of the navigation messages.