The existing narrowband interference (NBI) detection and parameter estimation algorithms for direct sequence spread spectrum (DSSS) communications based on the Nyquist sampling theorem are confined to the high sampling rate. In order to solve this problem, the compressive sensing is used to the NBI detection and parameter estimation in DSSS communications, a newly emerged sparse approximation technique, block sparse Bayesian learning, is utilized to get the NBI feature vector from the compressed signal using the different features of DSSS signals and NBI in the compressed domain and the block sparsity feature of NBI in the frequency domain. The NBI detection and parameter estimation are realized by detecting and estimating parameters of the feature vector. Reported simulation results demonstrate that the proposed method is effective in the NBI detection and parameter estimation in DSSS communications, and significantly outperforms other conventional methods. The performance is mainly affected by the variety of interference intensity, interference bandwidth and compression ratio. The larger the interference intensity is, the smaller the interference bandwidth is and the greater the compression ratio is, the better the interference detection and parameter estimation performance are.