Blind adaptive multiuser detector has become a research hotspot in recent years due to a number of advantages, but many blind adaptive algorithms involve low convergence rate. This paper presents a novel stochastic blind adaptive multiuser detector without requiring training sequences, which needs only two system parameters: the signature sequence of the desired user i, s i and the variance of the additive white Gaussian noise (AWGN),σ 2. Simulation results show that by reasonably choosing time varying step size, the proposed detector can not only improve the convergence rate, but also reduce the limiting NSE (Normalized Squared Error) values, so it can effectively increase the performance of the system. Blind adaptive multiuser detector has become a research hotspot in recent years due to a number of advantages, but many blind adaptive algorithms involve low convergence rate. This paper presents a novel stochastic blind adaptive multiuser detector without requiring training sequences, which needs only two system parameters : the signature sequence of the desired user i, si and the variance of the additive white Gaussian noise (AWGN), σ 2. Simulation results show that by reasonably choosing time varying step size, the proposed detector can not only improve the convergence rate, but also reduce the limiting NSE (Normalized Squared Error) values, so it can able just increase the performance of the system.