A novel estimating system is proposed to find out the security weaknesses of a random number generator (RNG) based on a chaotic jerk system. Convergence of the estimating system is proved using auto-synchronization. Unknown parameter of the target RNG is revealed where the public information are the design of the RNG and a scalar time series observed from the target chaotic jerk system. Simulation and numerical results verifying the feasibility of the estimating system are given such that, next bit can be predicted while the same output bit sequence of the RNG can be regenerated.