Electric Network Frequency (ENF) fluctuations based forensic analysis is recently proposed for time-of-recording estimation, timestamp verification, and clip insertion/deletion forgery detection in multimedia recordings. Due to the load control mechanism of the electric grid, ENF fluctuations exhibit pseudo-periodic behavior and generally require a long duration of recording for forensic analysis. In this paper, a statistical study of the ENF signal is conducted to model it using an autoregressive process. The proposed model is used to understand the effect of the ENF signal duration and signal-to-noise ratio on the detection performance of a timestamp verification system under a hypothesis detection framework. Based on the proposed model, a decorrelation based approach is studied to match the ENF signals for timestamp verification. The proposed approach requires a shorter duration of the ENF signal to achieve the same detection performance as without decorrelation. Experiments are conducted on audio data to demonstrate an improvement in the detection performance of the proposed approach.