Abstract: Concurrency conflicts may bring data conflict covert channel in multilevel secure systems. The existing covert channel detection methods have the following flaws: 1) Analyzing conflict records with single point, so the invaders can evade to be detected; 2) Using single indicator will bring false positive and false negative. We present a detection method based on conflict interval time called CTIBDA in this paper. This method solves the above problems: 1) Analyzing the conflict records with subject and object can prevent intruders from dispersing; 2) Using both the distribution and the sequence of intervals between transactions conflicts as indicators. The experimental results show that this approach can reduce the false positive and false negative and increase the accuracy. CTIBDA is suitable for online implementation and can be universally applied to concurrency conflict covert channels in other scenarios.