Terrorist attacks through building ventilation systems are becoming an increasing concern. In case pollutants are intentionally released in a building with mechanical ventilation systems, it is critical to localize the source and characterize its releasing curve. Previous inverse modeling studies have adopted the adjoint probability method to identify the source location and used the Tikhonov regularization method to determine the source releasing profile, but the selection of the prediction model and determination of the regularization parameter remain challenging. These limitations can affect the identification accuracy and prolong the computational time required. To address the difficulties in solving the inverse problems, this work proposed a Markov-chain-oriented inverse approach to identify the temporal release rate and location of a pollutant source in buildings with ventilation systems and validated it in an experimental chamber. In the modified Markov chain, the source term was discrete by each time step, and the pollutant distribution was directly calculated with no iterations. The forward Markov chain was reversed to characterize the intermittently releasing profile by introducing the Tikhonov regularization method, while the regularized parameter was determined by an automatic iterative discrepancy method. The source location was further estimated by adopting the Bayes inference. With chamber experiments, the effectiveness of the proposed inverse model was validated, and the impact of the sensor performance, quantity and placement, as well as pollutant releasing curves on the identification accuracy of the source intensity was explicitly discussed. Results showed that the inverse model can identify the intermittent releasing rate efficiently and promptly, and the identification error for pollutant releasing curves with complex waveforms is about 20%.