In diffusion-based molecular communication (MC), the most common modulation technique is based on the concentration of information molecules. However, the random delay of molecules due to the channel with memory causes severe inter-symbol interference (ISI) among consecutive signals. In this paper, we propose a detection technique for demodulating signals, the increase detection algorithm (IDA), to improve the reliability of concentration-encoded diffusion-based molecular communication. The proposed IDA detects an increase (i.e., a relative concentration value) in molecule concentration to extract the information instead of detecting an absolute concentration value. To validate the availability of IDA, we establish a real physical tabletop testbed. And we evaluate the proposed demodulation technique using bit error rate (BER) and demonstrate by the tabletop molecular communication platform that the proposed IDA successfully minimizes and even isolates ISI so that a lower BER is achieved than the common demodulation technique.