Deception detectors are used as aids in crime detection and military security systems. Artificial intelligence has been used to classify deceptive and truthful messages using physiological parameters recorded from individuals. This study aims to utilize deep learning networks to detect deception using the extracted temperature signal on the face. Facial thermal data sets of 32 subjects were recorded. Data sets in the time of a mock crime scenario in two groups of deceptive and truthful using the control question test were divided. The five temperature signals were extracted from thermal videos by averaging the 30% of the maximum and minimum of the region of interests (periorbital, forehead, cheeks, perinasal, chin). This signal was used as the input of the deep learning network. The long short-term memory network was used to classify the extracted signals. We evaluated the network performance using sensitivity, specificity, and accuracy criterias. The obtained accuracy value was 93.75%, specificity 96.87%, sensitivity 90.63% which shows notable results compared to previous research.