Abstract:
"Deception detection is an important task carried out in high skate situations like suspect
interrogations. Criminal investigations are solely dependent on a suspect’s truthful confession
for a case to be solved. The only way to make a suspect confess is by conducting a suspect
interrogation with evidence. Detectives face difficulties in detecting deception manually due to
various reasons such as excessive time requirement, not specialized in deception detection,
judgements become human biased, necessity of more detective resources. Due to these reasons
manual deception detection has become a hard process to conduct. Therefore, researchers have
identified different automated deception detection approaches. These approaches also have
many limitations which leads to inaccurate and invalid results.
There is a necessity of an accurate, valid and high performing automated deception detection
system. For this reason, Caught in a Lie: an automated deception detection system using human
facial cues and facial chirality has been introduced. The system takes nonverbal indicators of
humans such as facial cues and facial chirality into account and detect deception of interrogation
videos. Human face is a display of emotions people feel. Hence, it is hard to hide or change nonverbal indicators. That is the main reason non-verbal indicators were focused in deception
detection on the system Caught in a Lie. To the best of the author’s knowledge, a Facial
Chirality deep learning model has not been used for any automated deception detection systems
up to this point. However, manual deception detection focuses on facial chirality which gives
accurate results.
The proposed system detects deception of interrogation videos accurately and efficiently by
outputting the deceptive and truthful percentage of the suspect as well as which indicators caused
the deceptiveness or truthfulness of the suspect. Therefore, the Caught in a Lie system is a
novel automated deception detection system which is the first to focus on facial chirality when
detecting deception. This facial chirality model will assist future researchers of the domain or
different domains which require facial chirality."
