The rise of deepfake technology has given rise to significant cyber security challenges, with easily accessible applications enabling the creation of convincing fraudulent videos. To address this, a pioneering approach is introduced in this paper, leveraging unique features like teeth and mouth movement as distinct indicators for video authenticity. These attributes are significantly problematic to repeat accurately, creating the proposed approach precisely and efficiently in identifying the videos of deepfake, surpassing present techniques. In recent times, fake videos pressurize several industries, from reportage to politics, hence this ingenious solution presents a convincing mechanism for supporting the visual content ethics and improving cybersecurity standards against growing digital extortions. False material created with deep learning algorithms is known as a “deepfake. Usually, they are in the form of short videos, but they can also be images or sounds. The complexity of deepfake material is rising in tandem with the rapid evolution of AI technologies and the decreasing efficacy of current detection methods. The majority of detection techniques now in use in the literature are AI-based, passive algorithms. In this paper, we present a proof-of-concept deepfake detection system that can distinguish fake news from voice-impersonated video clips.