The abundance of digital editing tools has made it gradually easier to modify visual content. Criminals and hackers misuse such images and videos for deceptive purposes. Researchers are not only identifying the risks like identity theft and dispersion of falsie content but also seeking Artificial Intelligence based solutions. The proposed work introduces a passive detection method that mainly focuses on identifying tampering in digital videos through specific features. Videos are classified into static and dynamic categories. The Backward Selection Method and Forward Selection Method are used for feature selection to enhance accuracy. An ensemble model based on Isolation Forests and One-Class SVM is employed for outlier detection. This method effectively distinguishes between original and tampered content without relying on embedded data or prior knowledge. Experimental evaluations on a comprehensive video dataset shows that this approach achieves high levels of accuracy, precision, and recall. It offers a robust solution with for the forensic analysis of video content and achieved the 93.0% accuracy. The results highlight the method's potential for use in legal and investigative contexts where the authenticity of visual evidence is critical.