Development of AI/ML based solution for detection of face-swap based deep fake videos

Source summary

Develop a forensic AI/ML detector specifically for face-swap videos. The source presents possible approaches including spatial and temporal inconsistencies, pose/texture analysis, adversarial training, audio-visual mismatch and provenance records. These are alternatives or complementary methods to investigate, not a mandatory stack. For an input video, produce an explanatory report describing detected abnormalities and the inferred manipulation characteristics.