Detection and Verification for Deepfake Bypassed Facial Feature Authentication

Facial features have become one of the leading standards for biometric authentication. With the growing demand for the Internet of Things, it has become imperative to have systems to detect any data breach used for authentication. New deep fake technologies have made it possible to compromise the database, making real-time authentication inefficient. The counterfeit to the real-time data can be challenging to detect. We have developed a deep fake image detection utilizing ResNet50 and Spatial Pyramidal Pooling to tackle this problem. Our model detects manipulation in real-time data, irrespective of its size and background, at an accuracy of 94%. This model will allow us to detect bypass in the database for facial biometric authentication and improve the system’s security.