Deep Trace: Unravelling Deepfakes Through LSTM Analysis

Abstract

Due to the intense rising of deepfake data which is causing huge risk in maintaining and protecting the consistency of different variety of data like video, audio, images and some interactive elements. These deepfakes are also harming the wellbeing and harmony of the society. Therefore, this paper brings in solution for detecting deepfakes by employing combination of different technology is called dual architecture. The spectaculated structure flawlessly, unified both Res-Net and LSTM with CNN. The Res-Net mainly emphasizes on categorizing and distinguishing between original or manipulated images. Additionally, Res-Net filter- outs patterns from images which includes facial contours and embedded text. LSTM with CNN unit pin points on temporal coherence and it also recognizes the realness of the image. Transfer Learning ways are used where understanding and intelligence is gained by training the dataset to magnify the functionality and performance of the spectaculated structure. In, Overview the entire deep fake detection system demonstrates unified collaboration of Res-Net and LSTM- CNN for delivering, a effective deep fake system which helps from attacks that is being created from AI tools. From the proposed model by increasing the accuracy and robustness a reliable system is generated to solve the attacks happening in digital world.