False Information Detection with Deep Learning

As we know, Artificial Intelligence (AI) technology is increasingly advanced and developing globally. AI technology has been widely used in many fields, one of which is digital forensics. With the development of AI, there are also many negative impacts such as deepfakes and voice cloning. This research aims to identify several deep learning models adapted for this detection task. The study also seeks to assess the effectiveness of various deep learning approaches in identifying deepfakes and deeper voice cloning, which can be a reliable technique to combat the spread of deepfakes and voice cloning as well. In this research we focus on looking for several models that can detect information manipulation by AI. In this experiment we want to between 2 different datasets for deepfakes and do several iterations to test the datasets. And for voice cloning, we want to test with 2 different methods, namely CNN and LSTM which are carried out in several iterations to get accurate results. The results obtained, for deepfakes, the quality of the image affects the results. For high quality, the highest percentage is 84.87% and low quality is 88.89%. Meanwhile, the voice AI model LSTM shows a higher percentage figure than CNN, namely 99.30%.