Deepfake technology is a new way to alter digital content and create videos that look very real. The responsible use of deepfake technology is essential, as its inappropriate application can lead to significant consequences, from harming an individual’s reputation to influencing public opinion. Nowadays, this technology is being misused for spreading false information or deceiving people as well, making it crucial to develop an effective method for the detection of synthetic media. The current research focuses on various aspects, such as datasets, features, tools, and techniques used in the field of deepfake detection. Further investigation of gaps like lack of a multi-modal approach, less work on hybrid models, and unseen datasets associated with current research work, has also been done. The existing deep learning models being used for deepfake detection face several challenges. There is no such model that works well with the different types of datasets. Also, the methods used to create deep fakes are changing quickly, making it even more difficult for existing detection models to obtain better performance. In order to overcome the challenges, it is proposed to design a hybrid learning framework for deepfake detection using a multi-modal approach.
