In recent times, the proliferation of deepfake videos, generated through sophisticated deep learning algorithms, has garnered significant attention. These videos, capable of convincingly manipulating faces, have inundated the internet, often targeting celebrities and politicians. This trend poses a grave threat to social stability as these manipulated videos are frequently employed to tarnish reputations and manipulate public opinion. People can simply learn how to create deepfake films using a small number of victims or target photos with little to no effort. As a result, numerous research initiatives have been launched, with a primary emphasis on refining detection techniques and creating thorough benchmarks within the academic domain. In this paper, we have reviewed some of the recent works done in deepfake detection, through deep learning methods. Additionally, we also look at the benchmark datasets in this domain, highlighting potential avenues for future research.