Posted by Admin: System Admin
Over the last few decades, rapid progress in AI, machine learning, and deep learning has resulted in new techniques and various tools for manipulating multimedia. Though the technology has been mostly used in legitimate applications such as for entertainment and education, etc., malicious users have also exploited them for unlawful or nefarious purposes. For example, high-quality and realistic fake videos, images, or audios have been created to spread misinformation and propaganda, foment political discord and hate, or even harass and blackmail people. The manipulated, high-quality and realistic videos have become known recently as Deepfake. Various approaches have since been described in the literature to deal with the problems raised by Deepfake. To provide an updated overview of the research works in Deepfake detection, we conduct a systematic literature review (SLR) in this paper, summarizing 112 relevant articles from 2018 to 2020 that presented a variety of methodologies. We analyze them by grouping them into four different categories: deep learning-based techniques, classical machine learning-based methods, statistical techniques, and blockchain-based techniques. We also evaluate the performance of the detection capability of the various methods with respect to different datasets and conclude that the deep learning-based methods outperform other methods in Deepfake detection. Machine learning is an important component of the growing field of data science. Through the use of statistical methods, different type of algorithms is trained to make classifications or predictions, and to uncover key insights in this project. These insights subsequently drive decision making within applications and businesses, ideally impacting key growth metrics. Machine learning algorithms build a model based on this project data, known as training data, in order to make predictions or decisions without being explicitly programmed to do so. Machine learning algorithms are used in a wide variety of datasets, where it is difficult or unfeasible to develop conventional algorithms to perform the needed tasks.
In [22], the consistency of the biological signs are measured along with the spatial and temporal [23]_[25] directions to use various landmark [26] points of the face (e.g., eyes, nose, mouth, etc.) as unique features for authenticating the legitimacy of GANs generated videos or images. Similar characteristics are also visible in Deepfake videos, which can be discovered by approximating the 3D head pose [27]. In most cases, facial expressions are associated initially with the head's movements. Habeeba et al. [88] applied MLP to detect Deepfake video with very little computing power by exploiting visual artifacts in the face region. As far as the performance concern in machine learning based Deepfake methods, it is observed that these approaches can achieve up to 98% accuracy in detecting Deepfakes. However, the performance entirely relies on the type of dataset, the selected features, and the alignment between the train and test sets. The study can obtain a higher result when the experiment uses a similar dataset by splitting it into a certain level of ratio, for example, 80% for a train set and 20% for a test set. The unrelated dataset drops the performance close to 50%, which is an arbitrary assumption. Zhang et al. [33] introduced a GAN simulator that replicates collective GAN-image artifacts and feeds them as input to a classifier to identify them as Deepfake. Zhou et al. [34] proposed a network for extracting the standard features from RGB data, while [35] proposed a similar but generic resolution. Besides, in [36]_[38], researchers proposed a new detection framework based on physiological measurement, for example, Heartbeat. At first, the deep learning-based method was proposed in [40] for Deepfake video detection. Two inception modules, (i) Meso-4 and (ii) MesoInception-4, were used to build their proposed network. In this technique, the mean squared error (MSE) between the actual and expected labels is used as the loss function for training. An enhancement of Meso-4 has been proposed in [41]. Disadvantages High dimensional features are not preserved as a binary coded structure. The information is not stored in a permission-based Blockchain, which gives the owner control over its contents.
_ We perform a comprehensive survey on existing literature in the Deepfake domain. We report current tools, techniques, and datasets for Deepfake detection-related research by posing some research questions. _ We introduce a taxonomy that classifies Deepfake detection techniques in four categories with an overview of different categories and related features, which is novel and the first of its kind. _ We conduct an in-depth analysis of the primary studies' experimental evidence. Also, we evaluate the performance of various Deepfake detection methods using different measurement metrics. _ We highlight a few observations and deliver some guidelines on Deepfake detection that might help future research and practices in this spectrum. Advantages Presents a generic framework based on Blockchain technology by setting up a proof of digital content's authenticity to its trusted source. Presents the proposed solution's architecture and design details to control and administrate the interactions and transactions among participants. Integrates the critical features of IPFS [114]-based decentralized storage ability to Blockchain-based Ethereum Name service.