Human motion copy is an intriguing yet challenging task in artificial intelligence and computer vision, which strives to generate a fake video of a target person performing the motion of a source ...person. The problem is inherently challenging due to the subtle human-body texture details to be generated and the temporal consistency to be considered. Existing approaches typically adopt a conventional GAN with an L1 or L2 loss to produce the target fake video, which intrinsically necessitates a large number of training samples that are challenging to acquire. Meanwhile, current methods still have difficulties in attaining realistic image details and temporal consistency, which unfortunately can be easily perceived by human observers. Motivated by this, we try to tackle the issues from three aspects: (1) We constrain pose-to-appearance generation with a perceptual loss and a theoretically motivated Gromov-Wasserstein loss to bridge the gap between pose and appearance. (2) We present an episodic memory module in the pose-to-appearance generation to propel continuous learning that helps the model learn from its past poor generations. We also utilize geometrical cues of the face to optimize facial details and refine each key body part with a dedicated local GAN. (3) We advocate generating the foreground in a sequence-to-sequence manner rather than a single-frame manner, explicitly enforcing temporal inconsistency. Empirical results on five datasets, iPER, ComplexMotion, SoloDance, Fish, and Mouse datasets , demonstrate that our method is capable of generating realistic target videos while precisely copying motion from a source video. Our method significantly outperforms state-of-the-art approaches and gains 7.2% and 12.4% improvements in PSNR and FID respectively.
Deepfakes—artificial but hyper-realistic video, audio, and images created by algorithms—are one of the latest technological developments in artificial intelligence. Amplified by the speed and scope ...of social media, they can quickly reach millions of people and result in a wide range of marketplace deceptions. However, extant understandings of deepfakes’ implications in the marketplace are limited and fragmented. Against this background, we develop insights into the significance of deepfakes for firms and consumers—the threats they pose, how to mitigate those threats, and the opportunities they present. Our findings indicate that the main risks to firms include damage to image, reputation, and trustworthiness and the rapid obsolescence of existing technologies. However, consumers may also suffer blackmail, bullying, defamation, harassment, identity theft, intimidation, and revenge porn. We then accumulate and present knowledge on the strategies and mechanisms to safeguard against deepfake-based marketplace deception. Furthermore, we uncover and report the various legitimate opportunities offered by this new technology. Finally, we present an agenda for future research in this emergent and highly critical area.
Currently, face-swapping deepfake techniques are widely spread, generating a significant number of highly realistic fake videos that threaten the privacy of people and countries. Due to their ...devastating impacts on the world, distinguishing between real and deepfake videos has become a fundamental issue. This paper presents a new deepfake detection method: you only look once–convolutional neural network–extreme gradient boosting (YOLO-CNN-XGBoost). The YOLO face detector is employed to extract the face area from video frames, while the InceptionResNetV2 CNN is utilized to extract features from these faces. These features are fed into the XGBoost that works as a recognizer on the top level of the CNN network. The proposed method achieves 90.62% of an area under the receiver operating characteristic curve (AUC), 90.73% accuracy, 93.53% specificity, 85.39% sensitivity, 85.39% recall, 87.36% precision, and 86.36% F1-measure on the CelebDF-FaceForencics++ (c23) merged dataset. The experimental study confirms the superiority of the presented method as compared to the state-of-the-art methods.
Nowadays, multimedia is vulnerable to hacking because of insecurity. The traditional security mechanism is insufficient to deal with multimedia to protect them against malicious events. So, the ...present study has introduced a novel grey wolf-based YOLO spatiotemporal framework (GW-YSTF) for predicting frames, whether it is fake or real from the trained video data. After initializing the data, the function pre-processing is activated in the hidden layer of the GW-YSTF to eliminate the noisy features in the introduced video frames. Then, a feature analysis function was performed to select the needed parts. Henceforth, the fake video frames are predicted based on the different classes in the trained deepfake video database. Moreover, the presented model is tested in the Python environment. The improvement measure was validated in comparative analysis by comparing the proposed model performance with other existing models based on accuracy, recall, F-score, and precision. The proposed model has recorded the most comprehensive fake score for the accuracy of video frame prediction of 99.8%, higher than the traditional approaches.
While a fast-growing body of research is concerned with the detrimental consequences of disinformation for democracy, the role of visuals in this context has so far only been discussed superficially. ...Visuals are expected to amplify the impact of disinformation, but it is rarely specified how, and what exactly distinguishes them from text. This article is one of the first to treat visual disinformation as its own type of falsehood, arguing that it differs from textual disinformation in its production, processing and effects. We suggest that visual disinformation is determined by varying levels of modal richness and manipulative sophistication. Because manipulated visuals are processed differently on a psychological level, they have unique effects on citizens’ behaviours and attitudes.
Deepfake technology presents serious cybersecurity challenges that have become more prevalent with the availability of easily accessible applications. An effective method for detection and prevention ...of this is necessary. This paper introduces a robust approach and software implementation to detect fake videos constructed with Deep Learning technology that depends on utilizing teeth and mouth movement as distinguishing features that remain very difficult to perfect when faking videos. The proposed methodology has a higher efficiency and accuracy of fake video detection than similar approaches. The work in this article is an extension of previous work that introduced the main concepts with additional application of multi-transfer learning approaches including DenseNet121, DenseNet169, EfficientNetB0, EfficientNetB7, InceptionV3, MobileNet, ResNet50, vgg16, vgg19 and Xception to enhance the algorithm’s ability to detect and classify Deepfake videos based on features extracted from the teeth and mouth frames as a biological signal.
Deepfake is a technology that creates fake images and videos with replaced or synthesized faces. Deepfakes are becoming a concerning social phenomenon, as they can be maliciously used to generate ...false political news, disseminate dangerous information, falsify electronic evidence, and commit digital harassment and fraud. The ease and accuracy of creating Deepfakes have been bolstered by the popularity of wearing face masks since the beginning of the infectious disease outbreak (2020). Because these masks obstruct defining facial features, fake videos are now even more challenging to identify, increasing the necessity for advanced Deepfake detection technology. The research also creates a real/fake video dataset with face masks because the field lacks the dataset required for detection-model training. The proposed research proposes a Deepfake Face Mask Dataset (DFFMD) based on a novel Inception-ResNet-v2 with preprocessing stages, feature-based, residual connection, and batch normalization. The combination of preprocessing stages, feature-based, residual connection, and batch normalization increases the detection accuracy of deepfake videos in the presence of facemasks, unlike the traditional methods. The study's results compared with existing state-of-the-art methods detect face-mask-Deepfakes with 99.81% accuracy compared to the traditional InceptionResNetV2 and VGG19, whose accuracy is 77.48%, and 99.25%, respectively. Future work should evaluate the accuracy of developing a subsequent experimental work for increased detection of deepfake with facemasks.
The emergence of technologies based on artificial intelligence is accelerating the digital transformation of media organizations, directly impacting work processes, the relationship with audiences, ...the generation of content, and the emergence of new professional profiles. It is also, and notably, transforming the processes of detecting and verifying false content. This descriptive-exploratory research analyzes the impact that the use of AI is having on the transformation of the public entity Radiotelevisión Española (RTVE). Through a literature review and interviews with RTVE executives and experts, it reveals the transformative impact of AI on the corporation, highlighting its use to generate new content and verify the authenticity of fake and deepfake videos. In this area, RTVE combines traditional methodologies with others based on AI and leads the development of several tools in collaboration with several universities. These tools have already yielded satisfactory results in detecting these misleading materials, reinforcing RTVE’s role as a guarantor of the veracity of information and increasing citizens’ trust in its content. Similarly, AI is reinforcing RTVE’s identity as a public service by facilitating the generation of automated content, which guarantees access to information in depopulated territories, and others that connect new generations with cultural content. The arrival of artificial intelligence will also generate in a short time a transformation of profiles and professional roles that adapt to this new reality.
Recent progress of artificial intelligence makes it easier to edit facial movements in videos or create face substitutions, bringing new challenges to anti-fake-faces techniques. Although multimedia ...forensics provides many detection algorithms from a traditional point of view, it is increasingly hard to discriminate the fake videos from real ones while they become more sophisticated and plausible with updated forgery technologies. In this paper, we introduce a motion discrepancy based method that can effectively differentiate AI-generated fake videos from real ones. The amplitude of face motions in videos is first magnified, and fake videos will show more serious distortion or flicker than the pristine videos. We pre-trained a deep CNN on frames extracted from the training videos and the output vectors of the frame sequences are used as input of an LSTM at secondary training stage. Our approach is evaluated over a large fake video dataset Faceforensics++ produced by various advanced generation technologies, it shows superior performance contrasted to existing pixel-based fake video forensics approaches.