DFFMD: A Deepfake Face Mask Dataset for Infectious Disease Era With Deepfake Detection Algorithms

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 16711 - 16722

Published: Jan. 1, 2023

Deepfake is a technology that creates fake images and videos with replaced or synthesized faces. Deepfakes are becoming concerning social phenomenon, as they can be maliciously used to generate false political news, disseminate dangerous information, falsify electronic evidence, commit digital harassment fraud. The ease accuracy of creating have been bolstered by the popularity wearing face masks since beginning infectious disease outbreak (2020). Because these obstruct defining facial features, now even more challenging identify, increasing necessity for advanced detection technology. research also real/fake video dataset because field lacks required detection-model training. proposed proposes Face Mask Dataset (DFFMD) based on novel Inception-ResNet-v2 preprocessing stages, feature-based, residual connection, batch normalization. combination normalization increases deepfake in presence facemasks, unlike traditional methods. study’s results compared existing state-of-the-art methods detect face-mask-Deepfakes 99.81% InceptionResNetV2 VGG19, whose 77.48%, 99.25%, respectively. Future work should evaluate developing subsequent experimental increased facemasks.

Language: Английский

---

Deep Fake Face Detection Using LSTM

IARJSET, Journal Year: 2024, Volume and Issue: 11(3)

Published: March 30, 2024

Deep fake videos, which employ artificial intelligence to manipulate and generate highly convincing content, have emerged as a significant threat society, potentially undermining trust in visual media.Detecting these deceptive videos is outmost importance combat the spread of misinformation protect integrity digital media.In this study, we propose novel approach for deep face video detection utilizing Long Short-Term Memory (LSTM) networks, type Recurrent Neural Network (RNN).Our capitalizes on temporal patterns context within sequences, harnessing unique strengths LSTM capturing sequential information.We demonstrate effectiveness our methodology by training network diverse dataset comprising both real videos.The network's ability learn dependencies identify inconsistencies facial expressions, eye movements, other subtle cues allows it distinguish between genuine manipulated content.To further enhance accuracy robustness system, integrate pre-processing techniques framelevel analysis, such optical flow computation landmarks extraction.Additionally, comprehensive ensemble models machine learning algorithms improve overall performance.In experiments, evaluate LSTM-based system large-scale known unseen achieving high low false positive rates.We also compare with existing methods, demonstrating its superiority terms generalization.The results study signify potential mitigating adverse effects content society.As technology continues evolve, showcases promising step towards combating dissemination multimedia, promoting media integrity, upholding information.

Language: Английский

---

Enhancing Facemask Detection using Deep learning Models

International Journal of Advanced Computer Science and Applications, Journal Year: 2023, Volume and Issue: 14(7)

Published: Jan. 1, 2023

Face detection and mask are critical tasks in the context of public safety compliance with mask-wearing protocols. Hence, it is important to track down whoever violated rules regulations. Therefore, this paper aims implement four deep learning models for face detection: MobileNet, ResNet50, Inceptionv3, VGG19. The evaluated based on precision recall metrics both tasks. results indicate that proposed model ResNet50 achieves superior performance detection, demonstrating high (99.4%) (98.6%) values. Additionally, shows commendable accuracy detection. MobileNet Inceptionv3 provide satisfactory results, while VGG19 excels but slightly lower findings contribute development effective systems, implications safety.

Language: Английский

---

Deep Fake Face Detection Using Long Short-Term Memory with Deep Learning Approach

Journal of Image Processing and Intelligent Remote Sensing, Journal Year: 2022, Volume and Issue: 21, P. 28 - 36

Published: Jan. 30, 2022

Strong and effective detection techniques are desperately needed to lessen the possible effects of disinformation manipulation as frequency deepfake videos keeps rising. The use Long Short-Term Memory (LSTM) networks for video is examined in this abstract. Recurrent neural (RNNs), such LSTM, a viable option analysing dynamic movies because their ability capture temporal dependencies sequential data. study explores complexities using LSTM architectures identify films highlights need comprehending patterns present manipulated information. Preprocessing data part suggested methodology entails producing training datasets highest Caliber augmentation methods improve model generalization. To attain best results detection, procedure network-specific optimization investigated. Evaluation criteria including recall, accuracy, precision, F1 score used evaluate how well works discern between modified authentic content. abstract also covers potential directions future strengthen resilience LSTM-based systems, difficulties constraints specific minimizing false positives negatives. have significance practical uses, especially when it comes social media hosting services, where incorporation identification can enhance online safety security.

Language: Английский

---