Article
Computer Science, Information Systems
Marriam Nawaz, Ali Javed, Aun Irtaza
Summary: The great development in AI has brought advancements in information technology. Lightweight ML techniques allow limited storage and processing power. Deepfakes, a famous application of this era, poses serious risks to global security and confidentiality. Detecting and classifying deepfakes accurately is challenging due to convincingly manipulated content produced by GANs. This study proposes a DL-based approach called C-LSTM for deepfakes detection, utilizing pre-trained models like VGG16, VGG19, ResNet50, XceptionNet, GoogleNet, and DenseNet, as well as a novel feature descriptor called Dense-Swish-Net121. Results show that the proposed Dense-Swish-Net121 with Bi-LSTM approach performs well in deepfakes detection.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Theory & Methods
Yisroel Mirsky, Wenke Lee
Summary: Generative deep learning algorithms have advanced to a stage where distinguishing between real and fake has become increasingly challenging. The unethical and malicious applications of this technology, such as the creation of deepfakes for spreading misinformation and impersonating individuals, have raised concerns. This article delves into the creation, detection, current trends, shortcomings in defense solutions, and areas requiring further research and attention in the realm of deepfakes.
ACM COMPUTING SURVEYS
(2021)
Article
Computer Science, Information Systems
Noor ul Huda, Ali Javed, Kholoud Maswadi, Ali Alhazmi, Rehan Ashraf
Summary: This paper presents a fusion method that combines deep features with handcrafted texture features for accurate detection of Deepfake videos. Experimental results show that the proposed method achieves high accuracy and area under the curve on different datasets.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Faramarz Farhangian, Rafael M. O. Cruz, George D. C. Cavalcanti
Summary: The proliferation of social networks has posed challenges in combating fake news, but automatic fake news detection using artificial intelligence has become more feasible. This paper revisits the definitions and perspectives of fake news and proposes an updated taxonomy, based on multiple criteria, for the field. The study finds that optimal feature extraction techniques vary depending on the dataset, and context-dependent models based on transformer models consistently exhibit superior performance.
INFORMATION FUSION
(2024)
Article
Computer Science, Software Engineering
Marriam Nawaz, Ali Javed, Aun Irtaza
Summary: The development of artificial intelligence and generative adversarial networks (GANs) has revolutionized technologies and approaches used for malicious purposes, particularly in the creation of deepfake videos. These highly realistic fake videos have been employed for harassment, blackmail, and political manipulation. This poses a severe threat to privacy and calls for automated methods to detect deepfakes. The presented approach utilizes deep learning, specifically the ResNet-Swish-Dense54 model, to accurately identify deepfakes by extracting and classifying human faces in video frames. Through evaluation and experimentation, the proposed method has demonstrated robustness and effectiveness in detecting visual manipulation.
Article
Mathematics, Interdisciplinary Applications
Nida Aslam, Irfan Ullah Khan, Farah Salem Alotaibi, Lama Abdulaziz Aldaej, Asma Khaled Aldubaikil
Summary: The widespread adoption and growth of social media networks have facilitated the rapid spread of fake news, leading to negative impacts. A deep learning model was proposed to classify news as real or fake, achieving significant results.
Article
Engineering, Electrical & Electronic
Meixu Chen, Todd Goodall, Anjul Patney, Alan C. Bovik
Summary: This paper proposes a new deep learning video compression architecture that efficiently compresses videos without motion estimation. The framework captures the regularities of video motion using displaced frame differences and utilizes LSTM-UNet for space-time reconstruction. Experimental results show that the proposed MOVI-Codec outperforms H.264 and HEVC codecs, and even exceeds the performance of the latest H.266 codec on high resolution videos.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2022)
Article
Engineering, Civil
Hadi Ghahremannezhad, Hang Shi, Chengjun Liu
Summary: Traffic video analytics has become a vital component of transportation systems, applying computer vision techniques to efficiently monitor video feeds from surveillance cameras. Object detection, the most crucial step, has been extensively studied with various algorithms. This paper reviews these algorithms, categorizing them as motion-based and appearance-based techniques, and analyzes their advantages, disadvantages, challenges, limitations, and potential solutions.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Rosa Gil, Jordi Virgili-Goma, Juan-Miguel Lopez-Gil, Roberto Garcia
Summary: This study conducts a bibliometric analysis of articles on the evolution and trends of deepfakes technology. It answers six research questions and identifies 331 research articles on deepfakes. The analysis provides insights on research areas, current topics, trends, researchers, and funding institutions in the field of deepfakes. This paper presents current trends and opportunities for practitioners and researchers interested in deepfakes research.
Article
Computer Science, Information Systems
Saleh Albahli, Marriam Nawaz
Summary: The development of deep learning and artificial intelligence has raised concerns about the security and privacy of digital data, particularly in relation to deepfake technology. This study proposes a DL method called the MedNet model to detect deepfake samples based on lung CT scans. The study introduces a customized EfficientNetV2-B4 framework and a spatial-channel attention mechanism to improve classification performance. The results demonstrate the effectiveness of the proposed method in accurately detecting real lung CT scan samples from deepfake images.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Dilip Kumar Sharma, Sonal Garg
Summary: Spotting fake news is a critical problem in today's society, with social media playing a significant role in its propagation. Fake news on digital platforms causes confusion and biased perspectives. This paper addresses the challenges in identifying fake news due to the lack of a comprehensive benchmark dataset, by presenting the IFND dataset that focuses on Indian news. The dataset includes both text and images, and an intelligent augmentation algorithm is used to generate meaningful fake news statements. The proposed dataset and analysis using machine learning and deep-learning classifiers, including a multi-modal approach, achieved satisfactory results.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
S. T. Suganthi, Mohamed Uvaze Ahamed Ayoobkhan, Krishna Kumar, Nebojsa Bacanin, K. Venkatachalam, Hubalovsky Stepan, Trojovsky Pavel
Summary: This article discusses the application of deep learning in detecting deep fake images, identifies the issues with existing techniques, and proposes a method that combines Fisherface and LBPH algorithms for deep fake detection in face images.
PEERJ COMPUTER SCIENCE
(2022)
Article
Computer Science, Information Systems
Jiafeng Xu, Dawei Jia, Zhizhe Lin, Teng Zhou
Summary: This study designs and develops a deep learning-based passport security feature classification model to identify similar security features. By embedding pixel attention and introducing focal loss, the proposed model achieves an enhanced classification accuracy.
Article
Computer Science, Artificial Intelligence
Anshika Choudhary, Anuja Arora
Summary: Social media plays a crucial role in influencing user decisions as a primary source of news, but faces challenges in content authenticity. This study proposed a linguistic model to address fake news detection and classification, achieving an average accuracy of 86%. The sequential neural model showed comparable performance in discernibly less time compared to other models.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Information Systems
Zhiwen Fang, Joey Tianyi Zhou, Yang Xiao, Yanan Li, Feng Yang
Summary: This study introduces a novel Multi-Encoder Single-Decoder network to tackle the pattern bias problem in anomaly detection in videos. By encoding motion and content cues individually, it achieves end-to-end learning ability and real-time running speed.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Automation & Control Systems
Zhihua Xia, Chengsheng Yuan, Rui Lv, Xingming Sun, Neal N. Xiong, Yun-Qing Shi
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2020)
Article
Computer Science, Artificial Intelligence
Chengsheng Yuan, Zhihua Xia, Xingming Sun, Q. M. Jonathan Wu
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2020)
Article
Engineering, Electrical & Electronic
Guangyong Gao, Shikun Tong, Zhihua Xia, Bin Wu, Liya Xu, Zhiqiang Zhao
Summary: Recent advancements in data hiding technology have led to the popularity of reversible data hiding (RDH) as a research topic, with a focus on medical images. This paper proposes an automatic contrast enhancement algorithm, RDHACEM, which separates images into regions of interest and non-interest and shows improved visual quality and embedding capacity in the regions of interest.
Article
Computer Science, Information Systems
Zixuan Shen, Zhihua Xia, Peipeng Yu
Summary: This paper introduces the concept of Personalized Local Differential Privacy (PLDP) and its algorithm design, with experimental results showing that the proposed scheme can protect the privacy of crowdsourced data while maintaining high utility.
SECURITY AND COMMUNICATION NETWORKS
(2021)
Review
Computer Science, Artificial Intelligence
Peipeng Yu, Zhihua Xia, Jianwei Fei, Yujiang Lu
Summary: Deepfake videos generated by deep learning algorithms have raised widespread concerns due to their potential threats to social stability. Current detection methods are not yet sufficient for real-world applications, and future research should focus more on generalization and robustness.
Article
Computer Science, Information Systems
Yujiang Lu, Yaju Liu, Jianwei Fei, Zhihua Xia
Summary: The research focuses on developing techniques for detecting forged faces in videos and introduces a novel spatiotemporal feature fusion strategy, which has been proven to be effective.
SECURITY AND COMMUNICATION NETWORKS
(2021)
Article
Computer Science, Theory & Methods
Peipeng Yu, Jianwei Fei, Zhihua Xia, Zhili Zhou, Jian Weng
Summary: This paper proposes a commonality learning strategy for face video forgery detection, which improves the generalization ability by learning the common forgery features from different forgery databases. Experimental results demonstrate the effectiveness of this strategy in face forgery detection.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2022)
Article
Computer Science, Information Systems
Zhihua Xia, Leqi Jiang, Dandan Liu, Lihua Lu, Byeungwoo Jeon
Summary: In this paper, an outsourced CBIR scheme based on a novel bag-of-encrypted-words (BOEW) model is proposed. Image encryption is performed using color value substitution, block permutation, and intra-block pixel permutation. The cloud server calculates local histograms from the encrypted image blocks, clusters them together, and uses the cluster centers as encrypted visual words. The bag-of-encrypted-words (BOEW) model is then built to represent each image, and the similarity between images is measured using the Manhattan distance between feature vectors.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2022)
Article
Engineering, Multidisciplinary
Xianyi Chen, Zhentian Zhang, Anqi Qiu, Zhihua Xia, Neal N. Xiong
Summary: This study proposes a new method of Cover Less Image Steganography (CIS) based on image selection and Star Generative Adversarial Network (StarGAN). The method aims to increase the hidden capacity, maintain better image quality, and enhance robustness and security performance.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Information Systems
Zhihua Xia, Qiuju Ji, Qi Gu, Chengsheng Yuan, Fengjun Xiao
Summary: This article proposes a secure scheme for outsourced CBIR which encrypts JPEG images and extracts secure features from the encrypted images, protecting the image data and achieving improved accuracy according to experimental results.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2022)
Article
Computer Science, Theory & Methods
Qi Gu, Zhihua Xia, Xingming Sun
Summary: This paper discusses the challenges in Multi-Source Privacy-Preserving Image Retrieval (MSPPIR) and proposes a novel JPEG image Encryption Scheme called JES-MSIR for multi-source content-based image retrieval. The proposed scheme supports secure and efficient retrieval from multiple sources, providing better retrieval services.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2022)
Review
Computer Science, Artificial Intelligence
Sulong Ge, Jianwei Fei, Zhihua Xia, Yao Tong, Jian Weng, Jianan Liu
Summary: This paper proposes a screen-shooting resilient watermarking scheme using deep neural network, which can extract watermark from captured photographs and maintain high visual quality of the watermarked images.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Information Systems
Guangyong Gao, Shikun Tong, Zhihua Xia, Yun-Qing Shi
Summary: This paper proposes a reversible data hiding algorithm based on most significant bit (MSB) prediction and error embedding, which can achieve large embedding capacity and good reconstructed image quality simultaneously.
IEEE TRANSACTIONS ON CLOUD COMPUTING
(2023)
Article
Engineering, Multidisciplinary
Zhihua Xia, Lan Wang, Jian Tang, Neal N. Xiong, Jian Weng
Summary: The paper proposes a privacy-preserving image retrieval scheme that efficiently encrypts image content to protect privacy and enhance security. Experimental results demonstrate that the proposed scheme outperforms existing schemes in terms of security and retrieval accuracy.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2021)