Search Results (18)

Responding to the lack of academic research on how young people are impacted by deepfake sexual abuse or how schools should address these issues, this paper explores levels of awareness of AI technology and sexualized deepfakes in UK schools and how schools are responding to these newly emergent harms. Drawing on interviews with students and teachers from eight schools across the UK, we found that teachers and students express uncertainty about how AI deepfake technology works. Some teachers underestimated how easy the technology is to use, and they lacked uniform comprehension that sexualized deepfakes should be treated the same way as non-consensual nudes, leading to inconsistency and variations in school responses. Students similarly lacked basic literacy about AI, equating AI with LLMs like ChatGPT, and even though sexualized deepfakes were occurring in their school contexts, students reported having received no explicit education on the topic. Educators and students connected sexualized deepfakes to a rise in misogyny via social media influencers, with some of the students and teachers calling for more education on AI, sexual violence, and consent at earlier ages. We advance the concept of AI-generated image-based sexual abuse, arguing that these harms should be understood as elements of technology-facilitated gender-based violence (TFGBV). We argue this framing is necessary to support systematic understandings of this issue and develop appropriate school responses. Our discussion offers recommendations for improving AI literacy, including preventative AI education that engages critically with AI harms and supports victims. Full article

Deepfake technology can produce highly realistic manipulated media which pose as significant cybersecurity threats, including fraud, misinformation, and privacy violations. This research proposes a deepfake prevention approach based on symmetric and asymmetric ciphers. Post-quantum asymmetric ciphers were utilized to perform digital signature operations, which offer essential security services, including integrity, authentication, and non-repudiation. Symmetric ciphers were also employed to provide confidentiality and authentication. Unlike classical ciphers that are vulnerable to quantum attacks, this study adopts quantum-resilient ciphers to offer long-term security. The proposed approach enables entities to digitally sign media content before public release on other platforms. End users can subsequently verify the authenticity of content using the public keys of the media creators. To identify the most efficient ciphers to perform cryptography operations required for deepfake prevention, the study explores the implementation of quantum-resilient symmetric and asymmetric ciphers standardized by NIST, including Dilithium, Falcon, SPHINCS+, and Ascon-80pq. Additionally, this research provides comprehensive comparisons between the various classical and post-quantum ciphers in both categories: symmetric and asymmetric. Experimental results revealed that Dilithium-5 and Falcon-512 algorithms outperform other post-quantum ciphers, with a time delay of 2.50 and 251 ms, respectively, for digital signature operations. The Falcon-512 algorithm also demonstrates superior resource efficiency, making it a cost-effective choice for digital signature operations. With respect to symmetric ciphers, Ascon-80pq achieved the lowest time consumption, taking just 0.015 ms to perform encryption and decryption operations. Also, it is a significant option for constrained devices, since it consumes fewer resources compared to standard symmetric ciphers, such as AES. Through comprehensive evaluations and comparisons of various symmetric and asymmetric ciphers, this study serves as a blueprint to identify the most efficient ciphers to perform the cryptography operations necessary for deepfake prevention. Full article

Artificial Intelligence is advancing rapidly, raising critical concerns about the integrity of digital content, particularly in sensitive domains such as medical imaging. Recent AI techniques, such as Generative Adversarial Networks (GANs) and diffusion models, can generate highly realistic synthetic medical images, posing risks of misdiagnosis, inappropriate treatment, and other adverse outcomes. This paper presents a deep learning-based approach to distinguish between authentic and synthetic images of skin malignancies generated by DCGAN, Wasserstein GAN (WGAN), and Stable Diffusion. A comprehensive dataset was constructed using authentic malignant skin images from an open-source Kaggle repository, alongside artificially generated images. Multiple deep learning models were trained and evaluated, with DenseNet169 achieving the highest performance, reaching 99.67% training accuracy, 97.50% validation accuracy, and 98.50% test accuracy—along with substantial precision, recall, and F1 scores across all classes. These results demonstrate the model’s efficacy in identifying both real and fake medical images. This work contributes to the emerging field of medical image forensics, highlighting its potential integration into clinical and insurance workflows to prevent fraud, strengthen trust, and mitigate risks. Furthermore, it lays the groundwork for future studies involving larger datasets, additional Deepfake generation methods, and real-time clinical applications. Full article

Generative Artificial Intelligence (GenAI) has transformed Australian higher education, amplifying online harms such as misinformation, fraud, and image-based abuse, with significant implications for cybercrime prevention. Combining a PRISMA-guided systematic review with MAXQDA-driven analysis of Australian university policies, this research evaluates institutional strategies against national frameworks, such as the Cybersecurity Strategy 2023–2030. Analyzing data from academic literature, we identify three key themes: educational strategies, alignment with national frameworks, and policy gaps and development. As the first qualitative analysis of 40 Australian university policies, this study uncovers systemic fragmentation in governance frameworks, with only 12 institutions addressing data privacy risks and none directly targeting AI-driven disinformation threats like deepfake harassment—a critical gap in global AI governance literature. This study provides actionable recommendations to develop the National GenAI Governance Framework, co-developed by TEQSA/UA and DoE, enhanced cyberbullying policies, and behavior-focused training to enhance digital safety and prevent cybercrime in Australian higher education. Mandatory annual CyberAI Literacy Module for all students and staff to ensure awareness of cybersecurity risks, responsible use of artificial intelligence, and digital safety practices within the university community. Full article

Deepfake technology, driven by advances in artificial intelligence (AI) and deep learning (DL), has become one of the foremost threats to digital trust and the authenticity of information. Despite the rapid development of deepfake detection methods, the dynamic evolution of generative models continues to outpace current mitigation efforts. This highlights the pressing need for more effective and proactive deepfake prevention strategy. This study introduces a comprehensive and multifaceted deepfake prevention framework that leverages both technical and non-technical countermeasures and involves collaboration among key stakeholders in a unified structure. The proposed framework has four modules: trusted content assurance, detection and monitoring, awareness and human-in-the-loop verification, and policy, governance, and regulation. The framework uses a combination of hybrid watermarking and embedding techniques, as well as cryptographic digital signature algorithms (DSAs) and blockchain technologies, to make sure that the media is authentic, traceable, and cannot be denied. Comparative experiments were conducted in this research using both classical and post-quantum DSAs to evaluate their efficiency, resource consumption, and gas costs in blockchain operations. The results revealed that the Falcon-512 algorithm outperformed other post-quantum algorithms while consuming fewer resources and lowering gas costs, making it a preferable option for real-time, quantum-resilient deepfake prevention. The framework also employed AI-based detection models and human oversight to enhance detection accuracy and robustness. Overall, this research offers a novel, multifaceted, and governance-aware strategy for deepfake prevention. The proposed approach significantly contributes to mitigating deepfake threats and offers a practical foundation for secure and transparent digital media ecosystems. Full article

A decentralized artificial intelligence (DAI) system is a human-oriented artificial intelligence (AI) system, which performs self-learning and shares its knowledge with other DAI systems like humans. A DAI device is an individual device (e.g., a mobile phone, a personal computer, a robot, a car, etc.) running a DAI system. A DAI device acquires validated knowledge data and raw data from a blockchain system as a trust anchor and improves its knowledge level by self-learning using the validated data. A DAI device using the proposed system reduces unreliable tasks, including the generation of unreliable products (e.g., deepfakes, fake news, and hallucinations), but the proposed system also prevents these malicious DAI devices from acquiring the validated data. This paper proposes a new architecture for a blockchain-based data management system for DAI devices, together with the service scenario and data flow, security threats, and security requirements. It also describes the key features and expected effects of the proposed system. This paper discusses the considerations for developing or operating the proposed system and concludes with future works. Full article

Artificial intelligence (AI) is a field of strategic interest for both the European Union (EU) and its member states, which are making significant efforts to develop and implement AI in a way that is economically and socially beneficial, as well as ethical and secure. This paper analyzes the importance of AI and its impact on the economy and society, highlighting the strategic and regulatory aspects agreed upon at the EU and Romanian levels, given this state’s status as an EU member. Based on the latest specialized literature, the first part addresses the concept of AI and emphasizes its role as a key driver of innovation and economic growth. Subsequently, we examine the EU’s institutional concerns, outlining the key guidelines and steps in harnessing AI opportunities, as well as the strategic and regulatory milestones that govern AI implementation within the EU. In this context, we focus on the complexities involved in the transition to the AI Era, recent developments, the process of drafting and adopting the EU AI Act, and the significance of the AI Pact. Our study fully reflects that Romania is also taking significant strategic and regulatory measures to align with the demands of the AI Era, with particular attention given to improving the legislative framework regarding the ethical implications of AI implementation and preventing deepfakes. Full article

Over the past few years, the rapid development of deepfake technology based on generative models has posed a significant threat to the field of information security. Despite the notable progress in deepfake-detection methods based on the spatial domain, the detection capability of the models drops sharply when dealing with low-quality images. Moreover, the effectiveness of detection relies on the realism of the forged images and the specific traces inherent to particular forgery techniques, which often weakens the models’ generalization ability. To address this issue, we propose the Frequency-Domain Masking and Spatial Interaction (FMSI) model. The FMSI model innovatively introduces masked image modeling in frequency-domain processing. This prevents the model from focusing too much on specific frequency-domain features and enhances its generalization ability. We design a high-frequency information convolution module for spatial and channel dimensions to help the model capture subtle forgery traces more effectively. Also, we creatively design a dual stream architecture for frequency-domain and spatial-domain information interaction and overcome single-domain detection limitations. Our model is tested on three public benchmark datasets (FaceForensics++, Celeb-DF, and WildDeepfake) through intra-domain and cross-domain experiments. The detection and generalization capabilities of the model are evaluated using the AUC and EER metrics. The experimental results demonstrate that our model not only possesses high detection capability but also exhibits excellent generalization ability. Full article

Generative adversarial networks (GANs) have revolutionised various fields by creating highly realistic images, videos, and audio, thus enhancing applications such as video game development and data augmentation. However, this technology has also given rise to deepfakes, which pose serious challenges due to their potential to create deceptive content. Thousands of media reports have informed us of such occurrences, highlighting the urgent need for reliable detection methods. This study addresses the issue by developing a deep learning (DL) model capable of distinguishing between real and fake face images generated by StyleGAN. Using a subset of the 140K real and fake face dataset, we explored five different models: a custom CNN, ResNet50, DenseNet121, MobileNet, and InceptionV3. We leveraged the pre-trained models to utilise their robust feature extraction and computational efficiency, which are essential for distinguishing between real and fake features. Through extensive experimentation with various dataset sizes, preprocessing techniques, and split ratios, we identified the optimal ones. The 20k_gan_8_1_1 dataset produced the best results, with MobileNet achieving a test accuracy of 98.5%, followed by InceptionV3 at 98.0%, DenseNet121 at 97.3%, ResNet50 at 96.1%, and the custom CNN at 86.2%. All of these models were trained on only 16,000 images and validated and tested on 2000 images each. The custom CNN model was built with a simpler architecture of two convolutional layers and, hence, lagged in accuracy due to its limited feature extraction capabilities compared with deeper networks. This research work also included the development of a user-friendly web interface that allows deepfake detection by uploading images. The web interface backend was developed using Flask, enabling real-time deepfake detection, allowing users to upload images for analysis and demonstrating a practical use for platforms in need of quick, user-friendly verification. This application demonstrates significant potential for practical applications, such as on social media platforms, where the model can help prevent the spread of fake content by flagging suspicious images for review. This study makes important contributions by comparing different deep learning models, including a custom CNN, to understand the balance between model complexity and accuracy in deepfake detection. It also identifies the best dataset setup that improves detection while keeping computational costs low. Additionally, it introduces a user-friendly web tool that allows real-time deepfake detection, making the research useful for social media moderation, security, and content verification. Nevertheless, identifying specific features of GAN-generated deepfakes remains challenging due to their high realism. Future works will aim to expand the dataset by using all 140,000 images, refine the custom CNN model to increase its accuracy, and incorporate more advanced techniques, such as Vision Transformers and diffusion models. The outcomes of this study contribute to the ongoing efforts to counteract the negative impacts of GAN-generated images. Full article

In the era of deepfakes and AI-generated content, digital image manipulation poses significant challenges to image authenticity, creating doubts about the credibility of images. Traditional image forensics techniques often struggle to detect sophisticated tampering, and passive detection approaches are reactive, verifying authenticity only after counterfeiting occurs. In this paper, we propose a novel full-resolution secure learned image codec (SLIC) designed to proactively prevent image manipulation by creating self-destructive artifacts upon re-compression. Once a sensitive image is encoded using SLIC, any subsequent re-compression or editing attempts will result in visually severe distortions, making the image’s tampering immediately evident. Because the content of an SLIC image is either original or visually damaged after tampering, images encoded with this secure codec hold greater credibility. SLIC leverages adversarial training to fine-tune a learned image codec that introduces out-of-distribution perturbations, ensuring that the first compressed image retains high quality while subsequent re-compressions degrade drastically. We analyze and compare the adversarial effects of various perceptual quality metrics combined with different learned codecs. Our experiments demonstrate that SLIC holds significant promise as a proactive defense strategy against image manipulation, offering a new approach to enhancing image credibility and authenticity in a media landscape increasingly dominated by AI-driven forgeries. Full article

The rapid growth of deepfake technology, generating realistic manipulated media, poses a significant threat due to potential misuse. Therefore, effective detection methods are urgently needed to prevent malicious use, as current approaches often focus on single modalities or the simple fusion of audio–visual signals, limiting their accuracy. To solve this problem, we propose a deepfake detection scheme based on bimodal temporal feature prediction, which innovatively introduces the idea of temporal feature prediction into the audio–video bimodal deepfake detection task, aiming at fully exploiting the temporal laws of audio–visual modalities. First, pairs of adjacent audio–video sequence clips are used to construct input quadruples, and a dual-stream network is employed to extract temporal feature representations from video and audio, respectively. A video prediction module and an audio prediction module are designed to capture the temporal inconsistencies within each single modality by predicting future temporal features and comparing them with reference features. Then, a projection layer network is designed to align the audio–visual features, using contrastive loss functions to perform contrastive learning and maximize the differences between real and fake video modalities. Experiments on the FakeAVCeleb dataset demonstrate superior performance with an accuracy of 84.33% and an AUC of 89.91%, outperforming existing methods and confirming the effectiveness of our approach in deepfake detection. Full article

Deepfakes have become widespread and have continued to develop rapidly in recent years. In addition to the use of deepfakes in movies and for humorous purposes, this technology has also begun to pose a threat to many companies and politicians. Deepfake detection is critical to the prevention of this threat. In this study, a Choquet fuzzy integral-based deepfake detection method is proposed to increase overall performance by combining the results obtained from different deepfake detection methods. Three different deepfake detection models were used in the study: XceptionNet, which has better performance in detecting real images/videos; EfficientNet, which has better performance in detecting fake videos; and a model based on their hybrid uses. The proposed method based on the Choquet fuzzy integral aims to eliminate the shortcomings of these methods by using each of the other methods. As a result, a higher performance was achieved with the proposed method than found when all three methods were used individually. As a result of the testing and validation studies carried out on FaceForensics++, DFDC, Celeb-DF, and DeepFake-TIMIT datasets, the individual performance levels of the algorithms used were 81.34%, 82.78%, and 79.15% on average, according to the AUC curve, while the level of 97.79% was reached with the proposed method. Considering that the average performance of the three methods across all datasets is 81.09%, it can be seen that an improvement of approximately 16.7% is achieved. In the FaceForensics++ dataset, in which individual algorithms are more successful, the performance of the proposed method reaches the highest AUC value, 99.8%. It can be seen that the performance rates can be increased by changing the individual methods discussed in the proposed method. We believe that the proposed method will inspire researchers and will be further developed. Full article

The detection of deepfake images and videos is a critical concern in social communication due to the widespread utilization of deepfake techniques. The prevalence of these methods poses risks to trust and authenticity across various domains, emphasizing the importance of identifying fake faces for security and preventing socio-political issues. In the digital media era, deep learning outperforms traditional image processing methods in deepfake detection, underscoring its significance. This research introduces an innovative approach for detecting deepfake images by employing transfer learning in a hybrid architecture that combines convolutional neural networks (CNNs) and long short-term memory (LSTM). The hybrid CNN-LSTM model exhibits promise in combating deep fakes by merging the spatial awareness of CNNs with the temporal context understanding of LSTMs. Demonstrating effective performance on open-source datasets like “DFDC” and “Ciplab”, the proposed method achieves an impressive precision of 98.21%, indicating its capability to accurately identify deepfake images with a limited false-positive rate. The model’s error rate is 0.26%, emphasizing the challenges and intricacies inherent in deepfake detection tasks. These findings underscore the potential of hybrid deep learning techniques for addressing the urgent issue of deepfake image detection. Full article

Deepfakes are digital audio, video, or images manipulated using machine learning algorithms. These manipulated media files can convincingly depict individuals doing or saying things they never actually did. Deepfakes pose significant risks to our lives, including national security, financial markets, and personal privacy. The ability to create convincing deep fakes can also harm individuals’ reputations and can be used to spread disinformation and fake news. As such, there is a growing need for reliable and accurate methods to detect deep fakes and prevent their harmful effects. In this paper, a hybrid convolutional neural network (CNN) and recurrent neural network (RNN) with a particle swarm optimization (PSO) algorithm is utilized to demonstrate a deep learning strategy for detecting deepfake videos. High accuracy, sensitivity, specificity, and F1 score were attained by the proposed approach when tested on two publicly available datasets: Celeb-DF and the Deepfake Detection Challenge Dataset (DFDC). Specifically, the proposed method achieved an average accuracy of 97.26% on Celeb-DF and an average accuracy of 94.2% on DFDC. The results were compared to other state-of-the-art methods and showed that the proposed method outperformed many. The proposed method can effectively detect deepfake videos, which is essential for identifying and preventing the spread of manipulated content online. Full article

The development of deepfake technology, based on deep learning, has made it easier to create images of fake human faces that are indistinguishable from the real thing. Many deepfake methods and programs are publicly available and can be used maliciously, for example, by creating fake social media accounts with images of non-existent human faces. To prevent the misuse of such fake images, several deepfake detection methods have been proposed as a countermeasure and have proven capable of detecting deepfakes with high accuracy when the target deepfake model has been identified. However, the existing approaches are not robust to partial editing and/or occlusion caused by masks, glasses, or manual editing, all of which can lead to an unacceptable drop in accuracy. In this paper, we propose a novel deepfake detection approach based on a dynamic configuration of an ensemble model that consists of deepfake detectors. These deepfake detectors are based on convolutional neural networks (CNNs) and are specialized to detect deepfakes by focusing on individual parts of the face. We demonstrate that a dynamic selection of face parts and an ensemble of selected CNN models is effective at realizing highly accurate deepfake detection even from partly edited and occluded images. Full article