Search Results (263)

SecureVision is an advanced and trustworthy deepfake detection system created to tackle the growing threat of ‘deepfake’ movies that tamper with media, undermine public trust, and jeopardize cybersecurity. We present a novel approach that combines big data analytics with state-of-the-art deep learning algorithms to detect altered information in both audio and visual domains. One of SecureVision’s primary innovations is the use of multi-modal analysis, which improves detection capabilities by concurrently analyzing many media forms and strengthening resistance against advanced deepfake techniques. The system’s efficacy is further enhanced by its capacity to manage large datasets and integrate self-supervised learning, which guarantees its flexibility in the ever-changing field of digital deception. In the end, this study helps to protect digital integrity by providing a proactive, scalable, and efficient defense against the ubiquitous threat of deepfakes, thereby establishing a new benchmark for privacy and security measures in the digital era. Full article

Social learning is the main policy-design mechanism that involves interactions between agents. This study provides an overview of the research on policy design using social learning. Descriptive and co-citation analyses were used to identify emerging research lines and thematic similarities between scientific publications. The database used for the bibliometric analysis contained 271 articles published between 1979 and 2022 in 152 journals indexed by the SSCI. We propose a study based on the origins and the future research agenda of social learning for policy design. The results reveal that “environment”, “governance”, and “social” represent the knowledge base. These topics have evolved over time and have become established as a consolidated intellectual structure. In addition, a new topic called “media and news” has emerged, focusing on the challenges of spreading fake news and learning manipulation in a post-truth world. The cluster “Media and news” is gaining significance due to its impact on the dissemination of information and the shaping of opinions in contemporary society. Full article

The proliferation of fake news across multiple modalities has emerged as a critical challenge in the modern information landscape, necessitating advanced detection methods. This study proposes a comprehensive framework for fake news detection integrating text, images, and videos using machine learning and deep learning techniques. The research employs a dual-phased methodology, first analyzing textual data using various classifiers, then developing a multimodal approach combining BERT for text analysis and a modified CNN for visual data. Experiments on the ISOT fake news dataset and MediaEval 2016 image verification corpus demonstrate the effectiveness of the proposed models. For textual data, the Random Forest classifier achieved 99% accuracy, outperforming other algorithms. The multimodal approach showed superior performance compared to baseline models, with a 3.1% accuracy improvement over existing multimodal techniques. This research contributes to the ongoing efforts to combat misinformation by providing a robust, adaptable framework for detecting fake news across different media formats, addressing the complexities of modern information dissemination and manipulation. 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

(1) Background: News deserts are communities without a local news outlet, or communities where residents face significantly reduced access to the news of the local public sphere. The demise of a local news outlet can have negative effects on community engagement and on the discussion of solutions to community problems. In Portugal, for example, 25% of municipalities do not have their own media outlets. When there are no journalists reporting on reality, studies show that much of the local information in these territories is obtained through social media, such as Facebook pages and groups, which can be a source of disinformation and manipulation that communities become vulnerable to. (2) Methods: Through focus groups in the municipality of Manteigas, we researched perspectives and behaviours, as well as the factors that influence people’s choices in the consumption of information. (3) Results: We found that citizens used a wide range of informational sources, with a strong dependence on social media and institutional channels to access local information. (4) Conclusions: Proximity relationships are the basis of fact-checking processes, and citizens showed less concern about disinformation and more trust in the information they accessed through official institutions’ pages and through word-of-mouth in their communities. Full article

Wavefront shaping is a well-known method of restoring a focus deep within scattering media by manipulating the incident light. However, the achievable focus enhancement depends on and is limited by the optical and geometrical properties of the medium. These properties contribute to the number of linearly independent transmission channels for light propagating through the turbid medium. Correlations occur when the number of incident waves coupled into the scattering medium exceeds this finite number of transmission channels. This paper investigates the wavefront shaping of such correlated electric fields. The influence of the observed correlations persists even though the average electric field distribution at positions in the focal plane follows a circular complex Gaussian. We show that correlations of the transmitted electric fields reduce the achievable intensity enhancement, even deep in the turbid medium. The investigations are carried out using a Monte Carlo algorithm. It is based on the speckle statistics of independent waves and introduces correlations of neighbouring electric fields via a Cholesky decomposition of the covariance matrix. Additional investigations include scenarios where the electric fields are not completely randomized, such as for ballistic or insufficiently scattered light. Significant contributions from such little-scattered light are observed to reduce the intensity enhancement further. Data from simulations solving Maxwell’s equations are compared with the results obtained from the Monte Carlo simulations for validation throughout this paper. Full article

The emergence of deep learning has sparked notable strides in the quality of synthetic media. Yet, as photorealism reaches new heights, the line between generated and authentic images blurs, raising concerns about the dissemination of counterfeit or manipulated content online. Consequently, there is a pressing need to develop automated tools capable of effectively distinguishing synthetic images, especially those portraying faces, which is one of the most commonly encountered issues. In this work, we propose a novel approach to synthetic face discrimination, leveraging deep learning-based image compression and predominantly utilizing the quality metrics of an image to determine its authenticity. Full article

There are few studies on the representation of older people regarding aids and assistive devices and even fewer that incorporate more inclusive views (gender, emotions, anti-ageist, territorial or land approach) as well as virtual or land ethnography or artificial intelligence. The general objective was to evaluate digital images of aids and assistive aids in the older population, from the perspectives mentioned above. Method. A descriptive and cross-sectional study that searched, observed and analyzed images. An evaluation of intentionally selected images from Freepik, Pixabay, Storyblocks, Splitshire, Gratisography and ArtGPT, included in an original database constructured by several authors of this article, was carried out in the context of the ENCAGEn-CM project (2020–2023, financed by the CAM and FSE). This base was updated and expanded in October and November 2023. In addition, an image generation process was carried out using artificial intelligence, and this was also part of the analysis (ArtGPT). Finally, algorithms were used to solve and retrain with the images. Results. Of the total final images included in the expanded database until November 2023 (n = 427), only a third (28.3%, 121/427) included the aids and assistive aids label. Representations of mixed groups predominated (38.8%) and, to a lesser extent, those of women. A large proportion of the devices were ‘glasses’ (74.6%) and the ‘use of a cane’ (14.9%). To a lesser extent, ‘wheelchairs’ (4.4%) or ‘hearing aids’ (0.9%) and the presence of more than one device (simultaneously) (5.3%) were noted. The main emotions represented were ‘joy’ (45.6%) and ‘emotion not recognized’ (45.6%), with, to a lesser extent, ‘sadness’ (3.5%), ‘surprise’ (4.4%) and ‘anger’ (0.9%). Differences by sex were found in the represented emotions linked to aids and assistive aids. The representation of images of the built environment predominated significantly (70.2%), and it was observed that older women were less represented in natural environments than men. Based on the previous findings, a method is proposed to address stereotypes in images of older individuals. It involves identifying common stereotypical features, like glasses and hospital settings, using deep learning and quantum computing techniques. A convolutional neural network identifies and suppresses these elements, followed by the use of quantum algorithms to manipulate features. This systematic approach aims to mitigate biases and enhance the accuracy in representing older people in digital imagery. Conclusion. A limited proportion of images of assistive devices and older people were observed. Furthermore, among them, the lower representation of images of women in a built environment was confirmed, and the expressions of emotions were limited to only three basic ones (joy, sadness and surprise). In these evaluated digital images, the collective imagination of older people continues to be limited to a few spaces/contexts and emotions and is stereotyped regarding the same variables (sex, age, environment). Technology often overlooks innovative support tools for older adults, and AI struggles in accurately depicting emotions and environments in digital images. There is a pressing need for thorough pretraining analysis and ethical considerations to address these challenges and ensure more accurate and inclusive representations of older persons in digital media. Full article

Facial biometrics are widely used to reliably and conveniently recognize people in photos, in videos, or from real-time webcam streams. It is therefore of fundamental importance to detect synthetic faces in images in order to reduce the vulnerability of biometrics-based security systems. Furthermore, manipulated images of faces can be intentionally shared on social media to spread fake news related to the targeted individual. This paper shows how fake face recognition models may mainly rely on the information contained in the background when dealing with generated faces, thus reducing their effectiveness. Specifically, a classifier is trained to separate fake images from real ones, using their representation in a latent space. Subsequently, the faces are segmented and the background removed, and the detection procedure is performed again, observing a significant drop in classification accuracy. Finally, an explainability tool (SHAP) is used to highlight the salient areas of the image, showing that the background and face contours crucially influence the classifier decision. Full article

Social networks have become deeply integrated into our daily lives, leading to an increase in image sharing across different platforms. Simultaneously, the existence of robust and user-friendly media editors not only facilitates artistic innovation, but also raises concerns regarding the ease of creating misleading media. This highlights the need for developing new advanced techniques for the image copy detection task, which involves evaluating whether photos or videos originate from the same source. This research introduces a novel application of the Vision Transformer (ViT) model to the image copy detection task on the DISC21 dataset. Our approach involves innovative strategic sampling of the extensive DISC21 training set using K-means clustering to achieve a representative subset. Additionally, we employ complex augmentation pipelines applied while training with varying intensities. Our methodology follows the instance discrimination concept, where the Vision Transformer model is used as a classifier to map different augmentations of the same image to the same class. Next, the trained ViT model extracts descriptors of original and manipulated images that subsequently underwent post-processing to reduce dimensionality. Our best-achieving model, tested on a refined query set of 10K augmented images from the DISC21 dataset, attained a state-of-the-art micro-average precision of 0.79, demonstrating the effectiveness and innovation of our approach. Full article

Electromagnetic scattering is a routine tool for rapid, non-contact characterization of particle media. In previous work, the interaction targets of scattering intensity, scattering efficiency, and extinction efficiency of Bessel pincer light-sheet beams were all aimed at dielectric spheres. However, most particles in nature are charged. Considering the boundary condition on a charged sphere, the beam shape coefficients (BSCs) ($p_{mn},q_{mn}$) of the charged spherical particle illuminated by a Bessel pincer light-sheet beam are obtained. The extinction, scattering, and absorption efficiencies are derived under the generalized Lorenz–Mie theory (GLMT) framework. This study reveals the significant differences in scattering characteristics of Bessel pincer light-sheet beams on a charged particle compared to traditional beams. The simulations show a few apparent differences in the far-field scattering intensity and efficiencies between charged and natural spheres under the influence of dimensionless size parameters. As dimensionless parameters increase, the difference between the charged and neutral spheres decreases. The effects of refractive index and beam parameters on scattering, extinction, and absorption coefficients are different but tend to converge with increasing dimensionless parameters. When applied to charged spheres with different refractive indices, the scattering, extinction, and absorption efficiencies of Bessel pincer light-sheet beams change with variations in surface charge. However, once the surface charge reaches saturation, these efficiencies become stable. This study is significant for understanding optical manipulation and super-resolution imaging in single-molecule microbiology. Full article

The aim of this study is to investigate how amplifiers are used in Twitter (now called “$\mathbb{X}$”) during election campaigns. Specifically, the main purpose is to identify the role and engagement of Twitter amplifiers in the 2019 European elections, the visibility of political parties and leaders, and the way in which automated tools are used to manipulate public opinion by influencing voting decisions. The countries considered in the study are two economic powers of Western Europe, France and Germany, as well as two countries of the European South, which are affected by the economic and financial crisis, Greece and Italy. The countries from Southern Europe were included in the sample as they are often used by mass media as political campaign tools. This paper emphasizes the Twitter platform through which the data collection was implemented using the official API of the social networking tool, focusing on the 2019 European elections. We collected data on 88 party leaders and MEP candidates between 10 May and 30 May 2019, as well as on 44,651 accounts that retweeted them. We concluded using 237,813 election-related tweets and used network theory to analyze and visualize the data. The results demonstrate that all political parties use amplifiers to promote their tweets, and some use the same amplifiers between different countries. Full article

Solid crystalline suspensions (SCSs) containing submicron particles were introduced as a competitive solution to increase dissolution rates and the bioavailability of poorly water-soluble drugs. In an SCS, poorly water-soluble drug crystals are finely dispersed in a hydrophilic matrix. Lately, melt milling as an adapted wet milling process at elevated temperatures has been introduced as a suitable batch manufacturing process for such a formulation. In this work, the transfer from batch operation to a two-step continuous process is demonstrated to highlight the potential of this technology as an alternative to other dissolution-enhancing methods. In the first step, a powder mixture of a model drug (griseofulvin) and a carrier (xylitol) is fed to an extruder, where a uniform suspension is obtained. In the second step, the suspension is transferred to a custom-built annular gap mill, where comminution down to the submicron region takes place. The prototype’s design was based on batch grinding results and a narrow residence time distribution, intended to deliver large quantities of submicron particles in the SCS. The throughput of the mill was found to be limited by grinding media compression. By inclining the mill at an angle, the grinding media position was manipulated, such that compression was avoided. Different states of the grinding media in the grinding chamber were identified under surrogate conditions. This strategy allows the maintenance of an energy-optimized comminution without adaption of the associated process parameters, even at high throughputs. Using this new process, the production of an SCS with 80–90 % submicron particles in a single passthrough was demonstrated. Full article

Previous studies of formation of extremely compressed wave packets during femtosecond filamentation in the region of anomalous group velocity dispersion in solid dielectrics mostly considered the case of linearly polarized laser pulses. However, recent results suggest potential applications of polarization state manipulation for ultrafast laser writing of optical structures in bulk solid-state media. In the present work, evolution of radiation polarization parameters during formation of such extreme wave packets at the pump wavelength of 1900 nm in fused silica is studied numerically on the basis of the carrier-resolved unidirectional pulse propagation equation (UPPE). It was revealed that initial close-to-circular polarization leads to higher intensity of the anti-Stokes wing in the spectrum of the generated supercontinuum. Numerical simulations indicate a complex, space–time variant polarization state, and the resulting spatiotemporal electric field distribution exhibits a strong dependence on the initial polarization of the femtosecond pulse. At the same time, electric field polarization tends to linear one in the region with the highest field strength regardless of the initial parameters. The origin of this behavior is attributed to the properties of the supercontinuum components generation during filament-induced plasma formation. Full article

The rapid advancement of deepfake technology presents significant challenges in detecting highly convincing fake videos, posing risks such as misinformation, identity theft, and privacy violations. In response, this paper proposes an innovative approach to deepfake video detection by integrating features derived from ant colony optimization–particle swarm optimization (ACO-PSO) and deep learning techniques. The proposed methodology leverages ACO-PSO features and deep learning models to enhance detection accuracy and robustness. Features from ACO-PSO are extracted from the spatial and temporal characteristics of video frames, capturing subtle patterns indicative of deepfake manipulation. These features are then used to train a deep learning classifier to automatically distinguish between authentic and deepfake videos. Extensive experiments using comparative datasets demonstrate the superiority of the proposed method in terms of detection accuracy, robustness to manipulation techniques, and generalization to unseen data. The computational efficiency of the approach is also analyzed, highlighting its practical feasibility for real-time applications. The findings revealed that the proposed method achieved an accuracy of 98.91% and an F1 score of 99.12%, indicating remarkable success in deepfake detection. The integration of ACO-PSO features and deep learning enables comprehensive analysis, bolstering precision and resilience in detecting deepfake content. This approach addresses the challenges involved in facial forgery detection and contributes to safeguarding digital media integrity amid misinformation and manipulation. Full article