Deepfake Research Articles

LLM Supply Chain Provenance: A Blockchain-based Approach

The burgeoning size and complexity of Large Language Models (LLMs) introduce significant challenges in ensuring data integrity. The proliferation of "deep fakes" and manipulated information raises concerns about the vulnerability of LLMs to misinformation. Traditional LLM architectures often lack robust mechanisms for tracking the origin and history of training data. This opaqueness can leave LLMs susceptible to manipulation by malicious actors who inject biased or inaccurate data. This research proposes a novel approach integrating Blockchain Technology (BCT) within the LLM data supply chain. With its core principle of a distributed and immutable ledger, BCT offers a compelling solution to address this challenge. By storing the LLM's data supply chain on a blockchain, we establish a verifiable record of data provenance. This allows for tracing the origin of each data point used to train the LLM, fostering greater transparency and trust in the model's outputs. This decentralised approach minimises the risk of single points of failure and manipulation. Additionally, the immutability of blockchain records ensures that the data provenance remains tamper-proof, further enhancing the trustworthiness of the LLM. Our approach leverages three critical features of BCT to strengthen LLM security: 1) Transaction Anonymity: While data provenance is recorded on the blockchain, identities of data contributors can be anonymised, protecting their privacy while ensuring data integrity. 2) Decentralised Repository: Enhances the system's resilience against potential attacks by distributing the data provenance record across the blockchain network. 3) Block Validation: Rigorous consensus mechanisms ensure the validity of each data point added to the LLM's data supply chain - minimising the risk of incorporating inaccurate or manipulated data into the training process. Using the experimental approach, initial evaluations using simulated LLM training data on a blockchain platform demonstrate the feasibility and effectiveness of the proposed approach in enhancing data integrity. This approach has far-reaching implications for ensuring the trustworthiness of LLMs in various applications.

The Legal Dilemma Of Deepfakes: Ai Liability And The Challenges Of Digital Identity Theft

The European Union's Artificial Intelligence Act (EU AI Act) defines a "Deep Fake" as an “AI-generated or manipulated image, audio, or video content that mimics existing persons, objects, places, or other entities or events, creating a false appearance of authenticity or truthfulness to viewers.” In other words, Using machine learning algorithms, the bogus videos are transformed into realistic videos. Several AI-related catastrophes, including automated car accidents that resulted in physical injuries, robots causing harm to labourers, AI-assisted online privacy assaults, AI-assisted fraud that involves face, speech, or signature imitations, AI-assisted digital fingerprints falsely classifying innocent people as criminals at airports, and AI-assisted fraud in elections, have made headlines. Cryptographic signing and the hashing of a video into a fingerprint are used to confirm and reconfirm whether a video came from its original source. Many strategies for detecting deepfake videos have been developed over the last decade as a result of the rising threat of these videos. However, the fundamental issue with such procedures is that they are inaccurate and time-consuming. It also brings additional concerns such as misinformation and disinformation. The deep fakes raise concerns regarding digital identity theft. Identity theft, a criminal offence under the laws of most countries, including India, is addressed in the digital context under the Information Technology Act. However, as deepfakes are generated using artificial intelligence, the issue of fixing liability becomes increasingly complex. Commentators have argued that existing legal liability concepts may fail to address future conflicts involving AI systems. The challenge with the current legal system is primarily posed by AI systems that function without human interaction, but also by AI systems that operate with minimal human assistance. The current legal framework does not clearly define whether artificial intelligence can be regarded as a legal person, which complicates the issue of determining the extent of AI liability for the creation and dissemination of deepfakes. This ambiguity presents significant risks to the protection of digital identities, highlighting the need for a thorough review of existing laws to ensure they are adequately equipped to tackle the unique challenges posed by AI-generated content.

Concerns related to new information tool-synthetic media and Deep Fake content

Media content, form, and technology innovation are all being significantly impacted by artificial intelligence. These days, artificial intelligence is incorporated into every news media program to some extent. In particular, improvements in character-based artificial intelligence began to assist journalists in their daily work. On the other hand, artificial intelligence has given rise to a whole new kind of misleading information that is virtually identical to actual information in both form and structure. Synthetic media and Deep Fake information are terms used to describe false information and news produced with artificial intelligence. Artificial intelligence that creates deep fake content uses the three main methods of machine learning—supervised, unsupervised, and reinforcement learning—as well as the technology referred to as Generative Adversarial Networks (GANs), which produce fake images that resemble face swaps or real-world video simulations and can identify and learn from deep fake information. They have also advanced to a point where they can create convincingly fake videos on their own without assistance from humans, which will have a big impact on the entertainment, media, and film sectors. Therefore, three things need to be done to stop this kind of information from spreading: firstly, create an algorithm with countermeasures; secondly, teach journalists and editors to double-check the information to make sure it makes sense; and thirdly, public media education must be enhanced. If we do nothing about deep fake content right now, people will stop believing in news sources completely, and the infocalypse is quickly approaching. To be more precise, several studies have revealed that the people in our country lack media literacy. Consequently, if this kind of information reaches us, it will be quite challenging to control and halt it. Мэдээллийн шинэ хэрэгсэл дэх синтетик медиа буюу гүн худал (synthetic media-Deep Fake) контентийн асуудалд Хураангуй. Хиймэл оюун ухааны ололт амжилтууд хэвлэл мэдээллийн хэрэгслүүдэд цөөнгүй дэвшилийг авчирчээ. Гэвч үүнийг дагаад шинэ технологийг ашигласан синтетик медиа, гүн худал контентийн асуудал бий болж, үүнтэй тэмцэх, сэргийлэх явдал дэлхий дахины хэвлэл мэдээлэл, олон нийтийн харилцааны салбарын мэргэжилтнүүдийн хувьд нэн чухал асуудал болов. Гүн худал мэдээлэл нь L(μG,μD)≔EX~y~μD(x)[In y]+EX~y~μD(x)[In(1-y)]. математик тэгшитгэлийн дагуу бэлтгэгдэж байна. Мөн алгоритм нь хиймэл оюун ухааны машин сургалтын аргын зэрэгцээ бодит хүний видео симуляци эсвэл нүүр царайг солих байдлаар хуурамч дүрсийг боловсруулдаг, зураг бичлэгийг зохиомлоор бүтээдэг, гүн хуурамч контентын алдааг олж, түүн дээрээ суралцдаг Generative Adversarial Networks технологид суурилан хөгжих болсноор хэн нэгний оролцоогүйгээр бие даан худал видео дүрслэлийг бэлтгэх ур чадвартай болжээ. Иймээс гүн хуурамч контенттой эртхэн тэмцэж эхлэхгүй бол иргэдийн мэдээ мэдээлэлд итгэх итгэл нэг мөсөн буурах цаашлаад мэдээллийн сүйрэл (infocalypse)-ыг бий болгоход ч ойрхон байна.

AI for chemistry teaching: responsible AI and ethical considerations

Abstract This paper discusses the ethical considerations surrounding generative artificial intelligence (GenAI) in chemistry education, aiming to guide teachers toward responsible AI integration. GenAI, driven by advanced AI models like Large Language Models, has shown substantial potential in generating educational content. However, this technology’s rapid rise has brought forth ethical concerns regarding general and educational use that require careful attention from educators. The UNESCO framework on GenAI in education provides a comprehensive guide to controversies around generative AI and ethical educational considerations, emphasizing human agency, inclusion, equity, and cultural diversity. Ethical issues include digital poverty, lack of national regulatory adaptation, use of content without consent, unexplainable models used to generate outputs, AI-generated content polluting the internet, lack of understanding of the real world, reducing diversity of opinions, and further marginalizing already marginalized voices and generating deep fakes. The paper delves into these eight controversies, presenting relevant examples from chemistry education to stress the need to evaluate AI-generated content critically. The paper emphasizes the importance of relating these considerations to chemistry teachers’ content and pedagogical knowledge and argues that responsible AI usage in education must integrate these insights to prevent the propagation of biases and inaccuracies. The conclusion stresses the necessity for comprehensive teacher training to effectively and ethically employ GenAI in educational practices.

Hoaxes’ anatomy: Analysis of disinformation during the coronavirus pandemic in Europe (2020-2022)

The escalating proliferation of disinformation throughout Europe, exacerbated by the onset of the COVID-19 pandemic, necessitates a rigorous examination of its ramifications for public health. This study undertakes an investigation into trends in pandemic-related disinformation in Spain, Germany, the United Kingdom, and Poland spanning the period from 2020 to 2022. In this way, a meticulous content analysis of 704 publications from fact-checking sites was conducted. To augment the study’s robustness and provide a nuanced perspective on the evolving nature of disinformation, case studies were meticulously incorporated. The data reveals that despite of the professionalization of deep fakes, 43% of disinformation propagated through textual means, predominantly leveraging social networks (75%), especially Facebook and WhatsApp. Notably, misinformation pertaining to vaccines constituted the most prevalent narrative among disinformation stories, comprising 25.85% of the total. Moreover, as the pandemic unfolded, the thematic focus of disinformation adapted to the distinct contexts of each country. Furthermore, 47.16% of the content was found to be fabricated, with the primary objective of amplifying a common adversary accounting for 39.77% of cases. Given the contemporary relevance and public significance of this subject, the overarching objective of this research has been to scrutinise the dynamics of disinformation related to the COVID-19 pandemic in Europe. Consequently, the findings of this study carry a distinct European public service orientation.

A Multidisciplinary Threats to Emerging Cybersecurity, Legal and Ethical Threats Posed by Deepfake Technology

The creation of remarkably realistic but phony audio and video content through the use of deepfake technology raises a number of issues with cybersecurity, law and ethics. Deep fakes undermine the credibility of online information sources by posing risks including identity theft and the dissemination of false information. Legally speaking, the rise of deepfakes highlights regulatory inadequacies and enforcement concerns while posing difficult considerations about privacy rights and responsibility. Due to the potential for reputational injury and the perpetuation of negative stereotypes, deep fakes raise ethical questions about permission, privacy and harm to society. Deep fakes have become more common as a result of artificial intelligence's quick development which calls for preventative steps to counter new dangers. To effectively secure against misuse and maintain the integrity of digital material, this research paper proposes for a multidisciplinary strategy that takes legal, ethical and cybersecurity concerns into account

A Survey on Detecting Deep Fakes Using Advanced AI-Based Approaches

Today, artificial intelligence is used to clone human faces, which leads to a new technology known as “deepfakes.” Recently, machine learning (ML) approaches and the use of deep learning (DL) networks have captured researchers' competition to achieve the highest classification accuracy in building efficient models for digital content deepfake detection. Therefore, this review analyzes and compares existing deepfake detection methods based on advanced artificial intelligence algorithms. Thus, deepfake detection techniques were classified into three major categories based on the classifier model used (machine learning, deep learning, or hybrid) and then compared to show the aspects that influence the efficiency and accuracy of the algorithms. This research helps researchers develop efficient classification models for deepfake detection applications. Based on the survey information reviewed in this study, a discussion of open issues and future directions is presented. The most important challenges and research directions related to deepfake detection methods are discussed.

Detection of real-time deep fakes and face forgery in video conferencing employing generative adversarial networks

As facial modification technology advances rapidly, it poses a challenge to methods used to detect fake faces. The advent of deep learning and AI-based technologies has led to the creation of counterfeit photographs that are more difficult to discern apart from real ones. Existing Deep fake detection systems excel at spotting fake content with low visual quality and are easily recognized by visual artifacts. The study employed a unique active forensic strategy Compact Ensemble-based discriminators architecture using Deep Conditional Generative Adversarial Networks (CED-DCGAN), for identifying real-time deep fakes in video conferencing. DCGAN focuses on video-deep fake detection on features since technologies for creating convincing fakes are improving rapidly. As a first step towards recognizing DCGAN-generated images, split real-time video images into frames containing essential elements and then use that bandwidth to train an ensemble-based discriminator as a classifier. Spectra anomalies are produced by up-sampling processes, standard procedures in GAN systems for making large amounts of fake data films. The Compact Ensemble discriminator (CED) concentrates on the most distinguishing feature between the natural and synthetic images, giving the generators a robust training signal. As empirical results on publicly available datasets show, the suggested algorithms outperform state-of-the-art methods and the proposed CED-DCGAN technique successfully detects high-fidelity deep fakes in video conferencing and generalizes well when comparing with other techniques. Python tool is used for implementing this proposed study and the accuracy obtained for proposed work is 98.23 %.

Strengthening India’s Cybersecurity and Data Privacy Landscape: A Comprehensive Overview

This article chronicles India’s cybersecurity and data privacy journey, from the inception of the Information Technology Act in 2000 to the latest Digital Personal Data Protection Act of 2023. Examining key legislative developments, committee formations and judicial interventions, it highlights India’s commitment to fortifying its cybersecurity framework. The topic is analysed in a substantially analytical and qualitative manner, primarily based on literature, both primary and secondary. It is descriptive and puts forth a case explanation of cyber security in banking. In the face of escalating challenges in the digital era, the article contributes to the discourse by addressing prevalent threats like deep fakes, existing frameworks and innovative strategies for securing digital assets. Emphasising the critical interplay of technology, governance and the legal framework, the article underscores the imperative of robust cybersecurity measures to safeguard citizens’ and government interests.

A Systematic Review on Deep Fake Image Generation, Detection Techniques, Ethical Implications, and Overcoming Challenges

Deepfake technology, rooted in sophisticated machine learning techniques, utilizes deep neural networks to create highly realistic fake content such as videos, audio recordings, and images. This technology has rapidly evolved due to advancements in deep learning models, computational power, and data availability. The ethical implications, social impact, misuse, and legal frameworks surrounding Deepfake technology have been extensively studied. Detection techniques using deep learning approaches have been developed to combat the challenges posed by Deepfake content. Recommendations for future research include enhancing detection techniques, integrating explainable AI, and exploring real-time detection systems. Industry and policy implications emphasize the need for robust detection technologies, comprehensive legal frameworks, and collaborative efforts to address ethical concerns and regulate Deepfake content. This systematic review explores the landscape of deep fake image generation, detection techniques and challenges, in addition to ethical considerations. By synthesizing existing research, we aim to provide insights into deep-fake technology's advancements, limitations, and societal implications. This review underscores the urgent need for interdisciplinary collaboration and robust frameworks to address the multifaceted issues surrounding deep fakes in the digital age.

The Ethical Considerations of Artificial Intelligence (AI)

This paper examines the ethical implications of AI in 5 specific areas: Education, idea ownership, deep fakes, bias, and autonomous vehicles. The author examines these areas using the ethical traditions of deontology, consequentialism, and care ethics.

Artificial Intelligence and Its Challenges To Elections In Indonesia: A Legal Analysis

The improper utilization of AI technology poses difficulties to democracy, particularly the growing threat of unjust elections, exemplified by the deployment of bot accounts and deep fakes during electoral processes. Hence, it is crucial to build a strong and comprehensive framework to regulate the utilization of AI technology in Indonesia's political process. This article analyzes four main topics: (a) the advancement of AI technology and its connection to elections; (b) the influence of AI technology on election principles; (c) the pressing need for regulating AI in elections; and (d) the possibilities and difficulties of regulating AI technology within Indonesia's legal framework. The paper employs doctrinal legal research to examine the necessity of regulating AI technology in the context of conducting elections, taking into account the constitutional framework, established principles, and democratic election norms. The result shows that irresponsible use of AI technology remains a menace to democratic election ideals, and Indonesia must establish adequate legal mechanisms to tackle the problems stemming from the improper use of AI technology in the political process. The regulation of AI technology can be initiated by introducing a bill specifically focused on artificial intelligence (AI). This process should also involve the synchronization and harmonization of election rules, including election laws, laws governing the election of governors, regents, and mayors, laws concerning political parties, and other implementing regulations such as those established by the General Election Commission and the Election Supervisory Board.

Exploring Legal and Technical Challenges of Deep Fakes in India

The rapid advancement of deep fake technology has introduced significant challenges in the realms of data security, privacy, and intellectual property rights. Deep fakes, which involve the creation of highly realistic but manipulated digital content using artificial intelligence, have been misused for various malicious purposes, including political manipulation, identity theft, and defamation. This article examines the current state of deep fake regulation in India, highlighting the inadequacies of existing laws and the urgent need for specific legislation to address these issues. Despite the presence of broader laws under the Information Technology Act and the Indian Penal Code, there is a conspicuous absence of precise legal frameworks targeting the unique threats posed by deep fakes. Through an interdisciplinary approach that integrates legal analysis and technical insights, this study evaluates the effectiveness of current detection technologies and explores their limitations. The article also provides case studies illustrating the misuse of deep fakes in political and personal contexts, underscoring the real-world impact of this technology. In conclusion, the study advocates for comprehensive legal reforms and enhanced technological measures to mitigate the risks associated with deep fakes. By proposing targeted amendments to existing laws and recommending the development of advanced detection tools, this research aims to contribute to a more robust and effective regulatory environment for deep fakes in India.

Deep fakes and the Artificial Intelligence Act—An important signal or a missed opportunity?

AbstractThe Artificial Intelligence Act (AI Act) adopted by the European Union might serve as a global regulatory reference point. Heated negotiations over the AI Act have shown that reconciling the interests of numerous stakeholders is not an easy task. As well as creating clear and precise rules that would enable implementing effective safeguards for citizens against the manipulative potential of technology. The AI Act introduces the first legal definition of deep fakes and creates a system of protection against their harmful applications, that is based on transparency obligations. I argue that in the course of negotiations, the EU has made progress in regulating deep fakes, but the adopted solutions should only be an introduction to a more strict protection of individuals. The AI Act targets a specific group of deep fakes, leaving some of them unregulated or poorly regulated and disregarding the consequences of nonconsensual deep fake pornography. By analyzing the legal provisions related to deep fakes in the AI Act, I point to the rationale behind the chosen forms of protection, criticize the shortcomings and list the consequences of adopting the AI Act for deep fakes landscape.

Challenges faced by Nigerian fact-checkers with deep fakes during the 2023 general election

Challenges faced by Nigerian fact-checkers with deep fakes during the 2023 general election

Feature fusion Vision Transformers using MLP-Mixer for enhanced deepfake detection

Deepfake technology, utilizing deep learning and computer vision, presents significant security threats by generating highly realistic synthetic media, such as images and videos. In this paper, we propose a feature fusion-based method for deepfake detection, harnessing the capabilities of three Vision Transformer (ViT) architectures: Dual Attention Vision Transformers (DaViT), Inception Transformer (iFormer), and Group Propagation Vision Transformer (GPViT). The proposed feature fusion method integrates the local–global contextual understanding of DaViT, the frequency spectrum broadening of iFormer, and the high-resolution feature retention of GPViT to provide a comprehensive analysis of visual data, crucial for detecting subtle manipulations typical in deepfakes. An MLP-Mixer (Multi-Layer Perceptron-Mixer) model is proposed for the feature fusion process for effective feature integration, enhancing the detection capabilities while preserving the individual strengths of each ViT model. The performance of the proposed method is evaluated using standard datasets, namely FaceForensics++ and Celeb-DF, demonstrating its robustness and generalizability across various scenarios. For the four manipulation techniques in the FaceForensics++ dataset, namely DeepFakes (DF), Face2Face (F2F), FaceSwap (FS), and NeuralTextures (NT), the proposed method achieves area under the curve (AUC) of 99.75%, 97.04%, 98.78%, and 87.49% respectively, Moreover, it demonstrates an AUC of 99.84% on the Celeb-DF dataset. The performance of the proposed model has been evaluated against many state-of-the-art methods to ensure its robustness and generalization capabilities.

Exploring Cyber Fraud within the South African Cybersecurity Legal Framework

All countries are globally struggling with the challenges cybercrime presents to the cybersecurity legal framework. Fraud is not a new crime and existed long before the internet. The internet provides a threat actor access to a lot of potential victims and the use of various threat vectors to gain access to personal information by means of social engineering. It is therefore not surprising that cyber fraud has become a serious threat which continues to escalate globally. In 2021, around $100 million was lost in Canada due to online fraud. The United Kingdom (UK) Finance indicated that cyber fraud costs consumers more than £1.2 billion in 2022. The South African (SA) Fraud Prevention Services noted a 356% surge in identity fraud between April 2022 and April 2023. The cybersecurity threat landscape is ever-evolving with the UK Finance warning that the number of cyber frauds could surge out of control as threat actors begin to incorporate the use of Artificial intelligence (AI) to make their operations far more sophisticated and not as easily detected. In 2023 the United States (US) also warned that the irresponsible use of AI could exacerbate societal harms such as fraud. Cyber fraud, also referred to as a “white collar” or commercial crime, is an umbrella term to describe the commission of different types of cyber fraud by means of the use of various threat vectors. The threat vector used to commit the different type of fraud is continuously evolving, such as the use of sophisticated phishing to quishing and deep fakes which are aimed at deceiving the recipient in sharing information. The information obtained from a data breach may be used to commit cyber fraud. Irrespective of the threat vector used to commit fraud, all types of fraud present with the same elements, namely a threat actor who unlawfully and intentionally deceives a victim to benefit and cause harm. The discussion focuses on cyber fraud in general and not a specific type of cyber fraud. The purpose of the discussion is to provide an overview of the challenges cyber fraud present to the South African cybersecurity legal landscape.

Artificial Intelligence for Media: Opportunities or Threats

The emergence of artificial intelligence (AI) has ushered in an era of unprecedented possibilities and, at the same time, poses significant threats in the media landscape. As a potent tool, AI can streamline operations, offering innovative solutions like content personalization, predictive analytics, and augmented reality experiences that can enhance user engagement and boost the media industry’s economic viability. These opportunities allow for a more immersive and personalized user experience, leveraging data to create content that is finely tuned to individual preferences and trends. However, the integration of AI in media does not come without its challenges. The threat of deep fakes and the dissemination of misinformation stand as substantial concerns, with AI technologies facilitating the creation of highly convincing fake content that can manipulate perceptions and sow discord. Moreover, the autonomy given to AI systems can potentially lead to job displacements and raise ethical concerns regarding privacy and data security. Despite these threats, when wielded responsibly and ethically, AI can usher in a new golden age for media characterized by creativity, efficiency, and innovation. As the media industry stands at this crossroads, it must navigate the fine balance between leveraging AI’s opportunities and mitigating its potential threats, fostering a landscape that upholds truth, ethics, and human value at its core. It is imperative for stakeholders to collaborate, devising strategies and regulatory frameworks to ensure AI serves as a force for good, steering media towards a future that embodies progress and inclusivity.

DL MODEL FOR IMITATION FALSE FACE, AGE, GENDER AND SCENARIO DISPARITIES

Scrutinises many methods that are employed by numerous scholars to identify Deep phoney films. We offer a method for spotting identity changes, such face swapping, in a single photograph. Deep Fake and other facial swapping methods alter the face region in an effort to replicate the appearance of the context while changing only the face. We show that the two regions differ as a result of this mode of operation. These variations raise red flags for manipulative activity. Our method makes use of a pair of networks: one for face identification, that takes into account face region included within accurate semantic segmentation & specifies gender and age, and another for context recognition, which takes into account face's surrounding context. Recognizing the Meaning of Emoticons, we provide a strategy which makes use of recognition signals from our 2 networks to pinpoint these deviations, so offering an enhanced, non-traditional detection signal. We propose a deep fake-detection method based on an organ-level transformer model for extracting deep fake characteristics. By giving less weight to identification of deformed or unclear organs, we give priority to detection of obvious and intact organs. Our system can also detect subtle alterations to facial expressions and details, as well as heavily tainted, digitally-generated phony images. So-called "micro expressions" are fast, fleeting facial changes. This kind of unrestrained display of emotion is a window into a person's true feelings. examined the seven most common micro expressions seen in human faces: happiness, sadness, anger, disgust, contempt, fear, & surprise.

Deepfake Detection System

Abstract: Because deep fake technology allows for the creation of incredibly realistically modified material that has the potential to mislead viewers and possibly cause instability across a range of businesses, it poses a severe threat to modern society. These days, detecting such modified content is crucial to maintaining the trustworthiness and integrity of digital media. In response, this research proposes a robust deep learning-based technique for detecting deep fakes in videos. Our method uses the Deep fake Detection Challenge dataset, which contains both real and deep fake films, to train and assess our deep learning model. We want to provide a dependable solution that can differentiate between artificially generated information and actual content using stateof-the-art neural networks. Artificial intelligence-generated phony videos are more likely to spread thanks to deep learning algorithms, which raises serious concerns about official blackmail, terrorist propaganda, revenge pornography, and political manipulation. To alleviate these concerns, our approach is designed to automatically identify several forms of deep forgery, including replacement and re-enactment techniques. By means of extensive testing and analysis, we demonstrate the effectiveness of our deep fake detection system in precisely distinguishing bogus films from authentic ones. Our system, which combines cutting-edge deep learning methods with a large dataset, offers a promising way to lessen the hazards related to the spread of deep fake technologies