Top Research Papers on Deepfake Detection

A detection model using convolution neural network (CNN) for face detection and Recurrent neural network (RNN) for video classification is developed for detecting fake videos.

A wild testbed is built by gathering texts from various human writings and deepfake texts generated by different LLMs to showcase the challenges of deep-fake text detection in a wild testbed and show empirical results on automatic detection meth-ods.

This research introduces a Multimodal Deepfake Detection system capable of identifying manipulated content by combining visual and auditory cues, which employs convolutional neural networks to analyse video frames and process audio spectrograms, providing a comprehensive approach to detecting deepfake content.

A new multi-attentional deepfake detection network that consists of three key components: multiple spatial attention heads to make the network attend to different local parts, a new regional independence loss and an attention guided data augmentation strategy, and state-of-the-art performance.

Analytic thinking and political interest were positively associated with identifying deepfakes and negatively associated with the perceived accuracy of a fake news piece about a leaked video (whether or not the deepfake video itself was presented).

This paper proposes the first parameter-efficient tuning approach for deepfake detection, namely DeepFake-Adapter, to effectively and efficiently adapt the generalizable high-level semantics from large pre-trained ViTs to aid deepfake detection.

This work develops an open-source online platform, known as DeepFake-o-meter, that integrates state-of-the-art DeepFake detection methods and provide a convenient interface for the users.

EfficientNet-B5 is used to pluck out the spatial options of those faces they are fed as a batch of input series into a two-way long- and short-term memory (BiLSTM) to extract temporal characteristics.

The survey shows that future research should address the lack of large scale datasets in the wild, poor generalization of existing detection methods to unknown fake attacks, as well as interpretability of detection results.

No single method can reliably detect all Deepfakes and, therefore, combining multiple methods is often necessary to achieve high detection rates, and it is suggested that more extensive and diverse datasets are needed to improve the accuracy of detection algorithms.

This is the first time to employ automated machine learning to adaptively search a neural architecture for deepfake detection, and based on the explored search space, this method achieves competitive prediction accuracy compared to previous methods.

It has been revealed that current detection methods are still insufficient to be applied in real scenes, and further research should pay more attention to the generalization and robustness.

In a fake video, the face, expression or speech is replaced with an image of another's face, with a distinct speech or emotion with the help of the technology of deep learning.

This work makes the first attempt to improve deepfake detection fairness by proposing novel loss functions that handle both the setting where demographic information is available as well as the case where this information is absent.

This work introduces an open-source and user-friendly online platform, DeepFake-O-Meter v2.0, that integrates state-of-the-art methods for detecting DeepFake images, videos, and audio and serves as an evaluation and benchmanrking platform for researchers in digital media forensics to compare the performance of multiple algorithms on the same input.

The Face Deepfake Detection and Reconstruction Challenge was described and two different tasks were proposed to the participants: creating a Deepfake detector capable of working in an “in the wild” scenario; and creating a method capable of reconstructing original images from Deepfakes.

It is shown that deepfake detectors proven to generalize well on multiple research datasets still struggle in real-world scenarios with well-crafted fakes, and it is suggested that their training datasets should be complemented by high-quality fakes since training on mere research data is insufficient.

This paper aims to investigate deepfake challenges, and to detect deepfake videos by using eye blinking, using convolutional neural networks to classify the eye states and long short term memory for sequence learning and the eye aspect ratio was used to calculate the height and width of open and closed eyes and to detects the blinking intervals.

A novel DEEPFAKE C-L-I (Classification-Localization-Inference) in which the idea of accelerating Quantized Deepfake Detection Models using FPGAs due to their ability of maximum parallelism and energy efficiency compared to generalized GPUs is explored.

This work proposes the first method to address the fairness generalization problem in deepfake detection by simultaneously considering features, loss, and optimization aspects, and employs disentanglement learning to extract demographic and domain-agnostic forgery features, fusing them to encourage fair learning across a flattened loss landscape.

It is argued that the promise of algorithmic detectability falls short and that the ethico-politics of deepfakes cannot be reduced solely to a framework of detection algorithms.

Inspired by the success of masked modeling, Masked Relation Learning is proposed which decreases the redundancy to learn informative relational features and outperforms the state of the art by 2% AUC on the cross-dataset DeepFake video detection.

Novel synthetic training data called self-blended images (SBIs) to detect deepfakes are presented and extensive experiments show that the method improves the model generalization to unknown manipulations and scenes.

This work performs the first large scale test of PRNU-based deepfake detection methods on a variety of standard datasets and shows the impact of often neglected parameters of the face extraction stage on detection accuracy.

A gender-balanced and annotated deepfake dataset, GBDF, is contributed to mitigate the performance differential and to promote research and development towards fairness-aware deep fake detectors.

The readily available face editing apps and the vulnerability (or performance degradation) of face recognition systems under various face manipulations are outlined and expected to aid the readers in comprehending deepfake generation and detection mechanisms, together with open issues and future directions.

The 1M-Deepfakes Detection Challenge is designed to engage the research community in developing advanced methods for detecting and localizing deepfake manipulations within the large-scale high-realistic audio-visual dataset.

A simple and efficient pipeline for detecting fake environmental sounds based on the CLAP audio embedding is proposed and it is shown that using an audio embedding trained specifically on environmental audio is beneficial over a standard VGGish one as it provides a 10% increase in detection performance.

A systematic literature review (SLR) is conducted, summarizing 112 relevant articles from 2018 to 2020 that presented a variety of methodologies in Deepfake detection and concluding that the deep learning-based methods outperform other methods in Deep fake detection.

A key component of FaceGuard is a new deep-learning-based watermarking method, which is robust to normal image post-processing such as JPEG compression, Gaussian blurring, cropping, and resizing, but fragile to deepfake manipulation.

This work systematize audio spoofing detection by re-implementing and uniformly evaluating architectures from related work and identifies overarching features for successful audio deepfake detection, such as using cqtspec or logspec features instead of melspec features.

This paper uses a method known as frequency domain analysis after which a classifier will be used to differentiate the real and fake image, and can show promising performance for detecting these deepfake images.

A novel research problem called Detecting Sequential DeepFake Manipulation (Seq-DeepFake), where face images are manipulated sequentially with corresponding annotations of sequential facial manipulation vectors, which requires correctly predicting a sequential vector of facial manipulation operations.

A large-scale empirical study of facial deepfake/detection models and a set of key findings are drawn that the detection methods have poor generalization on samples generated by different deepfake methods, and there is no significant correlation between anti-detection ability and visual quality of deepfake samples.

This paper proposes an approach for detecting deepfake videos using Resnext CNN and LSTM, which can help in preventing the spread of misinformation and safeguarding the authors' society.

This work presents the first comprehensive benchmark for deepfake detection, called DeepfakeBench, which offers three key contributions: a unified data management system to ensure consistent input across all detectors, an integrated framework for state-of-the-art methods implementation, and standardized evaluation metrics and protocols to promote transparency and reproducibility.

This survey comprehensively reviews the latest developments in deepfake generation and detection, summarizing and analyzing current state-of-the-arts in this rapidly evolving field, and discusses the development of several related sub-fields.

This article explores the creation and detection of deepfakes and provides an in-depth view as to how these architectures work and the current trends and advancements in this domain.

The results lead to the hypothesis that DMs produce fewer detectable artifacts and are thus more difficult to detect compared to GANs.

A novel approach for detecting deepfake video sequences by leveraging data multimodality that extracts audio-visual features from the input video over time and analyzes them using time-aware neural networks, and indicates that a multimodal approach is more effective than a monomodal one, even if trained on disjoint monommodal datasets.

This work considers the case where the network is trained on one Deepfake algorithm, and tested on Deepfakes generated by another algorithm, so as to achieve SOTA results.

The SingFake dataset is presented, the first curated in-the-wild dataset consisting of 28.93 hours of bonafide and 29.40 hours of deepfake song clips in five languages from 40 singers, and four state-of-the-art speech countermeasure systems trained on speech utterances are evaluated.

A deepfake database (DFLIP-3K) is introduced for the development of convincing and explainable deepfake detection and is envisioned as an open database that fosters transparency and encourages collaborative efforts to further enhance its growth.

The proposed approach results in accuracy scores comparable to and surpassing several SOTA(State-of-the-art) approaches on three benchmark datasets, while consuming considerably lesser computational overhead, and containing over 100x lesser trainable parameters which was achieved using the extraction and manipulation of geometrical features.

This work proposes a novel visual / auditory deepfake joint detection task and shows that exploiting the intrinsic synchronization between the visual and auditory modalities could benefit deepfake detection.

This work proposes the Local-&Temporal-aware Transformer-based Deepfake Detection (LTTD) framework, which adopts a local-to-global learning protocol with a particular focus on the valuable temporal information within local sequences.

Insight into the effectiveness of different deep learning architectures (transformers, CNNs), training strategies (supervised, self-supervised), and deepfake detection benchmarks can help guide the development of more accurate and reliable deep fake detection systems.

The United States should pursue a two-pronged strategy against deepfakes composed of enhancing detection initiatives and implementing content provenance across the internet to prevent people from believing in the disinformation spread across the internet.

The proposed Xception- based approach holds promise in addressing this pressing challenge, providing a reliable means to distinguish deepfakes from authentic content, ultimately safeguarding the integrity of digital media and information dissemination.

This research presents a thorough comparative analysis of current state-of-the-art deepfake detection methods and identifies the factors that contribute to the performance degradation of deep fake detection models currently being used when tested against a comprehensive dataset.