Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
28
datasets available to search
ShareScore release 0.7.1
Dataset results
28 results for “deepfake”
CVoiceFake (crafted by SafeEar: Content Privacy-Preserving Audio Deepfake Detection)
<h1><strong>Introduction:</strong></h1> <p>CVoiceFake (small) is a dataset that features a random selection of 10% of samples from the entire collection. This dataset encompasses <strong>five common languages (English, Chinese, German, French, and Italian)</strong> and utilizes <strong>multi-advanced and classical voice cloning techniques</strong> (Parallel WaveGAN, Multi-band MelGAN, Style MelGAN, Griffin-Lim, WORLD, and DiffWave) to produce audio samples that bear a high resemblance to authentic audio.</p> <ol> <li><strong>Parallel WaveGAN</strong>: As a non-autoregressive vocoder-based model, Parallel WaveGAN produces high-fidelity audio rapidly, ideal for efficient and quality deepfake generation.</li> <li><strong>Multi-band MelGAN</strong>: Multi-band MelGAN is a variant of MelGAN that divides the frequency spectrum into sub-bands for faster and more stable multi-lingual vocoder training, enhancing the robustness and scalability of the dataset.</li> <li><strong>Style MelGAN</strong>: Style MelGAN is designed to capture fine prosodic and stylistic nuances of speech, making it particularly compelling for deepfake applications that require high levels of expressivity and variation in speech synthesis.</li> <li><strong>Griffin-Lim</strong>: This algorithm reconstructs waveforms from spectrograms using an iterative phase estimation method. Though less high-fidelity than neural vocoders, it serves as a traditional baseline for comparing deepfake generation.</li> <li><strong>WORLD</strong>: WORLD is a statistical parameter-based voice synthesis system that offers fine control over the spectral and prosodic features of the synthesized audio. Its fine manipulation is useful for crafting the nuanced variations needed in deepfake datasets.</li> <li>We have also built the SOTA diffusion-based deepfake audio (DiffWave); please contact the author at <code>xinfengli@zju.edu.cn</code> if you are interested in the dataset, particularly the DiffWave portion. Furthermore, any additional discussions are welcomed.<br><strong>DiffWave</strong>: DiffWave is a diffusion probability model for waveform generation. It converts the white noise signal into structured waveform through a Markov chain, capable of both conditional and unconditional generation tasks. DiffWave represents the advanced synthesis method for its fast synthesis speed and high synthesis quality.</li> </ol> <h1><strong>🔥</strong><strong>News:</strong></h1> <p>Please note that we recently released our DiffWave subset in Version 2 in comparison to Version 1, which is available on <a href="../records/14062964" target="_blank" rel="noopener">CVoiceFake Full</a>. You can download the file named CVoiceFake_Large_diffwave_update.tar.gz.xx, and after unzipping it, you will find it retains the same file structure as before.<br> <strong>| CVoiceFake_Large_diffwave_update.tar.gz.00 |<br> | CVoiceFake_Large_diffwave_update.tar.gz.01 |</strong></p> <p> </p> <h1><strong>Full Dataset & Project Page:</strong></h1> <p>The whole dataset is available on <a href="../records/14062964" target="_blank" rel="noopener">CVoiceFake Full</a> as well. Please kindly also refer to the project page: <a title="SafeEar Website" href="https://safeearweb.github.io/Project/" target="_blank" rel="noopener">SafeEar Website</a>.</p> <p> </p> <h1><strong>Citation:</strong></h1> <p>If you find our paper/code/benchmark helpful, please kindly consider citing this work with the following reference:</p> <pre><code>@inproceedings{li2024safeear,<br> author = {Li, Xinfeng and Li, Kai and Zheng, Yifan and Yan, Chen and Ji, Xiaoyu, and Xu, Wenyuan},<br> title = {{SafeEar: Content Privacy-Preserving Audio Deepfake Detection}},<br> booktitle = {Proceedings of the 2024 {ACM} {SIGSAC} Conference on Computer and Communications Security (CCS)}<br> year = {2024},<br>} </code></pre> <div> <div> </div> </div>
WaveFake: A data set to facilitate audio DeepFake detection
<p>The main purpose of this data set is to facilitate research into audio DeepFakes. We hope that this work helps in finding new detection methods to prevent such attempts. These generated media files have been increasingly used to commit <a href="https://www.vice.com/en/article/pkyqvb/deepfake-audio-impersonating-ceo-fraud-attempt">impersonation attempts</a> or <a href="https://www.wired.com/story/telegram-still-hasnt-removed-an-ai-bot-thats-abusing-women/">online harassment</a>. You can find the accompanying code repository on <a href="https://github.com/RUB-SysSec/WaveFake">GitHub</a>.</p> <p>The data set consists of 104,885 generated audio clips (16-bit PCM wav). We examine multiple networks trained on two reference data sets. First, the <a href="https://keithito.com/LJ-Speech-Dataset/">LJSpeech</a> data set consisting of 13,100 short audio clips (on average 6 seconds each; roughly 24 hours total) read by a female speaker. It features passages from 7 non-fiction books and the audio was recorded on a MacBook Pro microphone. Second, we include samples based on the <a href="https://sites.google.com/site/shinnosuketakamichi/publication/jsut">JSUT</a> data set, specifically, basic5000 corpus. This corpus consists of 5,000 sentences covering all basic kanji of the Japanese language (4.8 seconds on average; roughly 6.7 hours total). The recordings were performed by a female native Japanese speaker in an anechoic room. Finally, we include samples from a full text-to-speech pipeline (16,283 phrases; 3.8s on average; roughly 17.5 hours total). Thus, our data set consists of approximately 175 hours of generated audio files in total. Note that we do not redistribute the reference data.</p> <p>We included a range of architectures in our data set:</p> <ul> <li><a href="https://arxiv.org/abs/1910.06711">MelGAN</a></li> <li><a href="https://arxiv.org/abs/1910.11480">Parallel WaveGAN</a></li> <li><a href="https://arxiv.org/abs/2005.05106">Multi-Band MelGAN</a></li> <li><a href="http://arxiv.org/abs/2005.05106">Full-Band MelGAN</a></li> <li><a href="https://arxiv.org/abs/2010.05646">HiFi-GAN</a></li> <li><a href="https://arxiv.org/abs/1811.00002">WaveGlow</a></li> </ul> <p>Additionally, we examined a bigger version of MelGAN and include samples from a full TTS-pipeline consisting of a conformer and parallel WaveGAN model.</p> <p><strong>Collection Process</strong></p> <p>For WaveGlow, we utilize the <a href="https://github.com/NVIDIA/waveglow">official implementation</a> (commit 8afb643) in conjunction with the official pre-trained network on <a href="https://pytorch.org/hub/nvidia_deeplearningexamples_waveglow/">PyTorch Hub</a>. We use a popular implementation available on <a href="https://github.com/kan-bayashi/ParallelWaveGAN">GitHub</a> (commit 12c677e) for the remaining networks. The repository also offers pre-trained models. We used the pre-trained networks to generate samples that are similar to their respective training distributions, <a href="https://keithito.com/LJ-Speech-Dataset/">LJ Speech</a> and <a href="https://sites.google.com/site/shinnosuketakamichi/publication/jsut">JSUT</a>. When sampling the data set, we first extract Mel spectrograms from the original audio files, using the pre-processing scripts of the corresponding repositories. We then feed these Mel spectrograms to the respective models to obtain the data set. For sampling the full TTS results, we use the <a href="https://github.com/espnet/espnet">ESPnet</a> project. To make sure the generated phrases do not overlap with the training set, we downloaded the <a href="https://commonvoice.mozilla.org/en/datasets">common voices data set</a> and extracted 16.285 phrases from it.</p> <p>This data set is licensed with a CC-BY-SA 4.0 license.</p> <p>This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy -- EXC-2092 CaSa -- 390781972.</p>
Analysis Script of Prediction of Self-efficacy in Recognizing Deepfake based on HEXACO Personality Traits
<p><em><strong>Analysis Script (JASP) </strong></em>of Prediction of Self-efficacy in Recognizing Deepfake based on HEXACO Personality Traits </p>
SVDD Challenge 2024: A Singing Voice Deepfake Detection Challenge (WildSVDD Track)
<p>For more information about SVDD Challenge 2024, please refer to https://challenge.singfake.org/.<br><br>WildSVDD track dataset is an extension of <a href="https://singfake.org/">SingFake</a> dataset. </p>
System Fingerprint Recognition for Deepfake Audio (SFR) - Compressed Set
<div>The rapid progress of deep speech synthesis models has posed significant threats to society such as malicious manip</div> <div>ulation of content. This has led to an increase in studies aimed at detecting so-called “deepfake audio”. However, existing works</div> <div>focus on the binary detection of real audio and fake audio. In real-world scenarios such as model copyright protection and</div> <div>digital evidence forensics, it is needed to know what tool or model generated the deepfake audio to explain the decision. This</div> <div>motivates us to ask: ‘Can we recognize the system fingerprints of deepfake audio?’ In this paper, we present the first deepfake</div> <div>audio dataset for System Fingerprint Recognition (SFR) and conduct an initial investigation. We collected the dataset from</div> <div>the speech synthesis systems of seven Chinese vendors that use the latest state-of-the-art deep learning technologies, including</div> <div>both clean and compressed sets. In addition, we provide extensive benchmarks and research findings to facilitate the further development of system fingerprint recognition methods. The dataset is publicly available. </div> <div> </div> <div>The subsets 01, 02, and 03 represent the training set, development set, and test set, respectively.</div> <div> </div> <div> <div>This data set is licensed with a CC BY-NC-ND 4.0 license.</div> </div>
System Fingerprint Recognition for Deepfake Audio (SFR) - Clean Set
<div>The rapid progress of deep speech synthesis models has posed significant threats to society such as malicious manip</div> <div>ulation of content. This has led to an increase in studies aimed at detecting so-called “deepfake audio”. However, existing works</div> <div>focus on the binary detection of real audio and fake audio. In real-world scenarios such as model copyright protection and</div> <div>digital evidence forensics, it is needed to know what tool or model generated the deepfake audio to explain the decision. This</div> <div>motivates us to ask: ‘Can we recognize the system fingerprints of deepfake audio?’ In this paper, we present the first deepfake</div> <div>audio dataset for System Fingerprint Recognition (SFR) and conduct an initial investigation. We collected the dataset from</div> <div>the speech synthesis systems of seven Chinese vendors that use the latest state-of-the-art deep learning technologies, including</div> <div>both clean and compressed sets. In addition, we provide extensive benchmarks and research findings to facilitate the further development of system fingerprint recognition methods. The dataset is publicly available. </div> <div> </div> <div>The subsets 01, 02, and 03 represent the training set, development set, and test set, respectively.</div> <div> </div> <div> <div>This data set is licensed with a CC BY-NC-ND 4.0 license.</div> </div>
CVoiceFake-Full ("SafeEar: Content Privacy-Preserving Audio Deepfake Detection", ACM CCS 2024)
<h1><strong>Introduction:</strong></h1> <p>CVoiceFake (Full) encompasses <strong>five common languages (English, Chinese, German, French, and Italian)</strong> and utilizes <strong>multi-advanced and classical voice cloning techniques</strong> (Parallel WaveGAN, Multi-band MelGAN, Style MelGAN, Griffin-Lim, WORLD, and DiffWave) to produce audio samples that bear a high resemblance to authentic audio.</p> <ol> <li><strong>Parallel WaveGAN</strong>: As a non-autoregressive vocoder-based model, Parallel WaveGAN produces high-fidelity audio rapidly, ideal for efficient and quality deepfake generation.</li> <li><strong>Multi-band MelGAN</strong>: Multi-band MelGAN is a variant of MelGAN that divides the frequency spectrum into sub-bands for faster and more stable multi-lingual vocoder training, enhancing the robustness and scalability of the dataset.</li> <li><strong>Style MelGAN</strong>: Style MelGAN is designed to capture fine prosodic and stylistic nuances of speech, making it particularly compelling for deepfake applications that require high levels of expressivity and variation in speech synthesis.</li> <li><strong>Griffin-Lim</strong>: This algorithm reconstructs waveforms from spectrograms using an iterative phase estimation method. Though less high-fidelity than neural vocoders, it serves as a traditional baseline for comparing deepfake generation.</li> <li><strong>WORLD</strong>: WORLD is a statistical parameter-based voice synthesis system that offers fine control over the spectral and prosodic features of the synthesized audio. Its fine manipulation is useful for crafting the nuanced variations needed in deepfake datasets.</li> <li>We have also built the SOTA diffusion-based deepfake audio (DiffWave); please contact the author at <code>xinfengli@zju.edu.cn</code> if you are interested in the dataset, particularly the DiffWave portion. Furthermore, any additional discussions are welcomed.<br><strong>DiffWave</strong>: DiffWave is a diffusion probability model for waveform generation. It converts the white noise signal into structured waveform through a Markov chain, capable of both conditional and unconditional generation tasks. DiffWave represents the advanced synthesis method for its fast synthesis speed and high synthesis quality.</li> </ol> <h1><strong>🔥 News:</strong></h1> <p>Please note that we recently released our DiffWave subset in Version 2 in comparison to Version 1. You can download the file named CVoiceFake_Large_diffwave_update.tar.gz.xx, and after unzipping it, you will find it retains the same file structure as before.<br> <strong>| CVoiceFake_Large_diffwave_update.tar.gz.00 |<br> | CVoiceFake_Large_diffwave_update.tar.gz.01 |</strong></p> <p> </p> <h1><strong>Full Dataset & Project Page:</strong></h1> <p>The sampled small dataset is available on <a href="11124319" target="_blank" rel="noopener">CVoiceFake Small</a> as well. Please kindly also refer to the project page: <a title="SafeEar Website" href="https://safeearweb.github.io/Project/" target="_blank" rel="noopener">SafeEar Website</a>.</p> <p> </p> <h1><strong>Citation:</strong></h1> <p>If you find our paper/code/dataset helpful, please kindly consider citing this work with the following reference:</p> <pre>@inproceedings{li2024safeear,<br> author = {Li, Xinfeng and Li, Kai and Zheng, Yifan and Yan, Chen and Ji, Xiaoyu, and Xu, Wenyuan},<br> title = {{SafeEar: Content Privacy-Preserving Audio Deepfake Detection}},<br> booktitle = {Proceedings of the 2024 {ACM} {SIGSAC} Conference on Computer and Communications Security (CCS)}<br> year = {2024},<br>}</pre>
Apps para dispositivos Android que generan Deepfakes
<p><em>Apps</em> para Android que generan <em>Deepfakes</em></p>
Dataset of Prediction of Self-efficacy in Recognizing Deepfake based on Personality Traits
<p>Dataset of <em>Prediction of Self-efficacy in Recognizing Deepfake based on HEXACO Personality Traits</em> </p>
DEepfake CROss-lingual evaluation dataset (DECRO)
<p>Deepfake cross-lingual evaluation dataset (DECRO) is constructed to evaluate the influence of language differences on deepfake detection. </p> <p><strong>If you use DECRO dataset for deepfake detection, please cite the paper "<em>Transferring Audio Deepfake Detection Capability across Languages</em>" published in <em>www'23</em>.</strong></p>
Splicing Detection and Localization for Speech Deepfakes using Audio Novelty
<p>With the rapid progress of artificial intelligence over the last few years, the possibility of generating highly realistic multimedia content is within everyone's reach. This has paved the way for the creation of deepfakes, synthetic data produced using deep learning techniques that realistically represent people in behaviors not belonging to them. <br>In the audio case, speech deepfakes can be used alone or combined with splicing techniques.<br>This consists of replacing portions of authentic speech with synthetic elements, thereby altering the conveyed message and leading to significant threats.<br>In this work, we address the problem of splicing detection and localization in speech deepfakes.<br>We consider spliced audio tracks created by substituting parts of a pristine speech with synthetically generated segments. We design a system able to detect whether a manipulation takes place and localize it in time.<br>The proposed method, trained exclusively on splicing-free audio tracks, extracts embeddings from the input signal through a sliding window. Then, it employs audio novelty to measure the similarity among consecutive signal sections and uses it to detect and localize splicing points.<br>We evaluate our method on a state-of-the-art dataset as well as a novel, bias-free corpus specifically developed and released in this paper. <br>The proposed approach is benchmarked against multiple baselines for both splicing detection and localization tasks.</p>
SVDD Challenge 2024: A Singing Voice Deepfake Detection Challenge (CtrSVDD Track, Test Set)
<p>For more information about SVDD Challenge 2024, please refer to https://challenge.singfake.org/.<br><br>We have released the test set here.</p> <p>The training and development set is at https://zenodo.org/records/10467648.</p> <p>The Interspeech paper that describes the dataset details and baseline analysis is https://arxiv.org/abs/2406.02438.</p>
Dataset Survei Penggunaan Teknologi Deepfake: Analisis Kesadaran Etis di Kalangan Mahasiswa ITS
Open the record for dataset details and reuse information.
Participant Data and Code for "Human Detection of Political Speech Deepfakes across Transcripts, Audio, and Video"
<p>All code produced to analyze the participant response data and the participant response data itself are included in this repository. </p>
Towards the Detection of Diffusion Model Deepfakes (Dataset)
<p>This upload contains the dataset used in <a href="https://arxiv.org/abs/2210.14571">Towards the Detection of Diffusion Model Deepfakes</a>.</p> <p>Further details and instructions are provided in the <a href="https://github.com/jonasricker/diffusion-model-deepfake-detection">project's Github repository</a>.</p>
Resultados preexperimento "Retos de la Alfabetización Mediática e Informacional en la ecología de la Inteligencia Artificial: deepfakes y desinformación"
Open the record for dataset details and reuse information.
Resultados del cuasiexperimento "Retos de la Alfabetización Mediática e Informacional en la ecología de la Inteligencia Artificial: deepfakes y desinformación"
Open the record for dataset details and reuse information.
A-destructive-active-defense-algorithm-for-deepfake-face-images
Open the record for dataset details and reuse information.
audio-deepfake-perceptual-study-results
<p>Supporting data, including speaker and listener respondent demographics, for the Farid lab deepfake detection perceptual studies. The audio data that this study uses can be found at https://huggingface.co/datasets/faridlab/deepspeak_v1. </p>
VIOLA JONES ALGORITHM WITH CAPSULE GRAPH NETWORK FOR DEEPFAKE DETECTION
<p><strong>dataset of this paper VIOLA JONES ALGORITHM WITH CAPSULE GRAPH NETWORK FOR DEEPFAKE DETECTION</strong></p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.