Exposing DeepFake Videos Using Attention Based Convolutional LSTM Network

Su, Yishan; Xia, Huawei; Qi, Liang; Nie, Weizhi

doi:10.1007/s11063-021-10588-6

Cited by 12 publications

(3 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…LSTM has one more input and one more output compared to RNN. The two additional values are the input and output values of the cell state of the memory and forgetting mechanism of LSTM, which is usually represented by C. The cell state can determine the storage and deletion of relevant data during specific network information processing [21,22]. The emergence of cell states not only effectively avoids the problem of gradient vanishing in RNN neural networks, but also avoids the gradient explosion problem that often occurs in models.…”

Section: Theory Of Lstm Neural Networkmentioning

confidence: 99%

Construction and Analysis of QPSO-LSTM Model in Network Security Situation Prediction

Wentao

2024

JCSANDM

View full text Add to dashboard Cite

The continuous improvement of artificial intelligence technology has deepened its application in many fields and provided more support for predicting network security situations. QPSO-LSTM model based on LSTM neural network and fused with QPSO algorithm provides more options for improving network security situation prediction, further enhancing the effectiveness of network security situation prediction, and enabling more efficient and accurate prediction and analysis of network security situations. By comparing the applications of different types of algorithms in network security situation prediction, it was found that the QPSO-LSTM model has smaller prediction errors, can achieve higher prediction accuracy, and can also obtain higher F1-score and AUC values; the shorter identifying runtime also lays the foundation for improving the speed and efficiency of network security situation prediction. Therefore, in the field of network security situation prediction, the application of QPSO-LSTM model can provide more support for further improvement and improvement of security situation prediction performance in this field.

show abstract

Section: Theory Of Lstm Neural Networkmentioning

confidence: 99%

Construction and Analysis of QPSO-LSTM Model in Network Security Situation Prediction

Wentao

2024

JCSANDM

View full text Add to dashboard Cite

show abstract

“…Convolutional Neural Networks or CNNs detect audio deepfakes by analyzing the spectral and temporal characteristics of audio data. Paper (Su et. al, 2021) mentions CNN model to extract temporal and spectral audio features from data and apply them to detect manipulated audio clips.…”

Section: Deepfake Attack Detection Techniquesmentioning

confidence: 99%

Deepfake audio detection and justification with Explainable Artificial Intelligence (XAI)

Govindu,

Kale,

Hullur

et al. 2023

Preprint

View full text Add to dashboard Cite

Deepfake audio refers to synthetically generated audio, often used as legal hoaxes to impersonate human voices. This paper generates fake audio from Fake or Real (FoR) dataset using Generative Adversarial Neural Networks (GANs). FoR dataset has the advantage of a diversity of speakers across 195,000 samples. The proposed work analyses the quality of the generated fake data using the Fréchet Audio Distance (FAD) score. FAD evaluation score of 23.814 indicates good quality fake has been produced by the generator. The study further enables glass box analysis of deepfake audio detection through Explainable Artificial Intelligence (XAI) models of LIME, SHAP and GradCAM. This research assists in understanding impact of frequency bands in audio classification based on the quantitative analysis of SHAPLey values and qualitative comparison of explainability masks of LIME and GradCAM. The use of FAD metric provides a quantitative evaluation of generator performance. XAI and FAD metrics help in the development of deepfake audio through GANs with minimal data input. The results of this research are applicable to detection of phishing audio calls and voice impersonation.

show abstract

“…This allows the model to achieve more robust performance for deployment, providing real-life business solutions. It has been shown both in theory [9,10] and practice [7,8,[11][12][13] that depth is one of the most important factors leading to the success of deep learning. However, the bottleneck lies in the training of deep neural network which is computationally heavy and timeconsuming.…”

Section: Introductionmentioning

confidence: 99%

Decoupled neural network training with re-computation and weight prediction

Peng

Lin

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

To break the three lockings during backpropagation (BP) process for neural network training, multiple decoupled learning methods have been investigated recently. These methods either lead to significant drop in accuracy performance or suffer from dramatic increase in memory usage. In this paper, a new form of decoupled learning, named decoupled neural network training scheme with re-computation and weight prediction (DTRP) is proposed. In DTRP, a re-computation scheme is adopted to solve the memory explosion problem, and a weight prediction scheme is proposed to deal with the weight delay caused by re-computation. Additionally, a batch compensation scheme is developed, allowing the proposed DTRP to run faster. Theoretical analysis shows that DTRP is guaranteed to converge to crical points under certain conditions. Experiments are conducted by training various convolutional neural networks on several classification datasets, showing comparable or better results than the state-of-the-art methods and BP. These experiments also reveal that adopting the proposed method, the memory explosion problem is effectively solved, and a significant acceleration is achieved.

show abstract

Exposing DeepFake Videos Using Attention Based Convolutional LSTM Network

Cited by 12 publications

References 39 publications

Construction and Analysis of QPSO-LSTM Model in Network Security Situation Prediction

Construction and Analysis of QPSO-LSTM Model in Network Security Situation Prediction

Deepfake audio detection and justification with Explainable Artificial Intelligence (XAI)

Decoupled neural network training with re-computation and weight prediction

Contact Info

Product

Resources

About