Anti-Forensics of Audio Source Identification Using Generative Adversarial Network

Li, Xiaowen; Yan, Diqun; Dong, Li; Wang, Rangding

doi:10.1109/access.2019.2960097

Cited by 13 publications

(7 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The scope of the work is limited to forgery detection created using copy‐paste and insertion operations. However, there are counter‐measures [40–43] that use anti‐forensic operations to mislead or confuse forgery detection methods through image processing operations, such as sharpening, blurring, and so on. In this case, detecting forgeries becomes more challenging, which is a topic beyond the scope of this work.…”

Section: Resultsmentioning

confidence: 99%

Forged text detection in video, scene, and document images

et al. 2020

View full text Add to dashboard Cite

Rapid advances in artificial intelligence have made it possible to produce forgeries good enough to fool an average user. As a result, there is growing interest in developing robust methods to counter such forgeries. This study presents a new Fourier spectrum‐based method for detecting forged text in video images. The authors' premise is that brightness distribution and the spectrum shape exhibit irregular patterns (inconsistencies) for forged text, while appearing more regular for original text. The method divides the spectrum of an input image into sectors and tracks to highlight these effects. Specifically, positive and negative coefficients for sectors and tracks are extracted to quantify the brightness distribution. Variations in the shape of the spectrum are analysed by determining the angular relationship between the principal axes and the sectors/tracks of the spectrum. Next, it combines these two features to detect forged text in the images of IMEI (International Mobile Equipment Identity) numbers and document. For evaluation, the following datasets are used: own video dataset and standard datasets, namely, IMEI number, ICPR 2018 Fraud Document Contest, and a natural scene text dataset. Experimental results show that the proposed method outperforms existing methods in terms of average classification rate and F‐score.

show abstract

Section: Resultsmentioning

confidence: 99%

Forged text detection in video, scene, and document images

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Mel-frequency cepstral coefficients (MFCCs) of speech recordings are used in [17] for microphone identification. Audio source identification in the scope of anti-forensics using SVM and MFFC is proposed in [21]. In [23] the smartphone is identified based on MFCC of audio recordings.…”

Section: Related Workmentioning

confidence: 99%

Fingerprinting Smartphones Based on Microphone Characteristics From Environment Affected Recordings

et al. 2022

View full text Add to dashboard Cite

Fingerprinting devices based on unique characteristics of their sensors is an important research direction nowadays due to its immediate impact on non-interactive authentications and no less due to privacy implications. In this work, we investigate smartphone fingerprints obtained from microphone data based on recordings containing human speech, environmental sounds and several live recordings performed outdoors. To comply with real-world circumstances, we also consider the presence of several types of noise that is specific to the scenarios which we address, e.g., traffic and market noise at distinct volumes, which may reduce the reliability of the data. We analyze several classification techniques based on traditional machine learning algorithms and more advanced deep learning architectures that are put to test in recognizing devices from the recordings they made. The results indicate that the classical Linear Discriminant classifier and a deep-learning Convolutional Neural Network have comparable success rates while outperforming all the rest of the algorithms.

show abstract

“…Note that, GAN has also entered into the digital forensics filed. For instance, to resist the audio source identification (ASI) forensics, Li et al [21] proposed to use GAN to falsify the source information of an audio clip by adding specific disturbance. The doctored audio clips can effectively deceive numerous ASI forensic methods.…”

Section: A Generative Adversarial Networkmentioning

confidence: 99%

Targeted Speech Adversarial Example Generation With Generative Adversarial Network

et al. 2020

Self Cite

View full text Add to dashboard Cite

Although neural network-based speech recognition models have enjoyed significant success in many acoustic systems, they are susceptible to be attacked by the adversarial examples. In this work, we make first step towards using generative adversarial network (GAN) for constructing the targeted speech adversarial examples. Specifically, we integrate the target speech recognition network with GAN framework, which can then be formulated as a three-party game. The generator in GAN aims at generating perturbation that could make the target network misclassified to a specific target, while simultaneously fooling the discriminator treating the adversarial example as a beguine one. The discriminator is to distinguish the crafted examples from the geniue samples. The classification error of the target network is back-propagated via gradient flow to the generator for updating. The target network is responsible for back-propagating the classification error via gradients to the generator for updating, but the target network itself is freezed. With the carefully designed network architecture, loss function and training strategy, we successfully train a generator that could generate the adversarial perturbation for a given speech clip and a target label. Experiential results show that the generated adversarial examples could effectively fool the state-of-the-art speech classification networks, while attaining an acceptable auditory perception quality. In addition, our proposed method runs much faster than the prevalent optimization-based schemes. To facilitate reproducible research, codes, models and data are publicly available at https://github.com/winterwindwang/SpeechAdvGan.

show abstract

Anti-Forensics of Audio Source Identification Using Generative Adversarial Network

Cited by 13 publications

References 18 publications

Forged text detection in video, scene, and document images

Forged text detection in video, scene, and document images

Fingerprinting Smartphones Based on Microphone Characteristics From Environment Affected Recordings

Targeted Speech Adversarial Example Generation With Generative Adversarial Network

Contact Info

Product

Resources

About