A Deep Learning Framework for Audio Deepfake Detection

Khochare, Janavi; Joshi, Chaitali; Yenarkar, Bakul; Suratkar, Shraddha; Kazi, Faruk

doi:10.1007/s13369-021-06297-w

Cited by 43 publications

(16 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some ML models require more complex pre-processing phase, so in these cases DL models are a better choice. Khochare et al ( 2022 ) used different Ml and DL methods on a new dataset called FOR (Reimao and Tzerpos, 2019 ). Machine learning models, such as Support Vector Machine, Random Forest, and K-Nearest Neighbors, could not achieve very high metrics, and the best of them stopped at 0.67 accuracy.…”

Section: Deepfake Categoriesmentioning

confidence: 99%

“…It is also improved to reduce EER metric as well as solve the generalization problem (Chen T. et al, 2020 ). Some also used temporal types of neural networks, namely Temporal Convolutional Networks (TCN) and achieved great results (Khochare et al, 2022 ). TCN has outperformed multi-layer perception in audio spoof detection (Tian et al, 2016 ).…”

Section: Deepfake Categoriesmentioning

confidence: 99%

“…TCN has outperformed multi-layer perception in audio spoof detection (Tian et al, 2016 ). TCN has a good ability in capturing temporal dependencies in data (Chen Y. et al, 2020 ), which could be used in audio deepfake detection (Khochare et al, 2022 ). Some used RNN-based biLSTM networks to detect the deepfakes and have good performance (Arif et al, 2021 ).…”

Section: Deepfake Categoriesmentioning

confidence: 99%

See 2 more Smart Citations

Audio deepfakes: A survey

Khanjani

Watson²,

Janeja³

2023

Front. Big Data

View full text Add to dashboard Cite

A deepfake is content or material that is synthetically generated or manipulated using artificial intelligence (AI) methods, to be passed off as real and can include audio, video, image, and text synthesis. The key difference between manual editing and deepfakes is that deepfakes are AI generated or AI manipulated and closely resemble authentic artifacts. In some cases, deepfakes can be fabricated using AI-generated content in its entirety. Deepfakes have started to have a major impact on society with more generation mechanisms emerging everyday. This article makes a contribution in understanding the landscape of deepfakes, and their detection and generation methods. We evaluate various categories of deepfakes especially in audio. The purpose of this survey is to provide readers with a deeper understanding of (1) different deepfake categories; (2) how they could be created and detected; (3) more specifically, how audio deepfakes are created and detected in more detail, which is the main focus of this paper. We found that generative adversarial networks (GANs), convolutional neural networks (CNNs), and deep neural networks (DNNs) are common ways of creating and detecting deepfakes. In our evaluation of over 150 methods, we found that the majority of the focus is on video deepfakes, and, in particular, the generation of video deepfakes. We found that for text deepfakes, there are more generation methods but very few robust methods for detection, including fake news detection, which has become a controversial area of research because of the potential heavy overlaps with human generation of fake content. Our study reveals a clear need to research audio deepfakes and particularly detection of audio deepfakes. This survey has been conducted with a different perspective, compared to existing survey papers that mostly focus on just video and image deepfakes. This survey mainly focuses on audio deepfakes that are overlooked in most of the existing surveys. This article's most important contribution is to critically analyze and provide a unique source of audio deepfake research, mostly ranging from 2016 to 2021. To the best of our knowledge, this is the first survey focusing on audio deepfakes generation and detection in English.

show abstract

Section: Deepfake Categoriesmentioning

confidence: 99%

Section: Deepfake Categoriesmentioning

confidence: 99%

Section: Deepfake Categoriesmentioning

confidence: 99%

See 1 more Smart Citation

Audio deepfakes: A survey

Khanjani

Watson²,

Janeja³

2023

Front. Big Data

View full text Add to dashboard Cite

show abstract

“…Bartusiak et al used normalized gray-scale spectrograms of audio signal for synthetic speech detection using a CNN and a convolution transformer [114,156,157]. While in [158], the authors trained a temporal CNN and a spatial transform network using melspectrograms. For synthetic audio detection, these image-based methods outperformed feature-based methods including the ones using features related to energy, bandwidth, frequency, and short-term transform features such as MFCCs.…”

Section: Image-based Approachesmentioning

confidence: 99%

An Overview of Recent Work in Media Forensics: Methods and Threats

Bhagtani¹,

Yadav²,

Bartusiak³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper, we review recent work in media forensics for digital images, video, audio (specifically speech), and documents. For each data modality, we discuss synthesis and manipulation techniques that can be used to create and modify digital media. We then review technological advancements for detecting and quantifying such manipulations. Finally, we consider open issues and suggest directions for future research.

show abstract

“…Malicious/Selfish nodes introduce variously attacks such as Packet-Flooding [8], [9], [18], [19], Packet-Drooping (there are various categories of Packet-Dropping attacks. Such as, some misbehavior nodes drop all packets, or few misbehavior nodes drop selective packets, not all ) [9], [20], and Fake-Packet attacks (FPA) [21]- [25] to overuse limited resources of networks. Moreover, this would lead to nodes unavailability, high PacketLossRatio (PLR), low PacketDeliveryRatio (PDR), and fake-packets in the network, which further degrade network performance (Due to resources consumption).…”

Section: Introductionmentioning

confidence: 99%

FAPMIC: Fake Packet and Selective Packet Drops Attacks Mitigation by Merkle Hash Tree in Intermittently Connected Networks

et al. 2023

View full text Add to dashboard Cite

Delay/Disruption Tolerant Networks (DTNs) are a special category of IntermittentlyCon-nectedNetworks (ICNs). It has features such as long-delay, frequent-disruption, asymmetrical-data-rates, and high-bundle-error-rates. DTNs have been mainly developed for planet-to-planet networks, commonly known as Inter-Planetary-Networks (IPNs). However, DTNs have shown undimmed potency in challenged communication networks, such as DakNet, ZebraNet, KioskNet and WiderNet. Due to unique characteristics (Intermittent-connectivity and long-delay) DTNs face tough/huge/several challenges in various research areas i.e bundle-forwarding, key-distribution, privacy, bundle-fragmentation, and malicious/selfish nodes particularly. Malicious/selfish nodes launch various catastrophic attacks, this includes, fake packet attacks, selective packet drops attacks, and denial-of-service/flood attacks. These attacks inevitably consume limited resources (persistent-buffer and bandwidth) in DTNs. Fake-packet and selective-packet-drops attacks are top among the challenging attacks in ICNs. The focus of this article is on critical analyses of fake-packet and selective-packet-drops attacks. The panoramic view on misbehavior nodes mitigation algorithms are analyzed, and evaluated mathematically through several parameters for detection probability/accuracy. This article presents a novel algorithm to detects/mitigates fake-packet and selective-packet-drops attacks. Trace-driven simulation results show the proposed algorithm of this article accurately (enhanced detectionaccuracy, reduces false-positive/false-negative rates) detects malicious nodes which launch fake-packet and selective-packet-drops attacks, unlike previously proposed algorithms which detect only one attack (fakepacket or packet-drops at a time) or detect only malicious path (do not exactly detect malicious nodes which launch attacks).

show abstract

A Deep Learning Framework for Audio Deepfake Detection

Cited by 43 publications

References 32 publications

Audio deepfakes: A survey

Audio deepfakes: A survey

An Overview of Recent Work in Media Forensics: Methods and Threats

FAPMIC: Fake Packet and Selective Packet Drops Attacks Mitigation by Merkle Hash Tree in Intermittently Connected Networks

Contact Info

Product

Resources

About