Searcharxiv⌕ Search

arXiv subjects

Isao Echizen

Publications and source records attributed to Isao Echizen.

119 records · Page 7Linked to original sources

Capsule-Forensics: Using Capsule Networks to Detect Forged Images and Videos

Recent advances in media generation techniques have made it easier for attackers to create forged images and videos. State-of-the-art methods enable the real-time creation of a forged version of a single video obtained from a social network. Although numerous methods have been developed for detecting forged images and videos, they are generally targeted at certain domains and quickly become obsolete as new kinds of attacks appear. The method introduced in this paper uses a capsule network to detect various kinds of spoofs, from replay attacks using printed images or recorded videos to computer-generated videos using deep convolutional neural networks. It extends the application of capsule networks beyond their original intention to the solving of inverse graphics problems.

cs.CV↗

Efficient (nonrandom) construction and decoding for non-adaptive group testing

The task of non-adaptive group testing is to identify up to $d$ defective items from $N$ items, where a test is positive if it contains at least one defective item, and negative otherwise. If there are $t$ tests, they can be represented as a $t \times N$ measurement matrix. We have answered the question of whether there exists a scheme such that a larger measurement matrix, built from a given $t\times N$ measurement matrix, can be used to identify up to $d$ defective items in time $O(t \log_2{N})$. In the meantime, a $t \times N$ nonrandom measurement matrix with $t = O \left(\frac{d^2 \log_2^2{N}}{(\log_2(d\log_2{N}) - \log_2{\log_2(d\log_2{N})})^2} \right)$ can be obtained to identify up to $d$ defective items in time $\mathrm{poly}(t)$. This is much better than the best well-known bound, $t = O \left( d^2 \log_2^2{N} \right)$. For the special case $d = 2$, there exists an efficient nonrandom construction in which at most two defective items can be identified in time $4\log_2^2{N}$ using $t = 4\log_2^2{N}$ tests. Numerical results show that our proposed scheme is more practical than existing ones, and experimental results confirm our theoretical analysis. In particular, up to $2^{7} = 128$ defective items can be identified in less than $16$s even for $N = 2^{100}$.

cs.IT↗

Efficiently Decodable Non-Adaptive Threshold Group Testing

We consider non-adaptive threshold group testing for identification of up to $d$ defective items in a set of $n$ items, where a test is positive if it contains at least $2 \leq u \leq d$ defective items, and negative otherwise. The defective items can be identified using $t = O \left( \left( \frac{d}{u} \right)^u \left( \frac{d}{d - u} \right)^{d-u} \left(u \log{\frac{d}{u}} + \log{\frac{1}ε} \right) \cdot d^2 \log{n} \right)$ tests with probability at least $1 - ε$ for any $ε> 0$ or $t = O \left( \left( \frac{d}{u} \right)^u \left( \frac{d}{d -u} \right)^{d - u} d^3 \log{n} \cdot \log{\frac{n}{d}} \right)$ tests with probability 1. The decoding time is $t \times \mathrm{poly}(d^2 \log{n})$. This result significantly improves the best known results for decoding non-adaptive threshold group testing: $O(n\log{n} + n \log{\frac{1}ε})$ for probabilistic decoding, where $ε> 0$, and $O(n^u \log{n})$ for deterministic decoding.

cs.IT↗

Transforming acoustic characteristics to deceive playback spoofing countermeasures of speaker verification systems

Automatic speaker verification (ASV) systems use a playback detector to filter out playback attacks and ensure verification reliability. Since current playback detection models are almost always trained using genuine and played-back speech, it may be possible to degrade their performance by transforming the acoustic characteristics of the played-back speech close to that of the genuine speech. One way to do this is to enhance speech "stolen" from the target speaker before playback. We tested the effectiveness of a playback attack using this method by using the speech enhancement generative adversarial network to transform acoustic characteristics. Experimental results showed that use of this "enhanced stolen speech" method significantly increases the equal error rates for the baseline used in the ASVspoof 2017 challenge and for a light convolutional neural network-based method. The results also showed that its use degrades the performance of a Gaussian mixture model-universal background model-based ASV system. This type of attack is thus an urgent problem needing to be solved.

cs.SD↗

MesoNet: a Compact Facial Video Forgery Detection Network

This paper presents a method to automatically and efficiently detect face tampering in videos, and particularly focuses on two recent techniques used to generate hyper-realistic forged videos: Deepfake and Face2Face. Traditional image forensics techniques are usually not well suited to videos due to the compression that strongly degrades the data. Thus, this paper follows a deep learning approach and presents two networks, both with a low number of layers to focus on the mesoscopic properties of images. We evaluate those fast networks on both an existing dataset and a dataset we have constituted from online videos. The tests demonstrate a very successful detection rate with more than 98% for Deepfake and 95% for Face2Face.

cs.CV↗

Transformation on Computer-Generated Facial Image to Avoid Detection by Spoofing Detector

Making computer-generated (CG) images more difficult to detect is an interesting problem in computer graphics and security. While most approaches focus on the image rendering phase, this paper presents a method based on increasing the naturalness of CG facial images from the perspective of spoofing detectors. The proposed method is implemented using a convolutional neural network (CNN) comprising two autoencoders and a transformer and is trained using a black-box discriminator without gradient information. Over 50% of the transformed CG images were not detected by three state-of-the-art spoofing detectors. This capability raises an alarm regarding the reliability of facial authentication systems, which are becoming widely used in daily life.

cs.CV↗

High-quality nonparallel voice conversion based on cycle-consistent adversarial network

Although voice conversion (VC) algorithms have achieved remarkable success along with the development of machine learning, superior performance is still difficult to achieve when using nonparallel data. In this paper, we propose using a cycle-consistent adversarial network (CycleGAN) for nonparallel data-based VC training. A CycleGAN is a generative adversarial network (GAN) originally developed for unpaired image-to-image translation. A subjective evaluation of inter-gender conversion demonstrated that the proposed method significantly outperformed a method based on the Merlin open source neural network speech synthesis system (a parallel VC system adapted for our setup) and a GAN-based parallel VC system. This is the first research to show that the performance of a nonparallel VC method can exceed that of state-of-the-art parallel VC methods.

eess.AS↗

Efficient Decoding Schemes for Noisy Non-Adaptive Group Testing when Noise Depends on Number of Items in Test

The goal of non-adaptive group testing is to identify at most $d$ defective items from $N$ items, in which a test of a subset of $N$ items is positive if it contains at least one defective item, and negative otherwise. However, in many cases, especially in biological screening, the outcome is unreliable due to biochemical interaction; i.e., \textit{noise.} Consequently, a positive result can change to a negative one (false negative) and vice versa (false positive). In this work, we first consider the dilution effect in which \textit{the degree of noise depends on the number of items in the test}. Two efficient schemes are presented for identifying the defective items in time linearly to the number of tests needed. Experimental results validate our theoretical analysis. Specifically, setting the error precision of 0.001 and $d\leq16$, our proposed algorithms always identify all defective items in less than 7 seconds for $N=2^{33}\approx 9$ billion.

cs.IT↗

Can we steal your vocal identity from the Internet?: Initial investigation of cloning Obama's voice using GAN, WaveNet and low-quality found data

Thanks to the growing availability of spoofing databases and rapid advances in using them, systems for detecting voice spoofing attacks are becoming more and more capable, and error rates close to zero are being reached for the ASVspoof2015 database. However, speech synthesis and voice conversion paradigms that are not considered in the ASVspoof2015 database are appearing. Such examples include direct waveform modelling and generative adversarial networks. We also need to investigate the feasibility of training spoofing systems using only low-quality found data. For that purpose, we developed a generative adversarial network-based speech enhancement system that improves the quality of speech data found in publicly available sources. Using the enhanced data, we trained state-of-the-art text-to-speech and voice conversion models and evaluated them in terms of perceptual speech quality and speaker similarity. The results show that the enhancement models significantly improved the SNR of low-quality degraded data found in publicly available sources and that they significantly improved the perceptual cleanliness of the source speech without significantly degrading the naturalness of the voice. However, the results also show limitations when generating speech with the low-quality found data.

eess.AS↗

Non-Adaptive Group Testing Framework based on Concatenation Code

We consider an efficiently decodable non-adaptive group testing (NAGT) problem that meets theoretical bounds. The problem is to find a few specific items (at most $d$) satisfying certain characteristics in a colossal number of $N$ items as quickly as possible. Those $d$ specific items are called \textit{defective items}. The idea of NAGT is to pool a group of items, which is called \textit{a test}, then run a test on them. If the test outcome is \textit{positive}, there exists at least one defective item in the test, and if it is \textit{negative}, there exists no defective items. Formally, a binary $t \times N$ measurement matrix $\mathcal{M} = (m_{ij})$ is the representation for $t$ tests where row $i$ stands for test $i$ and $m_{ij} = 1$ if and only if item $j$ belongs to test $i$. There are three main objectives in NAGT: minimize the number of tests $t$, construct matrix $\mathcal{M}$, and identify defective items as quickly as possible. In this paper, we present a strongly explicit construction of $\mathcal{M}$ for when the number of defective items is at most 2, with the number of tests $t \simeq 16 \log{N} = O(\log{N})$. In particular, we need only $K \simeq N \times 16\log{N} = O(N\log{N})$ bits to construct such matrices, which is optimal. Furthermore, given these $K$ bits, any entry in the matrix can be constructed in time $O \left(\ln{N}/ \ln{\ln{N}} \right)$. Moreover, $\mathcal{M}$ can be decoded with high probability in time $O\left( \frac{\ln^2{N}}{\ln^2{\ln{N}}} \right)$. When the number of defective items is greater than 2, we present a scheme that can identify at least $(1-ε)d$ defective items with $t \simeq 32 C(ε) d \log{N} = O(d \log{N})$ in time $O \left( d \frac{\ln^2{N}}{\ln^2{\ln{N}}} \right)$ for any close-to-zero $ε$, where $C(ε)$ is a constant that depends only on $ε$.

cs.IT↗

Mix-ORAM: Using delegate shuffles

Oblivious RAM (ORAM) is a key technology for providing private storage and querying on untrusted machines but is commonly seen as impractical due to the high overhead of the re-randomization, called the eviction, the client incurs. We propose in this work to securely delegate the eviction to semi-trusted third parties to enable any client to accede the ORAM technology and present four different designs inspired by mix-net technologies with reasonable periodic costs.

cs.CR↗