Image‐based CAPTCHAs based on neural style transfer

Cheng, Zhouhang; Gao, Haichang; Liu, Zhongyu; Wu, Huaxi; Yang, Zi; Pei, Ge

doi:10.1049/iet-ifs.2018.5036

Cited by 9 publications

(9 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Increasing the data variety in the dataset can enrich the features, thereby increasing recognition performance. We analyzed the performance between NSGT and other augmentation methods that only use Neural Style Transfer [11][12][13][14]. Neural Style Transfer can provide variations in the color of the augmented image.…”

Section: Analysis and Discussionmentioning

confidence: 99%

“…This is using all layers to extract style information from the VGG19 network. The amount of correlation present between feature maps in each layer is used to measure the style information [14]. Style loss is the difference in correlation present between feature maps, which are calculated by the resulting image and the style image.…”

Section: Neural Style Transfermentioning

confidence: 99%

“…Zhao [13] applied for Neural Style Transfer with a combination of local and global optimization to reduce artifacts in the resulting image. Related research on Neural Style Transfer has been conducted by Cheng [14] to produce more variations of image-based captcha aimed at increasing security. This research resulted in good usability captcha based on images and fonts.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Neural Style Transfer and Geometric Transformations for Data Augmentation on Balinese Carving Recognition using MobileNet

Darma¹,

Suciati²,

Siahaan³

2020

IJIES

View full text Add to dashboard Cite

The preservation of Balinese carving data is a challenge in recognition of Balinese carving. Balinese carvings are a cultural heritage found in traditional buildings in Bali. The collection of Balinese carving images from public images can be a solution for preserving cultural heritage. However, the lousy quality of taking photographs, e.g., skewed shots, can affect the recognition results. Research on the Balinese carving recognition has existed but only recognizes a predetermined image. We proposed a Neural Style Geometric Transformation (NSGT) as a data augmentation technique for Balinese carvings recognition. NSGT is combining Neural Style Transfers and Geometric Transformations for a small dataset solution. This method provides variations in color, lighting, rotation, rescale, zoom, and the size of the training dataset, to improve recognition performance. We use MobileNet as a feature extractor because it has a small number of parameters, which makes it suitable to be applied on mobile devices. Eight scenarios were tested based on image styles and geometric transformations to get the best results. Based on the results, the proposed method can improve accuracy by up to 16.2%.

show abstract

Section: Analysis and Discussionmentioning

confidence: 99%

Section: Neural Style Transfermentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Neural Style Transfer and Geometric Transformations for Data Augmentation on Balinese Carving Recognition using MobileNet

Darma¹,

Suciati²,

Siahaan³

2020

IJIES

View full text Add to dashboard Cite

show abstract

“…This method is characterized to avoid attack by manipulating the test data against the pre-trained CAPTCHA system. The third method is the combination of grid-CAPTCHA and font-CAPTCHA [24]. The grid-CAPTCHA method selects the relevant CAPTCHA image related to the provided sentence.…”

Section: Other Methods For Generating Captchasmentioning

confidence: 99%

CAPTCHA Image Generation: Two-Step Style-Transfer Learning in Deep Neural Networks

Kwon

Yoon

Park

2020

Sensors

View full text Add to dashboard Cite

Mobile devices such as sensors are used to connect to the Internet and provide services to users. Web services are vulnerable to automated attacks, which can restrict mobile devices from accessing websites. To prevent such automated attacks, CAPTCHAs are widely used as a security solution. However, when a high level of distortion has been applied to a CAPTCHA to make it resistant to automated attacks, the CAPTCHA becomes difficult for a human to recognize. In this work, we propose a method for generating a CAPTCHA image that will resist recognition by machines while maintaining its recognizability to humans. The method utilizes the style transfer method, and creates a new image, called a style-plugged-CAPTCHA image, by incorporating the styles of other images while keeping the content of the original CAPTCHA. In our experiment, we used the TensorFlow machine learning library and six CAPTCHA datasets in use on actual websites. The experimental results show that the proposed scheme reduces the rate of recognition by the DeCAPTCHA system to 3.5% and 3.2% using one style image and two style images, respectively, while maintaining recognizability by humans.

show abstract

“…The Internet Security Threat Reports (ISTR) on different intelligent threats, mentioned in Table-1, reflects that security aspects must be enhanced globally. The CAPTCHA tackles such threats by discriminating between human and automated programs (Cheng et al, 2019;Gao et al, 2019;Belk et al, 2015;Zi et al, 2020). A robust CAPTCHA design algorithm (CDA) can endow improved security, which is intractable to bots.…”

Section: Introductionmentioning

confidence: 99%

Two-stage human verification using HandCAPTCHA and anti-spoofed finger biometrics with feature selection

Bera

Bhattacharjee

Shum

2021

Expert Systems with Applications

View full text Add to dashboard Cite

Image‐based CAPTCHAs based on neural style transfer

Cited by 9 publications

References 35 publications

Neural Style Transfer and Geometric Transformations for Data Augmentation on Balinese Carving Recognition using MobileNet

Neural Style Transfer and Geometric Transformations for Data Augmentation on Balinese Carving Recognition using MobileNet

CAPTCHA Image Generation: Two-Step Style-Transfer Learning in Deep Neural Networks

Two-stage human verification using HandCAPTCHA and anti-spoofed finger biometrics with feature selection

Contact Info

Product

Resources

About