Publication detail

SEMI-SUPERVISED APPROACH TO TRAIN CAPTCHA LETTER POSITION DETETOR

BOŠTÍK, O.

Original Title

SEMI-SUPERVISED APPROACH TO TRAIN CAPTCHA LETTER POSITION DETETOR

Type

conference paper

Language

English

Original Abstract

Common Optical Character Recognition (OCR) methods benefit from the fact, that the text is distributed in images in a predictable pattern. This is not the situation with CAPTCHA systems. Utilizing OCR algorithms to overcome common web anti-abuse CAPTCHA systems is therefore a challenging task. To train a system to overcome any CAPTCHA scheme, an attacker needs a huge dataset of annotated images. And for some methods, the attacker needs not only the right answers but also an exact position of the character in the CAPTCHA image. Annotate the positions of the object in an image is a time-consuming task. In this paper, we propose a system, which can help to annotate the position of CAPTCHA character with minimal human interaction. After annotating a small sample of targeted CAPTCHA images, a YOLO-based region detection deep network is used to search for the characters’ locations.

Keywords

OCR, CAPTCHA, Deep learning, YOLO v2, semi-supervised learning, MATLAB

Authors

BOŠTÍK, O.

Released

26. 4. 2021

Publisher

Vysoké učené Technické, Fakulta elektrotechniky a komunikačních technologií

Location

Brno

ISBN

978-80-214-5942-7

Book

Proceedings of the 27nd Conference STUDENT EEICT 2018

Edition number

1

Pages from

436

Pages to

440

Pages count

5

BibTex

@inproceedings{BUT171159,
  author="Ondřej {Boštík}",
  title="SEMI-SUPERVISED APPROACH TO TRAIN CAPTCHA LETTER POSITION DETETOR",
  booktitle="Proceedings of the 27nd Conference STUDENT EEICT 2018",
  year="2021",
  number="1",
  pages="436--440",
  publisher="Vysoké učené Technické, Fakulta elektrotechniky a komunikačních technologií",
  address="Brno",
  isbn="978-80-214-5942-7"
}