While text-based CAPTCHAs have been the predominant type of human interaction proofs (HIPs) for many years, image recognition challenges have also gained significant attention. This trend is due, on one hand, to groundbreaking advancements in solving text CAPTCHAs and, on the other hand, to the intrinsic weakness of machines in dealing with cognitive tasks such as those presented in image CAPTCHAs. In addition to classic and even human-centric image CAPTCHA solvers, deep learning has recently emerged as a significant player, providing two unprecedented and, at the same time, contradictory advantages for designers and adversaries. Designers benefit from deep learning techniques to make CAPTCHAs as hard to break as possible, while adversaries utilize deep learning algorithms to attack novel and complicated image-based challenges. Given these premises, this paper presents an analytical study on the applications of deep learning for and against image CAPTCHAs. This study aims to provide a comprehensive overview of the latest advancements in the field, assisting researchers and practitioners in designing image CAPTCHAs that are both user-friendly and resilient against modern attacks.

Image CAPTCHAs: When Deep Learning Breaks the Mold

Moradi, Morteza
;
Moradi, Mohammad;Palazzo, Simone;Spampinato, Concetto
2024-01-01

Abstract

While text-based CAPTCHAs have been the predominant type of human interaction proofs (HIPs) for many years, image recognition challenges have also gained significant attention. This trend is due, on one hand, to groundbreaking advancements in solving text CAPTCHAs and, on the other hand, to the intrinsic weakness of machines in dealing with cognitive tasks such as those presented in image CAPTCHAs. In addition to classic and even human-centric image CAPTCHA solvers, deep learning has recently emerged as a significant player, providing two unprecedented and, at the same time, contradictory advantages for designers and adversaries. Designers benefit from deep learning techniques to make CAPTCHAs as hard to break as possible, while adversaries utilize deep learning algorithms to attack novel and complicated image-based challenges. Given these premises, this paper presents an analytical study on the applications of deep learning for and against image CAPTCHAs. This study aims to provide a comprehensive overview of the latest advancements in the field, assisting researchers and practitioners in designing image CAPTCHAs that are both user-friendly and resilient against modern attacks.
2024
deep learning
CAPTCHAs
Deep learning
Accuracy
Task analysis
Image recognition
Character recognition
Transformers
Object recognition
CAPTCHA protection
image CAPTCHA
object recognition
File in questo prodotto:
File Dimensione Formato  
Image_CAPTCHAs_When_Deep_Learning_Breaks_the_Mold.pdf

accesso aperto

Tipologia: Versione Editoriale (PDF)
Licenza: Creative commons
Dimensione 1.83 MB
Formato Adobe PDF
1.83 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11769/640733
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact