AI-Driven Mathematical Analysis of Handwritten Word and Pseudoword CAPTCHA Security

Fuente: Zenodo
Salvato in:
Dettagli Bibliografici
Autori principali: Dimple, Mohinder Kumar
Natura: Recurso digital
Pubblicazione: Zenodo 2025
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866902002933432320
author Dimple
Mohinder Kumar
author_facet Dimple
Mohinder Kumar
contents <p>Handwritten CAPTCHAs are widely used to differentiate human users from automated systems by exploiting the natural variability of handwriting. However, recent advances in artificial intelligence (AI), especially OCRbased recognizers, have raised concerns about their true security strength. This paper presents an AI-driven and mathematically grounded analysis of handwritten word and pseudoword CAPTCHA schemes from a cybersecurity perspective. A complete attack pipeline is constructed using convolution-based smoothing, probabilistic thresholding, and set-theoretic morphological transformations to prepare CAPTCHA images for AI-based decoding. OCR is modeled as a probabilistic classifier, and its breaking success is evaluated statistically across a large dataset. Entropy calculations and comparative difficulty analysis are used to assess the robustness of the CAPTCHA designs. The results demonstrate that many handwritten CAPTCHAs exhibit low entropy and are vulnerable to modern AI solvers, posing a potential cybersecurity risk. This work shows how applied mathematics and AI can jointly evaluate and reveal weaknesses in security mechanisms.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18043738
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle AI-Driven Mathematical Analysis of Handwritten Word and Pseudoword CAPTCHA Security
Dimple
Mohinder Kumar
<p>Handwritten CAPTCHAs are widely used to differentiate human users from automated systems by exploiting the natural variability of handwriting. However, recent advances in artificial intelligence (AI), especially OCRbased recognizers, have raised concerns about their true security strength. This paper presents an AI-driven and mathematically grounded analysis of handwritten word and pseudoword CAPTCHA schemes from a cybersecurity perspective. A complete attack pipeline is constructed using convolution-based smoothing, probabilistic thresholding, and set-theoretic morphological transformations to prepare CAPTCHA images for AI-based decoding. OCR is modeled as a probabilistic classifier, and its breaking success is evaluated statistically across a large dataset. Entropy calculations and comparative difficulty analysis are used to assess the robustness of the CAPTCHA designs. The results demonstrate that many handwritten CAPTCHAs exhibit low entropy and are vulnerable to modern AI solvers, posing a potential cybersecurity risk. This work shows how applied mathematics and AI can jointly evaluate and reveal weaknesses in security mechanisms.</p>
title AI-Driven Mathematical Analysis of Handwritten Word and Pseudoword CAPTCHA Security
url https://doi.org/10.5281/zenodo.18043738