AI-Driven Mathematical Analysis of Handwritten Word and Pseudoword CAPTCHA Security
Fuente:
Zenodo
Salvato in:
| Autori principali: | , |
|---|---|
| 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 |