GhostIris: A Password Strength Estimator for Multilingual Users
Popular password strength meters are built around English dictionaries, so they over-rate passwords that people write in their own languages using Latin letters. GhostIris is an open-source, client-side password strength estimator that models how people in South Asia, the Arab world and China actually build passwords. It combines language packs (Roman Urdu, Hinglish, Arabizi, Pinyin, Roman Punjabi, Roman Pashto and Roman Sindhi), a phonetic spelling-variant matcher for Romanised writing, per-language character-level Markov models, and a dynamic-programming guess estimator aligned with NIST SP 800-63B-4. We evaluated GhostIris against zxcvbn 4.4.2 on 1,760 synthetic, reproducible passwords. For culture-specific passwords built from known vocabulary, zxcvbn's median estimate was 10^8.6 guesses against 10^3.8 for GhostIris, and 56.9% were rated "strong" by zxcvbn although GhostIris estimated fewer than 10^10 guesses. On a held-out split, the median gap shrank to 1.5 orders of magnitude and 41.7% of passwords were still over-rated. On random and English-word controls, GhostIris did not produce lower estimates than zxcvbn. GhostIris runs entirely in the browser and is available as a web app, an installable PWA, a web component and a browser extension. Live tool: https://rubab-ashraf.github.io/ghostiris/Source code: https://github.com/rubab-ashraf/ghostiris
Authors
- Rubab Ashraf (ORCID: https://orcid.org/0009-0000-7489-616X)
Institutions
- Government College University, Faisalabad (PK)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24224091
- Primary Topic
- User Authentication and Security Systems
- Type
- preprint