antenore/voynich-toolkit: Phase 26: Crib Attack — 78 folios decoded

Read full text Summarise with VIRAT
Authors: Antenore Gatta
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Phase 26: Known-Plaintext Crib Attack with Cross-Page Bootstrap First large-scale decoding of the Voynich manuscript using a reverse cipher + LLM crib generation approach. Key Results 78 folios analyzed with word-level crib attack 883 exact matches (d=0) across 22 bootstrapped pages 81.2% average coverage after 2 bootstrap rounds z-score > 100: 0 matches in 200 permutations of real Hebrew words (176,600 attempts) 555 unique Hebrew words confirmed through cipher roundtrip 151 words confirmed across 2+ independent pages New Features Section-adaptive prompts (herbal, pharmaceutical, balneological, zodiac, astronomical, cosmological, text) Cross-page bootstrap vocabulary with leave-one-out validation build_crib_vocabulary() and enhanced identify_anchors() with crib-confirmed anchors --force flag for re-running pages with enriched vocabulary Batch processing: scripts/decode_all_pages.py --workers 2 --force Methodology Extract page constraints from EVA/IVTFF transcription Identify anchor words (DB glosses + cross-page crib vocabulary) Generate plausible Hebrew candidates via Claude Sonnet API Encode candidates through the cipher, compare to real EVA text Bootstrap: exact matches become anchors for subsequent rounds Files Included output/stats/crib_wordlevel_*.json — word-level match results per folio output/readings/*.txt — page-level translations/interpretations

Fields of Study & Subjects

Digital Humanities and Scholarship Libraries Manuscripts and Books Cryptographic Implementations and Security

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN