3 people 9dda608a86 feat: model-free keyed-Gumbel (EXP) same-key detector (#190) ...

Implements the replay test of ARBI's keyed-Gumbel technical report
(u = PRF(Hash(key, window), token); exact Gamma-tail p-value; repeated-window
masking) as a stdlib-only detector (detect_gumbel.py) — no GPU, model, or
logits — wired into the TextDetector registry, /capabilities and the
iterative Layer B rewrite loop (--gumbel-key; priority gumbel > markllm >
lexical divergence) with a gumbel.before/after/cleared report.

Also cites the ARBI article in the README bibliography (renamed from
References) and names EXP/Gumbel in the open-LLM vendor row. Same-key-only:
valid against the same key, tokenizer, and PRF layout used at generation;
not a vendor oracle. Key is never logged.

Co-authored-by: guillaumemeyer <guillaumemeyer@users.noreply.github.com>

2026-08-19 12:49:11 -07:00

..

audit_dir.py

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

audit_lib.py

fix: an unreadable text file is a failed scan, not a clean one (#169)

2026-08-19 11:27:19 -07:00

audit_website.py

fix: route website binary formats to their real scanners (#177)

2026-08-19 12:05:15 -07:00

av_meta.py

fix(server): preserve image format extensions in /clean and use safe writes in av_meta (#150)

2026-08-18 12:37:26 -07:00

bench_synthid_text.py

feat: multi-scheme MarkLLM text benchmark and detection (#188)

2026-08-19 08:08:44 -07:00

check_staged.py

feat: add pre-commit hook integration for staged-file checking/cleaning (#138)

2026-08-18 07:26:58 -07:00

clean_ctrlregen.py

chore: add Ruff linting and formatting with CI enforcement (#103)

2026-08-16 17:32:40 -07:00

clean_file.py

feat: audio/video AI/C2PA metadata stripping (MP4/MOV, WAV, MP3) (#139)

2026-08-18 07:47:43 -07:00

clean_image.py

feat: add BMP, GIF, TIFF, and EPUB format support (#107)

2026-08-16 19:47:11 -07:00

clean_staged.py

fix: distinguish a failed cleaner from an already-clean file (#159) (#161)

2026-08-19 10:42:45 -07:00

clean_text.py

chore: add Ruff linting and formatting with CI enforcement (#103)

2026-08-16 17:32:40 -07:00

common.py

fix: treat a failed c2patool run as inconclusive, not as "no C2PA" (#156)

2026-08-19 10:34:43 -07:00

container_meta.py

fix: keep collected evidence when a later zip member fails to read (#175)

2026-08-19 11:59:24 -07:00

detect_gumbel.py

feat: model-free keyed-Gumbel (EXP) same-key detector (#190)

2026-08-19 12:49:11 -07:00

detect_text_watermark.py

feat: multi-scheme MarkLLM text benchmark and detection (#188)

2026-08-19 08:08:44 -07:00

format_dispatch.py

feat: audio/video AI/C2PA metadata stripping (MP4/MOV, WAV, MP3) (#139)

2026-08-18 07:47:43 -07:00

image_meta.py

fix: truncated ISOBMFF containers still run the C2PA byte-scan fallback (#176)

2026-08-19 12:03:33 -07:00

inspect_file.py

feat: audio/video AI/C2PA metadata stripping (MP4/MOV, WAV, MP3) (#139)

2026-08-18 07:47:43 -07:00

inspect_image.py

feat: add BMP, GIF, TIFF, and EPUB format support (#107)

2026-08-16 19:47:11 -07:00

inspect_text.py

chore: add Ruff linting and formatting with CI enforcement (#103)

2026-08-16 17:32:40 -07:00

markdiffusion_harness.py

chore: add Ruff linting and formatting with CI enforcement (#103)

2026-08-16 17:32:40 -07:00

requirements-ctrlregen.txt

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

requirements-markdiffusion.txt

feat: split skill from service, add HTTP API and Docker distribution (#60)

2026-08-14 15:42:48 -07:00

requirements-markllm.txt

feat: add reproducible SynthID-text removal benchmark (#145)

2026-08-18 18:13:56 -07:00

requirements-synthid-scorer.txt

feat: split skill from service, add HTTP API and Docker distribution (#60)

2026-08-14 15:42:48 -07:00

rewrite_text.py

feat: model-free keyed-Gumbel (EXP) same-key detector (#190)

2026-08-19 12:49:11 -07:00

score_stylometry.py

fix: report unmeasurable burstiness instead of scoring it as max LLM-like (#141)

2026-08-18 12:14:27 -07:00

score_synthid.py

feat: vendor text-watermark detection (Gemini SynthID, Claude seam, MarkLLM) + SynthID image scorer sidecar (#109)

2026-08-17 18:31:22 -07:00

server.py

fix: treat a failed c2patool run as inconclusive, not as "no C2PA" (#156)

2026-08-19 10:34:43 -07:00

setup_ctrlregen.ps1

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

setup_ctrlregen.sh

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

setup_markdiffusion.sh

feat: split skill from service, add HTTP API and Docker distribution (#60)

2026-08-14 15:42:48 -07:00

setup_markllm.sh

feat: add reproducible SynthID-text removal benchmark (#145)

2026-08-18 18:13:56 -07:00

setup_synthid.ps1

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

setup_synthid.sh

fix: correctness and security hardening of the cleaning scripts (#122) (#126)

2026-08-17 17:15:35 -07:00

synthid_score_server.py

feat: vendor text-watermark detection (Gemini SynthID, Claude seam, MarkLLM) + SynthID image scorer sidecar (#109)

2026-08-17 18:31:22 -07:00

text_detectors.py

feat: model-free keyed-Gumbel (EXP) same-key detector (#190)

2026-08-19 12:49:11 -07:00

text_unicode.py

feat: consolidated Layer A hardening (missed carriers, reserved ignorables, noncharacters, layout format controls) (#133)

2026-08-19 10:12:54 -07:00