FractalTextGuard: Offline DFA/LRD & Fractal-Dimension Toolkit for Structural Text Authenticity (v2.5 Golden Master)

FractalTextGuard is an offline Python toolkit for structural text authenticity analysis using Detrended Fluctuation Analysis (DFA-n), Long-Range Dependence (LRD) proxies, and box-counting fractal dimension. The project targets reproducible, auditable, and dependency-free inspection of text and code-heavy documents where classical stylometry is brittle. What it measures (metrology scope) FractalTextGuard reports interpretable structural descriptors, including: Hurst exponent (H) via DFA-2/3/4 (trend-robust detrending orders) Standard error and CI95(H) from log-log regression diagnostics Box-counting dimension (D_box) of the signal/trajectory (scaling region selection) Compression ratio, Shannon entropy, and n-gram repetition signals Optional bootstrap stability and surrogate/null tests (for robustness) Intended use academic integrity checks (structure-based screening, not authorship proof) cybersecurity triage (bot/phishing content screening) document forensics & authenticity workflows high-volume batch inspection with audit logs Offline & privacy By default the tool runs fully offline: no data leaves the system. A strict offline mode is provided to enforce zero network activity. Reproducibility The tool supports deterministic runs (seeded randomness, deterministic subsampling) and reproducible reporting workflows (fixed anchors, stable cache keys, and portable cache entry import/export when delta-H comparisons are needed). Limitations (explicit) FractalTextGuard does not “prove authorship” and does not compute semantic truthfulness. Results are structural indicators; adversarial imitation is possible, so the recommended workflow is: metrics → stability tests → human review / external model. “Industrial highlights” stdlib-only (no external deps)strict-offline modedeterministic subsampling for box-countingDFA-2/3/4 with CI/SE diagnosticsexport/import cache entry for delta-H reproducibilitybatch processing + JSON reports for pipelines Licensing / non-commercial policy Non-commercial use is permitted under the selected license. Commercial use requires explicit permission from the author.