Repository navigation
Releases
Fortify v1.2.0 — The Keep
A medieval/gothic overhaul of the dashboard, plus a fleet-wide command view. The scanner and analysis engine are unchanged; this release is all about how you see and work with your findings.
Command dashboard
- A new Command landing screen with a stained-glass stat façade (patrols run / targets watched / reports drafted), a shield-based severity rollup, a Threats-sighted breakdown, and recent activity.
The Scriptorium
- An AI-reports panel with a candle chart over the last 7 days, provider-aware: Ollama burns blue, Gemini burns orange, and a day with both shows a candle for each side by side. A per-model share bar shows the real split.
Screens & chrome
- Siege Log — themed scan table with Watch/Siege toggle, severity chips, and pagination.
- Counsel — AI-provider settings as gothic radio cards (Ollama azure / Gemini orange).
- Carved-stone material system, hand-inked icon set, a layered stone background, and a collapsible sidebar.
Under the hood
- Each analysis now records which provider/model produced it (feeds the Scriptorium).
- The stained-glass façade scales responsively to fit narrower screens.
Full set: #42 (Command dashboard), #43 (Keep App theme), closes #41.
Fortify v1.1.0 — Expanded Scanner
Scanner-depth release: broader detection and reproducible AI analysis.
New detection
- Blind SQL injection (boolean-based) (#8) — infers injectability from TRUE/FALSE response differentials, catching injections that leak no database error (which the error-based check misses).
- Header version detection (#1, #29) — extracts software + version from leaky headers (e.g.
Server: nginx/1.18.0) and surfaces it in the Battle Report; version-less headers are ignored. - Per-scan XSS reflection tokens (#9) — unique random canary per scan instead of a static marker, eliminating false positives from pre-existing page content.
Trust & UX
- Reproducible analysis (#32) — temperature 0 (+ fixed seed for Ollama) so re-analyzing the same scan yields the same scores and summary.
- Specific failure messages (#27) — AI failures now say what went wrong (e.g. Gemini high demand / rate limit) instead of a generic error.
Run it
docker compose up --buildDashboard at http://localhost:5173, API at http://localhost:8500. See the README for full docs.
Follow-ups toward v1.2.0: time-based & union SQLi (#8 siblings), deterministic in-code scoring (#31), and dashboard enhancements (#33, #34).
Fortify v1.0.0
First feature-complete release of Fortify — an open-source web application security scanner with AI-assisted analysis.
Highlights
- Scanner — passive checks (TLS, security headers, sensitive-path exposure via content-signature detection) and opt-in, consent-gated active checks (SQLi, XSS, path traversal).
- AI Analyzer — grounded LLM risk reports (findings extracted in code, LLM only explains/scores), local by default via Ollama; optional Google Gemini (BYOK). Runs non-blocking as a background task.
- Dashboard — React + Tailwind console: Command (scan + live Siege Log), Battle Report (findings + AI analysis + PDF export), Settings (provider/BYOK). Hash-based routing so refresh/back work.
- Runnable by anyone — one-command Docker stack (
docker compose up --build) with a persistent volume; backend reaches host Ollama via host.docker.internal. - Hardening — unguessable scan IDs (no enumeration/IDOR), content-signature exposed-path detection (fewer false positives).
Run it
docker compose up --buildDashboard at http://localhost:5173, API at http://localhost:8500.
See the README for full docs. v1.0 milestone: all issues closed.