RUMBO Security Intelligence · v1.0.0
RUMBO Guardian is an independent security intelligence layer for detecting and explaining signals associated with phishing, fraud, impersonation and suspicious navigation.
It follows a privacy-first, local-first architecture: technical analysis runs on the operator's device, evidence remains inspectable, and contextual trust decisions never erase underlying technical indicators.
Portfolio demo: https://rumbo-ia.github.io/rumbo-guardian/
RUMBO public surfaces: rumbo.verso.fans · @RumboAGI on X
- Explainable 0–100 risk scoring for messages and URLs.
- Detection of urgency, credential requests, payment signals and suspicious domains.
- Local trusted / blocked domain context.
- Tamper-evident Evidence Ledger with SHA-256 hash chaining, local integrity verification and JSON export.
- Individual risk-report export.
- Installable PWA for desktop use.
- Brave / Chrome extension for explicit, on-demand inspection of the active tab.
- Context-menu analysis for links and selected text before navigation or follow-up action.
- Cross-origin form destination analysis, with stronger evidence when password fields submit off-site.
- Bounded redirect-chain analysis up to three hops, with cycle, depth and parameter guards.
- Ranking of visible links by risk.
- High-risk local alerts.
The shared risk engine is separated from the operator-facing web UI and browser extension. Trusted-domain context changes operator context, not the technical evidence: a trusted entry does not automatically convert a technically suspicious signal into a safe result.
On Windows, run START_GUARDIAN.cmd and open http://127.0.0.1:8766/.
The application can also be served with the included Node.js local server.
- Open
brave://extensionsorchrome://extensions. - Enable Developer mode.
- Choose Load unpacked.
- Select the
extensiondirectory. - Open Guardian from a normal web page and choose Analizar esta pestaña.
- Or right-click a link / selected text and choose Analizar … con RUMBO Guardian.
The extension requests activeTab, scripting, storage and contextMenus, with no broad host_permissions. Page inspection and context analysis are initiated explicitly by the operator. Context-menu payloads are passed through ephemeral storage.session entries and removed after the report consumes them.
The repository includes repeatable unit, portability, portfolio, real-browser extension and real-browser Evidence Ledger integrity tests. The integration harness uses an isolated Brave profile and a controlled high-risk fixture.
Run npm run check and npm test. On Windows, tests/run-extension-integration.ps1 exercises the extension and tests/run-ledger-browser-integration.ps1 verifies ledger integrity and tamper detection in isolated Brave sessions.
The portfolio screenshot is reproducible with scripts/capture-portfolio.ps1 and uses synthetic test data only.
See EVIDENCE.md for the verified baseline and SECURITY.md for security boundaries.
PRODUCT.md— product principles and scope.PORTFOLIO.md— portfolio narrative and demonstrated engineering capabilities.SECURITY.md— defensive scope and security boundaries.EVIDENCE.md— verification evidence and reproducibility notes.
V0.4.2 introduced a standalone Node.js verifier for exported Evidence Ledgers. It recalculates the SHA-256 chain without opening Guardian or trusting the embedded integrity field.
node tools/verify-ledger.js portfolio/demo-ledger-v042.json
node tools/verify-ledger.js <exported-ledger.json> --jsonExit code 0 means the chain is valid, 2 means the ledger is structurally invalid or tampered, and 1 means the file could not be read. The committed demo ledger is synthetic and reproducible.
This portfolio release does not grant an open-source license. Source is published for evaluation and demonstration; all rights remain reserved unless a later release states otherwise.
Guardian now inspects common redirect parameters such as url, redirect, continue, next, target and destination. When one of those parameters contains an absolute HTTP(S) destination on a different domain, Guardian analyzes that nested destination locally and surfaces both the cross-domain redirect and any risk signals found inside it. Ordinary query parameters that merely contain URL-shaped text are not treated as redirects.
Guardian now inspects form submission destinations during explicit active-tab analysis. Cross-origin form actions are surfaced as explainable evidence, and password-bearing forms that submit to another site receive a stronger signal. Empty and same-site form actions remain non-penalized. No form is submitted by Guardian.
Guardian now identifies javascript:, data: and vbscript: URL schemes as active-content navigation rather than treating them as ordinary low-risk URLs. The shared app/extension engine raises an explicit active_content_scheme signal while leaving normal HTTPS and non-active schemes such as mailto: outside this rule.
Guardian now performs bounded decoding of recognized redirect parameters before navigation analysis, allowing a destination hidden behind a second percent-encoding layer to be inspected locally. Decoding is capped, applies only to known redirect parameters, preserves same-site behavior, and does not fetch or follow the destination.
Guardian now follows recognized redirect parameters locally for up to three hops, preserving an explainable chain and propagating risk from nested destinations. Repeated routes stop recursion, excessive depth is surfaced explicitly, and each URL processes at most four recognized redirect targets. No network redirect is followed or fetched.