RBE·10 · RBE Deterministic Suite · AML/Fraud Velocity Live

Transaction Velocity Rule Simulator & Explainability Workbench

Author multi-dimensional AML/fraud velocity rules and calibrate them against seeded synthetic transaction streams with time-of-day patterns. Generates confusion matrices, deterministic rule contribution attribution, FATF typology mapping, and EBA GL/2021/05 pre-deployment audit trails. No ML. No PII. Fully client-side.

EBA GL/2021/05 FATF Rec. 10/11 FinCEN SAR 6AMLD PSD2 Art. 96 Policy Mandate Export Zero PII
⚡ Client-side · No server calls
🔒 Zero PII · Seeded PRNG only
📦 MCP-compatible · manifest.json
v RBE Suite v1.0
A — Velocity Rule Library
Scope & reliance — 🔒 All inputs are processed locally in your browser. No data is transmitted. Do not enter real personal data — use synthetic or anonymised inputs only. Embedded rates, thresholds, and regulatory citations are static reference values that may age — verify against current primary sources and your own data before relying on any output for commercial, legal, or compliance decisions. Deterministic logic · no inference · zero PII · CC BY 4.0.
Educational Use Only This tool provides a self-assessment / educational framework for internal planning purposes only. It is not a regulatory audit, legal advice, or a substitute for a formal compliance review by a qualified advisor. Verify all interpretations against the official source text and applicable RTS/ITS/guidance published by the relevant authority.
B — Synthetic Transaction Stream
1000
5%
Merchant Type Mix (total should ≈ 100%)
⏰ Time Distribution (build-critical — drives temporal velocity rules)
Generate stream and run analysis to see calibration results.
Run analysis to see FATF typology mapping.
Run analysis to preview the pre-deployment audit trail.
Run & Export
Analysis Results
About this tool
Seeded PRNGMulberry32 algorithm. Set a seed for bit-for-bit reproducible streams across sessions. Exponential inter-arrival times are drawn per-hour using configurable time distribution weights — temporal velocity rules fire at operationally representative rates.
Contribution attributionDeterministic, not SHAP/LIME. Contribution score for dimension D = weight(D) × normalised distance of observed value from its threshold. Closed-form arithmetic in plain JS — no ML library.
Confusion matrixTP = SAR-flagged transaction that triggered the rule. FP = non-SAR transaction that triggered. Precision, Recall, F1, FPR computed per rule over the full synthetic stream.
Regulatory coverageEBA GL/2021/05 · FATF Rec. 10/11 · FinCEN 31 CFR 1020.320 · PSD2 Art. 96 · 6AMLD
v1.0 
{
  "tool_id": "rbe-10-velocity-rule-simulator",
  "version": "1.0.0",
  "title": "Transaction Velocity Rule Simulator",
  "category": "ai",
  "ap2_export": true,
  "mandate_type": "velocity_rule_mandate",
  "mcp_tool_definition": {
    "name": "simulate_velocity_rules",
    "description": "Author and calibrate AML/fraud
velocity rules against a seeded synthetic
stream. Confusion matrices, contribution
attribution, FATF mapping, EBA audit trail.",
    "inputSchema": {
      "type": "object",
      "required": ["rules"],
      "properties": {
        "rules": { "type": "array" },
        "stream_length": { "type": "integer" },
        "sar_injection_rate": { "type": "number" },
        "random_seed": { "type": "integer" }
      }
    }
  },
  "execution": {
    "type": "browser-javascript",
    "entry": "tools/rbe-10-velocity-rule-simulator.html",
    "function_name": "runAnalysis",
    "timeout_ms": 5000
  }
}