How AFXO Works
AFXO delivers institutional-grade FX data through a robust, multi-layered pipeline that combines data from multiple sources, applies AI-powered quality control, and publishes verified rates on-chain.
Data Pipeline Architecture
Multi-Source Ingestion
We collect FX data from multiple institutional-grade providers, central banks, and market data sources for each currency pair.
- Minimum 3 active sources required per currency
- Real-time health monitoring of all sources
- Automatic failover when sources become unavailable
- Source diversity across provider types
Validation & Sanitization
Raw data passes through multiple validation layers to ensure accuracy and detect anomalies.
- Timestamp freshness verification
- Rate bounds checking against historical ranges
- Cross-source consistency validation
- Stale data detection and filtering
AI Quality Control
Machine learning models analyze incoming data in real-time to detect anomalies, outliers, and potential manipulation.
- Isolation Forest anomaly detection
- Time-series pattern analysis
- Source reliability scoring
- Outlier identification and handling
Weighted Aggregation
Validated rates are combined using a sophisticated weighted averaging algorithm.
- Dynamic source weighting based on reliability
- Recency-weighted calculations
- Outlier-resistant median alternatives
- Confidence-adjusted final rate
Confidence Scoring
Each aggregated rate receives a confidence score (0-100) reflecting data quality and reliability.
- Source agreement factor
- Data freshness factor
- Historical consistency factor
- Source count factor
On-Chain Publication
High-confidence rates are published to smart contracts on multiple blockchains.
- Chainlink-compatible interface
- Multi-chain deployment (Avalanche canonical, Celo, Base, Arbitrum, Solana)
- Gas-optimized update logic
- Minimum confidence threshold enforcement
Understanding Confidence Scores
Every rate published by AFXO includes a confidence score from 0-100. This score reflects the overall reliability and quality of the data.
Multiple sources in strong agreement. Data is fresh and consistent with historical patterns. Suitable for high-stakes applications.
Acceptable source agreement with minor discrepancies. Data quality is good but may warrant additional verification for critical uses.
Source disagreement or data quality concerns detected. Use with caution. On-chain oracles may reject updates below threshold.
Significant data quality issues. Updates are halted. System maintains last known good rate until quality improves.
Minimum Threshold
On-chain oracles enforce a minimum confidence threshold (default: 70%). Updates that don't meet this threshold are rejected, protecting downstream applications from unreliable data. Enterprise customers can configure custom thresholds.
Anti-Manipulation Safeguards
AFXO implements multiple layers of protection against data manipulation and oracle attacks, critical for DeFi applications that depend on accurate pricing.
- Multi-Source Requirement
No single source can influence the final rate. Minimum 3 sources required for publication.
- Rate Change Limits
On-chain contracts enforce maximum per-update change limits to prevent flash loan attacks.
- ML Anomaly Detection
Real-time machine learning identifies unusual patterns that may indicate manipulation attempts.
- Anti-Circularity
AFXO never uses on-chain DEX prices as primary input, preventing circular dependencies and manipulation vectors.
Security Audit Status
Security is a continuous process. We engage third-party auditors regularly and welcome responsible disclosure.
Technical Questions?
Our documentation provides detailed technical specifications, API references, and integration guides for developers.