This article was first published on Chainlink
Chainlink Price Feeds are purpose-built to provide DeFi applications with the maximum amount of price oracle security, reliability, and data quality. These properties are generated by focusing on several key design features:
- Decentralization at both the oracle node and data source levels to prevent any single points of failure
- Selection of secure node operators and premium data sources to ensure every individual component is highly reliable
- Provable security with on-chain performance metrics that users can independently monitor
- Crypto-economic incentives to ensure price feed updates occur even during extreme blockchain network congestion and unexpected hard forks.
These strong guarantees have led to the widespread adoption of Chainlink Price Feeds, which now secure billions of dollars in user funds across DeFi. As the amount of value being secured by Chainlink grows, we continue to explore additional ways in which developers can further secure their smart contracts. While Chainlink Price Feeds are designed to guarantee reliability even during the recent flash loan attacks, there are additional layers of redundancy in the form of circuit breakers and Chainlink client diversity that generate hyper-reliability for Chainlink Price Feeds, protecting user funds even amidst an unexpected black swan event.
The Multiple Layers of Redundancy in Chainlink Price Feeds
To ensure the highest level of reliability and tamper-resistance, Chainlink Price Feeds feature three layers of aggregation to filter out outlier data and hedge against black swan events.
First, Chainlink Price Feeds source data from professional data aggregators (e.g. BraveNewCoin), which fetch market data from all trading environments (both off-chain and on-chain exchanges) to create a single, volume-adjusted market-wide average price. Second, each Chainlink oracle node pulls data from multiple data aggregators and takes the median value, ensuring each node’s individual response is resistant to API downtime and doesn’t reflect merely a single source of data. ...
To keep reading, please go to the original article at: