Recently, while monitoring the market, my coffee has gone cold before I had a chance to drink it. Watching the same token in the same market at the same minute, one window shows no movement at all, while the other window suddenly plunges in price—it's really a dizzying experience. Which data is the true one that should be recorded on the blockchain?

You should know that once bad data is fed into a smart contract, the consequences are no small matter: forced liquidations, instant reversals of trading directions, liquidity pools being drained... All of this is under the pressure of oracles.

Oracles are essentially bridges—transferring real-world off-chain data, such as prices, into on-chain contracts safely and reliably. So when data sources conflict, the key is not about who is right or wrong, but about how to protect on-chain security first before figuring out the truth. There’s no room for ambiguity here.

**First Line of Defense: Using Median Instead of Average**

Don’t trust the mean; it’s too easily skewed by an outlier. The smarter approach is to use the median. Arrange all numbers in order and pick the middle one. Even if one data source is wildly off, it won’t affect the overall picture. The system simply ignores it and continues to operate steadily.

**Second Tactic: Consensus Among a Minimum Number of Data Sources**

There must be enough data sources reaching agreement within a reasonable range before the value is adopted. If the number isn’t sufficient, don’t update yet. Simple and straightforward, but highly effective.

**Third Tactic: Deviation Check Mechanism**

Compare new data with the last reliable value; if the jump is too extreme, trigger an alert. Multiple layers of defense are necessary to truly safeguard on-chain security.