Smart contracts have only ever been as useful as the data that they operate on. A lending protocol without an understanding of the value of its collateral can’t liquidate its bad debts. A derivatives platform that can’t observe the value of the market won’t settle its positions. Insurance on top of real-world events has no basis without observed real-world events. Blockchains are closed-off deterministic machines; they can only process things that happen on-chain. The only component of most useful decentralized financial applications that bridges this world’s on-chain vs. real-world gap is the oracle network.
What the oracle problem actually means
The blockchain fulfills its fundamental purpose: trustless verification via determinism. For a transaction, any two nodes that run it must end up with identical results given identical inputs. That fails as soon as an external datum comes into play. The nodes fetching the price feed even a few milliseconds apart may get slightly different numbers. When nodes disagree on their inputs, they cannot possibly agree on their outputs.
This is the oracle problem exactly. It is not just about “How do you bring data onto chain?” but rather, how do you bring real-world data from its inherently centralized and mutable source into a decentralized system without re-introducing the centralized trust the system was built to overcome? The act of having a single data provider supply that answer to the question effectively means one entity, one point of failure. This is the exact problem the blockchain was designed to solve.
The solution Oracle provides to solve this is to act as the middleman between two worlds. They call external information, verify and authenticate it, and then convert it to be able to be read on-chain by the smart contract. The architecture of this solution consists of 2 components: off-chain nodes, which retrieve the information from the real world, and on-chain smart contracts, which receive this information and forward it to the concerned applications.
How oracle networks work
A smart contract requires a value that is not available on chain. It will emit an event to request this value. Oracle nodes on the network then fetch it from exchanges, APIs, data aggregators, etc. And relay it back to the network. Several independent oracle nodes fetching the same value from various sources are averaged with a weighted median (or a similar consensus) and the final value is put on-chain as a single on-chain value. It is not possible for any one source to be tampered with when dozens of independent nodes reach an entirely different consensus.
There are two main delivery models of price feeds: push-based and pull-based oracles. The push-based oracle model has the price feed updated based on either an expected schedule or a price deviation of one percent; it’s the one that most lenders choose due to constant data feeds. Pull-based oracle model makes an offer on a price feed only when it is called, which reduces the latency and cost and thus is the preferred option of the high-frequency trading environment.
The attack surface that never goes away
It’s clear that oracle infrastructure is an unsolved problem. The second most costly attack vector in DeFi in 2024 was price oracle manipulation, which led to $52M in losses via 37 incidents. Attack mechanics are simple. Attackers use flash loans to dramatically increase/decrease the price of an asset within the same transaction, borrow against inflated collateral, and then cash out.
DeFi protocol hacks utilizing oracles resulted in the loss of $403.2M over 41 separate oracle manipulation attacks in 2022. Oracle networks, on the whole, were not the primary factor driving the majority of the losses; rather, poorly configured oracles, typically utilizing one on-chain source, such as AMM spot prices rather than diversified feeds from oracle networks, were targeted more frequently.
It is widely known how the defenses are successful. TWAP averages prices over time, averaging out block-level manipulation and making it uneconomical to maintain a price distortion for a sufficiently long period of time. Circuit breakers, which halt protocol operation when the reported price moves more than a threshold amount from the last recorded price, add an extra layer. One can also consider using Chainlink’s 900+ nodes to decentralize oracle responsibility, requiring a vast coordination problem to manipulate each independent operator to manipulate various data feeds.
An example of this, during June 2024, was the UwU Lend exploit that led to a loss of around $19.3M; due to the lack of these controls, an attacker used on-chain price feeds the protocol fed on, took out massive loans backed by this inflated collateral and drained the lending pools before prices could readjust.
Why oracle networks matter beyond DeFi
Price feeds in DeFi are just one of the use cases for oracle infrastructure. Protocols paying out claims based on documented weather outcomes, supply chain contracts based on verified delivery, and prediction markets settled on real-world results and these all require oracles.
The next evolution of real-world asset tokenization is here. The RWA market is expected to grow between $10 trillion and $16 trillion by 2030 and every tokenized bond, fund, or real asset needs constant, verified pricing and compliance information delivered on-chain. Oracle networks are the only ones capable of doing this.