Data sources are the origin points for external information, such as APIs from exchanges, weather stations, or IoT devices. Oracles aggregate data from multiple, independent sources to mitigate single points of failure and manipulation. For example, a price feed might pull from Coinbase, Binance, and Kraken. This redundancy is critical for ensuring the accuracy and censorship-resistance of the data supplied to smart contracts.

Oracle consensus refers to the process by which a decentralized oracle network agrees on a single, truthful data point before delivering it on-chain. This often involves nodes independently fetching data and using schemes like proof-of-stake slashing or reputation systems to penalize bad actors. For instance, a network might require a supermajority of nodes to report the same price. This mechanism is essential for preventing incorrect or manipulated data from being accepted by a dApp.

Cryptographic proofs allow users to cryptographically verify that data was delivered correctly from the oracle network to the blockchain. Techniques like TLSNotary proofs or zero-knowledge proofs can attest to the authenticity of API responses. A use case is verifying that a price submitted on-chain matches an off-chain signed attestation from a reputable node. This provides strong security guarantees, enabling trust-minimized applications without relying solely on economic incentives.

Pull oracles are on-demand, where a smart contract requests data only when needed, paying gas for the update. Push orcles proactively publish data at regular intervals, often used for constantly updating feeds like asset prices. For example, a lending protocol uses a push oracle for liquidations, while a prediction market might use a pull oracle to settle a bet. The choice impacts cost, latency, and which party initiates and pays for the transaction.

Node networks are the distributed set of independent operators that perform oracle duties. Decentralization here involves a diverse, permissionless set of nodes with separate infrastructure and data sources. Networks like Chainlink use staking and reputation to align node incentives with honest reporting. This design prevents collusion and ensures liveness, making the oracle service robust against targeted attacks or infrastructure outages that could cripple a centralized provider.

Security models define the economic and cryptographic assurances behind an oracle's operation. These include staking/slashing, where nodes post collateral that can be forfeited for malfeasance, and trusted execution environments (TEEs) that secure node computation. For users, this translates to quantifiable security budgets and attack costs. A robust model is why major DeFi protocols can secure billions in value, as it makes data manipulation prohibitively expensive for adversaries.

The oracle manipulation attack occurs when an adversary corrupts the data feed before it reaches the blockchain. Attackers can exploit centralized data sources or manipulate the price on a low-liquidity exchange to create a profitable arbitrage or liquidation opportunity. This matters because it directly undermines the integrity of price-dependent protocols like lending markets and derivatives.

Stale data refers to price or information that is no longer reflective of the current market state, often due to network congestion or oracle update delays. If a lending protocol uses an outdated price, it may fail to liquidate an undercollateralized position, leading to bad debt. This matters as it introduces systemic risk during periods of high volatility or network stress.

Centralized oracle single point of failure is a critical risk where reliance on a single data source or operator creates a vulnerable target. If that entity is compromised, goes offline, or acts maliciously, all dependent smart contracts receive corrupted data. This matters because it contradicts the decentralized ethos of DeFi and concentrates trust, making protocols susceptible to censorship and manipulation.

Decentralized oracle networks mitigate risk by aggregating data from multiple independent nodes. Security relies on a cryptoeconomic consensus model where nodes stake collateral and are slashed for providing incorrect data. A dispute period allows the community to challenge reported values. This matters as it aligns economic incentives with honesty, creating a more robust and tamper-resistant data layer.

Flash loan-powered oracle manipulation is a sophisticated attack vector where an attacker borrows a large, uncollateralized loan within a single transaction to distort an asset's price on a decentralized exchange. They then use this manipulated price to exploit a protocol's oracle that sources from that DEX. This matters because it demonstrates how DeFi primitives can be weaponized against each other, requiring oracle designs to use time-weighted average prices (TWAPs) or deeper liquidity sources.