Glossary
A blockchain oracle is the bridge that feeds outside data - prices, weather, sports results - into smart contracts. Chains cannot read the outside world on their own, so the oracle is the trust point every DeFi price hangs on.
A smart contract can only read its own chain: it is deterministic code with no window on the outside world. Anything external - a token's price, a weather reading, a sports result - must be delivered to it, and the thing that delivers it is an oracle. A centralised oracle is one source, and one source is a single point of failure. Decentralised oracle networks aggregate many independent reporters, and the consensus answer goes on-chain.
The DeFi layer leans on oracles completely. A lending protocol computing collateral ratios reads an oracle's price feed; a perp platform marking positions reads one; a parametric insurance contract paying on weather reads one. If the feed reports a wrong price, the protocol follows it, and liquidations follow the protocol - the oracle is the trust point the whole stack hangs on. Oracles move information; the blockchain-bridge entry moves assets - both are trust points with different exploit profiles. The difference matters when you are deciding what can drain your funds.
Walk a price feed end to end. A lending protocol needs the ETH price in AUD terms - call it $3,790 - to compute whether loans are collateralised. A Chainlink-style network has dozens of independent node operators, each pulling prices from multiple exchanges and submitting signed answers. The network aggregates them - commonly a median of the middle cluster - and writes the aggregated price to an on-chain reference contract. Your lending protocol reads that contract, and the collateral ratio computes from it.
Feeds update on thresholds, not clocks alone. A typical feed refreshes when the price moves a set percentage - say 0.5 percent - or after a set time, whichever comes first. That keeps the on-chain price fresh enough for collateral math without writing to the chain every second, which would be prohibitively expensive in gas. The threshold design matters in fast markets: a price moving 10 percent in minutes can outrun a 0.5 percent threshold, and the on-chain mark lags the market. The lag is where the exploit window lives.
The failure mode is the thin market. A venue's book can flash-crash - a price printing 30 percent low for two blocks - and a feed pulling from that venue relays the bad print. Robust networks weight many venues and drop outliers; single-source feeds relay faithfully. Protocols on thin feeds have liquidated collateral at manipulated prices, and the liquidations do not unwind when the price recovers. The liquidation-cascade entry covers the chain reaction, and the smart-contract entry covers the code that executes it.
Australian readers use DeFi protocols that lean on oracle prices whether they know it or not - every collateral ratio, every perp mark, every automated market runs on a feed. The oracle is the trust point: the code is deterministic, the feed is not. A protocol's audit covers its contracts; the feed is a separate question, and the deposit screen never asks it. Checking which feeds a protocol uses is the due diligence nobody mentions.
A colleague named Brent (yes, that Brent) had collateral liquidated during a flash crash because a feed marked the price 30 percent low for two blocks; the protocol followed its oracle, and the oracle followed a thin book. The price recovered in minutes; the liquidation did not unwind at all. The highest-leverage thing most Australians can do is check whether a protocol's feeds are multi-source with deviation thresholds before depositing, and treat thin single-source feeds as the exploit risk they are. The decentralised-finance entry covers the stack the feeds power. Oracles are the plumbing; plumbing is where the floods start.
A lending protocol reads an ETH/USD feed: dozens of operators pull from multiple exchanges, submit signed answers, and the network writes the median to a reference contract. The protocol computes collateral ratios from that mark - at $3,790 AUD per ETH, a loan against 2 ETH is collateralised to roughly $7,580 AUD. The feed refreshes on a 0.5 percent deviation threshold or a time window, whichever comes first. The code is deterministic; the feed is the part that can surprise, and the surprise is where the risk concentrates.
A venue's thin book prints a price 30 percent low for two blocks, and a feed pulling from that venue relays it. A lending protocol following the feed marks its loans against the bad price and liquidates collateral within the block. The price recovers in minutes; the liquidations stand, because the protocol executed exactly what its oracle reported. Multi-source feeds with outlier rejection are the fix, and they cost the protocol nothing but diligence - the liquidation-cascade entry covers what the forced selling becomes.
A flight-delay insurance contract pays out when an oracle confirms a delay of over two hours from an aviation data source. There is no claims form and no assessor - the contract executes on the data the oracle delivers, which is the entire point of the design. The trust question moves from the insurer to the data source: a wrong reading pays the wrong outcome. The oracle is the contract's eyes, and eyes can be fed - which is why the source list matters more than the promise.
Mistake: Assuming oracle prices are exchange prices.
Why it bites you: They are aggregated, threshold-updated estimates, and the mark can differ from the screen price exactly when markets are stressed. The lag is a design feature with an exploit window.
Mistake: Depositing into protocols on thin single-source feeds.
Why it bites you: A thin book's bad print relays faithfully through one source, and collateral liquidates at manipulated prices. Multi-source feeds with outlier rejection are the robust version.
Mistake: Confusing an oracle with a bridge.
Why it bites you: Oracles move information and bridges move assets - both are trust points with different exploit profiles. Deciding what can drain your funds starts with knowing which one you are using.
Mistake: Ignoring oracle risk in the liquidation arithmetic.
Why it bites you: A manipulated mark liquidates sound collateral, and the loss does not unwind when the price recovers. The feed is a separate question from the contract audit.
I spent a weekend reading one oracle network's whitepaper and two of its feed specifications so you do not have to, and the design is more honest than I expected: many reporters, signed answers, medians on-chain. The honesty ends where the aggregation choices begin. Weighting is a judgement, and judgements have opinions.
The oracle is the part of the stack nobody audits and everybody trusts, which is the combination I like least in finance. I ran the flash-crash scenario through a dozen feed configurations so you do not have to, and multi-source with outlier rejection survived all of them; the single-source versions did not. Brent trusts the audit badge and skips the feed question. The contract is the engine; the oracle is the fuel line, and fuel lines are where the fires start.
It is the system that delivers external data - prices, weather, results - to smart contracts, which cannot read the outside world on their own. Decentralised networks aggregate many independent reporters before posting the answer on-chain.
Because they are deterministic code that can only read their own chain, and anything external must be delivered to them. The oracle is the delivery mechanism, and the trust point that comes with it.
A decentralised oracle network whose node operators aggregate prices and other data from many sources and write consensus answers on-chain. It is the most-used example, not the only design - the questions to ask are the same of any feed.
Thin feeds can - a venue's flash crash relays through a single source, and protocols following it liquidate at manipulated prices. Multi-source aggregation with outlier rejection is the standard defence.
No - the protocol executed exactly what its oracle reported, and the liquidations stand even when the price recovers minutes later. Checking feed quality before depositing is the only prevention.
Disclaimer: This information is general in nature and does not constitute financial or legal advice. Always consult a qualified professional for your specific situation.

Financial Chaos Analyst
Ivy Sinclair-Wren is a Financial Chaos Analyst covering crypto markets, investing behaviour, and the Australian habit of checking prices during family dinners. Based in Melbourne, she reviews AUD Crypto Tracker's coin guides for accuracy, ATO alignment, and ruthless utility — no hype, no predictions, just what the data says.