DeFi lending just had its worst year yet for a specific kind of theft: the price manipulation exploit. Blockchain intelligence firm TRM Labs counted 32 price-manipulation attacks in 2026 through early September, up from 12 across all of 2025, according to a report first surfaced by Odaily Planet Daily and republished by KuCoin on September 2. That is not a small bump. It is close to a tripling in eight months, and TRM Labs says price manipulation now accounts for roughly one in eight crypto hacks, compared with about one in 17 back in 2022.
The attacks share a pattern. Someone finds a lending protocol that leans on a thin, easily moved price feed, pushes that price far from reality for a few minutes, and either borrows against inflated collateral or drains a vault before the market corrects. Three incidents from the past two weeks show how varied the mechanics can get: a $91,000 Sui exploit that killed an entire protocol, a $170,000 Optimism drain that took 13 minutes start to finish, and a Cronos lending exploit that forced a chain-level halt after roughly $75 million got borrowed against a pumped token. Together they explain why security researchers are treating 2026 as the year price manipulation stopped being a niche risk and became a mainstream one.
DeFi Lending’s Price-Manipulation Surge, By the Numbers
The 32-versus-12 figure comes directly from TRM Labs, a blockchain analytics firm that maintains a running database of crypto hacks going back several years. Odaily Planet Daily reported the number on September 2, 2026, framing it as the highest annual count TRM Labs has recorded for this specific exploit category, and KuCoin picked up the same data point in both a news flash and a longer blog post the same week. Neither outlet published a full dollar-loss total for the category, which matters: this is a story about frequency, not necessarily about a single blockbuster loss.
Independent coverage from Cryptonomist corroborated the same TRM Labs figures days later, describing the trend as a record surge specifically concentrated in DeFi lending rather than spread evenly across all of crypto. That framing lines up with what shattered.io has already tracked this year. Our own reporting on August’s record 50 crypto hacks showed total losses actually falling 49% month over month even as incident counts climbed, and a separate look at stolen keys overtaking code bugs put 2026’s aggregate DeFi hack losses above $1.3 billion. Price manipulation fits into that picture as a high-frequency, often lower-dollar category: individual incidents like Full Sail’s $91,000 loss and Cozy Finance’s $170,000 drain are small next to nine-figure bridge hacks, but there are a lot more of them, and the mechanism keeps recurring across different chains and protocol designs.
KuCoin’s blog post adds one more data point worth sitting with: attackers are increasingly targeting thinly traded collateral rather than blue-chip assets. One example the firm cites involves a token whose price got driven up roughly 100x in 20 minutes, enough to support an estimated $75 million in borrowing against it on a Cronos lending protocol before the position could be liquidated. That detail matches, almost exactly, the mechanics behind the Tectonic exploit that halted the Cronos chain in the same window, which we cover in more detail below.
What Actually Counts as a Price-Manipulation Exploit
Not every DeFi hack belongs in this bucket. TRM Labs and the outlets citing its data describe the category narrowly: incidents where an attacker deliberately distorts a collateral price or oracle feed in order to over-borrow, drain a lending pool, or mis-price a position, as opposed to a straight private-key theft, an access-control bug, or a governance takeover. That distinction matters for readers trying to make sense of headline numbers, because 2026’s overall DeFi loss figures blend several very different attack types.
Three sub-patterns show up repeatedly in this year’s incidents. The first is oracle-key compromise, where an attacker gains the ability to sign price updates directly, as happened to Full Sail. The second is market-based manipulation, where the attacker doesn’t touch any oracle code at all but simply buys up a thinly traded token on a low-liquidity market to move its reported price, which is closer to what happened on Cronos. The third is protocol-logic manipulation, where the exploit lives in how a lending or insurance contract calculates exposure rather than in the price feed itself, a pattern security researchers have pointed to in Cozy Finance’s second Optimism incident. Testing for the middle case specifically is why our oracle manipulation attack testing guide exists as a companion resource for teams auditing their own price feeds.
Inside the Attack: How Oracle and Price Manipulation Works
The mechanics are simpler than the dollar figures suggest. A lending protocol needs to know what collateral is worth in order to decide how much it can safely lend against it. Most protocols pull that number from an oracle, a piece of infrastructure that reports prices on-chain, often by aggregating data from several exchanges or liquidity pools. If an attacker can either corrupt the oracle’s inputs or physically move the market the oracle is watching, they can make an asset appear far more valuable than it is, for just long enough to borrow against it.
A simplified version of the sequence looks like this:
1. Attacker identifies a lending market accepting a thin, low-liquidity token as collateral
2. Attacker buys the token aggressively (or compromises the oracle feed directly)
3. Reported collateral price spikes far above fair value within minutes
4. Attacker deposits the inflated collateral and borrows the maximum allowed
5. Attacker withdraws the borrowed assets before the price corrects or liquidation triggers
6. Price reverts; protocol is left holding under-collateralized debtWhat makes 2026’s wave notable is how little capital some attackers need to pull this off. Full Sail’s attacker didn’t need to buy anything at all, they added a signing key to a live Switchboard oracle feed and simply declared a false price. The Cronos incident took the opposite route, real market activity that happened to be cheap because the target token traded so thinly. Both land in the same bucket because the outcome is identical: a protocol lends against a price that was never real.
Full Sail’s $91,000 Oracle Compromise on Sui
Full Sail, a capital-efficient yield protocol on the Sui blockchain, disclosed on August 29, 2026 that an attacker had exploited production code inside Switchboard, the oracle provider it relied on for price feeds. According to the project’s own account of the incident, the attacker added a signing key they controlled to the oracle’s authorized set, then used that access to push vault prices roughly 100x below market value. From there the attack was mechanical: deposit at the artificially low price, wait for the feed to correct, withdraw far more than had gone in.
The total drained across three affected vaults came to approximately $91,000, a modest sum by 2026’s standards. What made the incident notable wasn’t the size of the loss but the response: Full Sail said the breach did not involve a compromise of its own admin keys, pinning the root cause squarely on Switchboard’s oracle infrastructure, and then announced it would wind down operations on Sui entirely rather than continue running with a third-party dependency it no longer trusted. Switchboard, for its part, reportedly paused operations across Aptos, Sui, IOTA and Movement while it investigated how the signing-key vulnerability made it into production, according to reporting from Cryptobriefing.
Full Sail’s shutdown is a useful data point for the rest of the industry: a $91,000 loss, tiny next to a nine-figure bridge hack, was still large enough to end a protocol’s existence once the trust in its price infrastructure broke. Dollar losses alone understate the damage a manipulation exploit can do to a project’s viability.
Cozy Finance’s Second Optimism Exploit in Months
Cozy Finance, a DeFi risk-coverage protocol running on Optimism, suffered its second exploit of the year on the morning of September 7, 2026. Blockchain security firm Blockaid flagged the incident in near real time, and on-chain data reviewed by Crypto Economy showed the exploit transaction landing at 05:43 UTC, draining roughly 163,326 USDC.e, worth around $170,000 at the time. The attacker moved the funds through a bridge within 13 minutes of the initial drain, well inside the window most protocols need to freeze assets or coordinate a response.
The speed of the exit is the headline detail here. Real-time monitoring tools like Blockaid can now spot an exploit within seconds of it starting, but detection alone doesn’t stop a determined attacker from bridging funds out before anyone with the authority to pause a contract can act. That gap between detection and containment is becoming one of the defining problems in DeFi security, separate from whatever vulnerability let the exploit start in the first place.
Cozy Finance’s repeat exposure also raises a harder question for the sector: a protocol that has already been exploited once should, in theory, have hardened its attack surface. That it didn’t, at least not enough to stop a second incident within the same year, suggests that patching a single vulnerability doesn’t necessarily close off the broader class of exploit that produced it.
The $75 Million Tectonic Exploit That Halted Cronos
The largest and most disruptive incident in this wave targeted Tectonic, a lending protocol on the Cronos chain. Shattered.io’s earlier reporting on the Tectonic exploit and Cronos’s response detailed how the attack forced validators to halt the chain outright, an unusually aggressive intervention that most DeFi ecosystems avoid because it undercuts the premise of decentralization. KuCoin’s independent account of the broader price-manipulation trend describes what appears to be the same incident without naming the protocol: an attacker drove a thinly traded token’s price up roughly 100x in 20 minutes and borrowed an estimated $75 million from a Cronos lending protocol against the inflated collateral.
The two accounts line up closely enough on chain, dollar figure, and mechanism that they almost certainly describe the same event, which makes Tectonic the clearest large-scale example of 2026’s price-manipulation pattern. It is also the incident most directly responsible for the industry’s use of chain halts as an emergency brake. Halting an entire network to contain a single protocol’s exploit is a blunt instrument, and it invites obvious criticism about centralization, but Cronos’s willingness to do it shows how seriously ecosystem operators are now taking manipulation-driven insolvency risk.
2026 Price-Manipulation Incidents at a Glance
The table below lines up this year’s clearest price- and oracle-manipulation cases, drawing on TRM Labs’ aggregate count, KuCoin’s incident descriptions, and shattered.io’s own reporting on individual hacks.
| Date (2026) | Protocol | Chain | Loss / Exposure | Attack Vector |
|---|---|---|---|---|
| Aug 29 | Full Sail | Sui | ~$91,000 | Malicious signing key added to Switchboard oracle |
| Early Sept | Tectonic | Cronos | ~$75 million borrowed | Thin-liquidity token pumped ~100x in 20 minutes |
| Sept 7 | Cozy Finance | Optimism | ~$170,000 | Protocol-logic exploit, funds bridged out in 13 minutes |
| 2026 (per shattered.io) | Moonwell | Base | $8.7 million | Price oracle manipulation |
| 2026 (per shattered.io) | More Markets | Flow EVM | ~$410,000 | E-mode collateral flaw |
| Aug 2026 (per shattered.io) | Term Finance | — | $8.5 million | Governance-linked hack |
Why Illiquid Collateral Keeps Getting Exploited
Nearly every incident in the table above shares one ingredient: a collateral asset that doesn’t trade enough volume to resist a determined buyer. A blue-chip asset like ETH or BTC is hard to manipulate because moving its price meaningfully requires capital most attackers don’t have. A token trading a few thousand dollars a day on a single venue is a different story. Someone with a modest amount of capital can push its reported price up by a wide multiple in minutes, hold it there just long enough to borrow, and exit before the market or the protocol’s risk parameters catch up.
Lending protocols keep accepting these assets because listing more collateral types is how they grow. A protocol that only accepts ETH, BTC and major stablecoins is safer but smaller. One that accepts long-tail governance tokens, liquid-staking derivatives, or newly launched assets can attract more deposits and more borrowing volume, but every new listing is also a new attack surface. KuCoin’s write-up on the trend frames this directly: risk teams are being asked to move faster on listings than their oracle infrastructure can safely support, and the gap between the two is where 2026’s exploits keep happening.
The tokenization trend compounds the problem rather than solving it. As protocols experiment with real-world assets and other tokenized collateral, they’re often onboarding instruments with even less on-chain price history and even thinner secondary markets than the tokens already causing trouble, which suggests this category of exploit has room to keep growing before it shrinks.
Flash Loans: The Accelerant, Not the Root Cause
Flash loans, which let a borrower take out and repay a loan within a single transaction with no upfront collateral, show up in a large share of price-manipulation exploits, but they aren’t the underlying vulnerability. They’re a tool that makes the underlying vulnerability, an unreliable price feed, cheaper and faster to exploit. Without a flash loan, an attacker needs real capital to move a market. With one, they can borrow the capital, move the price, execute the exploit, and repay the loan, all in a single atomic transaction that either fully succeeds or fully reverts.
That distinction matters for defenders. Banning flash loans outright, which a few protocols have tried, doesn’t fix a bad oracle design, it just raises the capital requirement for an attacker who has other ways to source funds. The more durable fix is on the oracle side: time-weighted average pricing, multiple independent data sources, and circuit breakers that reject price updates outside a sane range within a short window. Teams that want to test whether their own contracts are exposed to this pattern can work through our flash loan attack testing guide, which walks through reproducing these attacks in a Foundry test environment before an attacker does it for real.
Price Manipulation vs Other DeFi Exploit Categories
Price manipulation is climbing fast, but it’s still not the biggest line item in 2026’s DeFi loss ledger by dollar value. Stolen private keys and access-control failures continue to produce the largest single losses, even though they happen less often. The table below places price manipulation alongside the other major categories shattered.io has tracked this year.
| Exploit Category | 2026 Signal | Trend vs 2025 | Representative Case |
|---|---|---|---|
| Price / oracle manipulation | 32 incidents recorded by TRM Labs | Up from 12 (+167%) | Full Sail, Tectonic, Cozy Finance |
| Stolen private keys | Leading category by dollar losses, part of $1.3B+ tracked total | Overtook code bugs as top loss driver | Covered in our stolen-keys report |
| Flash-loan-accelerated exploits | Present in a large share of manipulation cases | Growing alongside price manipulation | Used as an accelerant, not a standalone root cause |
| All DeFi exploits, Q2 2026 | 99 incidents, $746 million lost | Record quarter | Aggregate figure across all categories |
Read together, the two tables tell a consistent story: 2026 hasn’t seen one dominant attack type take over DeFi security. Instead, several categories are climbing in parallel, with stolen keys still leading on raw dollar losses while price manipulation leads on sheer incident count. Our earlier look at the record 99 hacks logged in Q2 2026 put total quarterly losses at $746 million across every category combined, a useful benchmark for how much of that figure any single exploit type, including this one, is likely responsible for.
Market Impact: What This Means for Lenders and TVL
For depositors, the practical risk isn’t necessarily losing funds outright, it’s a lending pool becoming under-collateralized after an attacker borrows against a manipulated price and disappears. When that happens, the protocol is left holding bad debt, and depositors either absorb the loss through a socialized shortfall or wait for a treasury bailout that may or may not materialize. Full Sail chose to shut down rather than manage that outcome. Other protocols have tried partial reimbursements funded by their own treasuries, an approach that works only as long as the treasury is large enough to cover the hole.
The broader market effect is a slow reallocation of deposits toward protocols with the most conservative collateral listings and the most heavily audited oracle stacks. Aave and Morpho, the two largest lending protocols by total value locked, have both built reputations around cautious asset onboarding and multi-source price feeds, and neither has appeared in TRM Labs’ 2026 incident count so far. That’s not proof of invulnerability, but it does suggest that scale and caution correlate with resilience in this specific category of attack, even if it comes at the cost of listing fewer speculative assets.
There’s a second-order effect worth watching too: insurance markets and risk-coverage protocols, the same category Cozy Finance operates in, are themselves becoming exploit targets. If the protocols meant to backstop DeFi risk are vulnerable to the same manipulation patterns as the lending markets they insure, the safety net has a hole in it that most depositors don’t know to check for.
Historical Context: From Mango Markets to Today’s Oracle Wars
Price manipulation isn’t a new idea in DeFi, it’s one of the oldest attack classes in the space. The 2022 Mango Markets exploit, in which an attacker used a similar collateral-pumping technique to borrow roughly $114 million from the Solana-based protocol before negotiating a partial return, is still the reference case most security researchers cite when explaining this attack family to newcomers. What’s changed by 2026 isn’t the core technique, it’s the frequency and the diversity of targets. Where Mango Markets was a singular, headline-grabbing event, this year’s incidents span multiple chains, oracle providers, and protocol types, suggesting the technique has become accessible to a much wider pool of attackers rather than a rare feat requiring deep expertise.
That shift mirrors what happened with flash loan attacks a few years earlier: an exotic technique used by a handful of sophisticated actors eventually became a commodity tool available to almost anyone willing to fork open-source exploit code. Security researchers increasingly expect price manipulation to follow the same trajectory, moving from a specialist attack into a routine one that smaller, less resourced protocols need to defend against by default rather than as an afterthought.
Can Aave and Morpho Avoid the Same Fate?
Size and caution have kept the biggest lending protocols out of this year’s incident count so far, but neither is architecturally immune. Aave’s risk framework relies on a governance-driven approval process for new collateral types and uses Chainlink’s decentralized oracle network for pricing, which spreads the attack surface across many independent data providers rather than a single feed. Morpho takes a related but distinct approach, isolating individual lending markets from one another so that a manipulated price in one market can’t cascade into the rest of the protocol’s liquidity.
Both designs raise the cost of an attack without eliminating it. A well-funded attacker could, in theory, still manipulate a thinly traded asset that either protocol has listed, and both have listed long-tail assets at various points to stay competitive on yield and borrowing options. The honest assessment from most security researchers is that Aave and Morpho are harder targets, not impossible ones, and that their absence from 2026’s incident list so far may reflect attacker preference for softer targets as much as any structural advantage.
Industry Response and Five Predictions for What’s Next
The immediate industry response to this wave has been operational rather than regulatory. Switchboard paused its oracle service across four chains after the Full Sail incident. Full Sail chose to shut down entirely. Blockaid and similar monitoring firms are catching exploits within seconds, and independent trackers like SlowMist’s real-time hack feed now log new DeFi incidents on a near-daily basis, even if that speed hasn’t yet translated into stopping funds from leaving before a bridge transaction clears. No dedicated regulatory action targeting price-manipulation exploits specifically has emerged as of this writing, though the pattern is likely feeding into broader DeFi risk-disclosure conversations already underway in the US and EU.
- Expect more protocols to follow Full Sail’s example and shut down voluntarily after a manipulation exploit rather than attempt a costly relaunch, particularly smaller projects without a large treasury cushion.
- Oracle providers will face pressure to publish security audits and incident-response track records the way smart contract auditors already do, turning oracle selection into a more visible due-diligence step for protocols.
- Real-time monitoring firms like Blockaid will keep shrinking detection time, but the harder problem, freezing or clawing back funds before a bridge transaction settles, will remain largely unsolved through 2026.
- Collateral listing standards will tighten industry-wide, with more protocols requiring minimum trading volume or liquidity depth thresholds before accepting a new asset type.
- TRM Labs’ 2026 count of 32 is unlikely to be the ceiling; based on the trajectory from 12 to 32 in under two years, next year’s tally plausibly clears 40 unless oracle design standards shift meaningfully across the sector.
None of this suggests DeFi lending is heading toward collapse. It suggests a maturing but still uneven security landscape, where the largest protocols have learned enough to avoid the worst outcomes while a long tail of smaller, faster-moving projects keeps making the same mistakes with new tokens on new chains. Teams building or auditing lending protocols can find a hands-on walkthrough of testing for this exact exploit family in our oracle manipulation attack testing guide, and readers who want the broader context on how crypto security has evolved this year can start from our cryptocurrency coverage hub.
Frequently Asked Questions
What is a DeFi price-manipulation exploit?
It’s an attack where someone deliberately distorts a collateral price or oracle feed, either by compromising the oracle directly or by moving a thinly traded market, in order to borrow more than they should be able to or drain a lending pool before the price corrects.
How many price-manipulation attacks happened in 2026?
TRM Labs recorded 32 such attacks in 2026 through early September, up from 12 across all of 2025, according to reporting picked up by KuCoin and Odaily Planet Daily.
What was the biggest price-manipulation exploit of 2026 so far?
The Tectonic exploit on Cronos is the largest documented case, with an attacker reportedly borrowing around $75 million against a token pumped roughly 100x in 20 minutes, an incident severe enough that Cronos halted the chain to contain it.
Are flash loans the cause of these attacks?
No. Flash loans make manipulation cheaper and faster to execute by removing the need for upfront capital, but the underlying vulnerability is almost always a price feed that can be moved or corrupted, not the flash loan mechanism itself.
Why do smaller protocols get hit more often than Aave or Morpho?
Larger protocols tend to use multi-source oracle networks and stricter collateral listing standards, which raises the cost of manipulating a price. Smaller protocols often list thinly traded assets to compete on yield, which gives attackers a cheaper target.
Is price manipulation the biggest source of DeFi losses in 2026?
Not by dollar value. Stolen private keys and access-control failures still produce larger aggregate losses. Price manipulation leads on incident frequency, now accounting for roughly one in eight crypto hacks.
Can users protect their deposits from this type of exploit?
Not directly, since the vulnerability sits in the protocol’s oracle and collateral design rather than in user behavior. The most practical step is checking whether a protocol lists thinly traded collateral or relies on a single oracle source before depositing significant funds.
What is TRM Labs and why does its data matter here?
TRM Labs is a blockchain intelligence firm that tracks crypto hacks and illicit activity across chains. Its incident counts, including the 32-versus-12 figure cited throughout 2026 coverage, are widely used by exchanges and researchers as a baseline for tracking exploit trends. More on its methodology is available through TRM Labs’ own resources.
