On July 19, 2024, a single transaction on Solana extracted $1.65 million in stablecoins from Allbridge’s Core bridge. The attacker borrowed $1.12 million via Kamino’s flash loan, manipulated the pool’s instantaneous price, swapped borrowed funds for profit, and vanished to Ethereum. The bridge was immediately paused. This is not a novel exploit. It is a textbook price oracle attack—one that should have been mitigated before mainnet deployment. The system failed because its pricing model assumed trust in a single-block snapshot, not a trust-minimized feed.
Allbridge is a cross-chain bridge aggregator that permits liquidity providers to deposit stablecoins into pools on various chains, including Solana. The Core bridge uses an Automated Market Maker (AMM) model on each chain to price assets—meaning the exchange rate is determined by the pool’s current balance. No time-weighted average price (TWAP) oracle. No slippage hard cap. The design relied on the assumption that no single actor could drastically shift the pool’s ratio. That assumption was wrong. Kamino’s flash loan provided the leverage; Allbridge’s naive pricing provided the vulnerability.
The exploit’s mechanics are clean. The attacker took a $1.12M flash loan from Kamino—a Solana lending protocol that, like all DeFi lending platforms, allows uncollateralized debt within one atomic transaction. They swapped that borrowed amount into the Allbridge Solana stablecoin pool, moving the pool’s internal price of USDC against USDT. Then, using the inflated USDT/USDC ratio, they swapped again to extract a profit in the other asset. The entire manipulation happened in seconds, within a single block. After the attack, they bridged the proceeds to Ethereum, disappearing from Solana. The bridge’s pause function kicked in, but only after the damage.
The core failure is algorithmic: the pool used the constant product formula (x*y=k) without any price guardrails. Flash loans are a standard DeFi primitive; they are not the problem. The problem is that Allbridge treated its pool as a self-contained pricing oracle, ignoring the fact that any large, instantaneous trade can distort the ratio. In my post-mortem audits of similar bridges, I have flagged this exact pattern—single-asset liquidity pools without TWAP or Chainlink feeds are basically honeypots. The fix is not complex: use a time-weighted price or a Chainlink oracle that aggregates external market data, and enforce a per-block trade volume limit. Allbridge did neither.
Here is where the contrarian angle cuts. Some bulls will argue that Kamino is complicit for providing the flash loan—that the lending protocol should have prevented the loan or charged a premium. That argument is wrong. Flash loans are a permissionless DeFi utility; they enable arbitrage, liquidation, and yes, exploits. The fault lies entirely with the bridge’s risk model. Blaming the lender is like blaming a bank for providing the cash that a robber uses in a mugging. Kamino serves its function. The exploit is a failure of the Allbridge team’s engineering judgment, not a failure of the flash loan mechanism.
Yet the attack reveals a deeper systemic risk: cross-chain bridges remain the weakest link in DeFi security. Even if Allbridge patches this specific hole, the fundamental design—a single chain pool with its own pricing—is fragile. The whole industry pretends that a quick audit patch restores trust, but trust is not restored by a code fix. Trust-minimized architecture requires immutable guardrails, not human intervention after an exploit. This incident should accelerate the adoption of zero-knowledge bridges that verify state proofs rather than pool balances.
Takeaway: Allbridge’s vulnerability was predictable, and the team’s oversight is now a public ledger entry. The $1.65M loss is a tuition fee for the market: price oracles on single-asset pools are bugs waiting to be hacked. Code speaks. The real question is whether the Allbridge team will deploy a trust-minimized fix or retreat into opacity. The wallet knows the truth.

