Crime & Security1 min readThe AssayerLook-back
Q2 2025 in crypto hacks: a code flaw at Cetus led $549.6 million in losses
Losses in the second quarter of 2025 fell to $549.6 million across 34 incidents, from $1.61 billion in the first quarter. That was close to the $541.4 million lost in the second quarter of 2024.
The kind of loss changed. In the first quarter, key compromises and social engineering made up 96.7% of losses. In the second, they made up 16%.
Cetus
The largest loss was a code flaw. Cetus CLMM, a decentralised exchange on the Sui network, lost about $223 million on May 22. A flawed overflow check in one of its functions let attackers use flash loans and narrow price ranges to mint huge amounts of liquidity for tiny deposits, then withdraw real reserves against it.
It was 40.6% of the quarter’s losses and ranks 20th of the 1,272 incidents with a known loss.
A trading bot at Bitget
On April 20, a market-making bot at Bitget on the VOXEL/USDT futures pair fell into a predictable trading loop. Leveraged traders exploited it for about $100 million. Bitget froze accounts and reversed the trades.
Burned on purpose
On June 18, attackers compromised the hot wallet keys of Nobitex, an Iranian exchange, and took roughly $82 million to $90 million. They then sent the funds to addresses no one can spend from, destroying them, and framed it as a political statement.
Other losses included UPCX, a payments platform, which lost roughly $70 million in its own tokens on April 1 after an attacker with a compromised key upgraded its admin contract. Taiwanese exchange BitoPro lost about $11.5 million after attackers tricked a cloud operations employee into installing malware. The attack was attributed to Lazarus Group.
Six incidents of $10 million or more made up 90.7% of losses. DeFi protocols lost $350.8 million across 28 incidents.
Sources
By The Assayer, the Bitminer.io house byline. Written by AI from the sources above and checked automatically. No person reviewed it before publication. How we work.
More on North Korea
