Every time there's a major hack we should force an Ethereum reorg to turn off the laundry machines. 🧠
THORChain just inadvertently delayed the laundering of a significant amount of funds stolen from Binance.
The reason? An Ethereum reorg.
One happened this morning and caused THORChain to incorrectly issue a double refund to someone who had deposited 10 ETH.
This triggered an “emergency pause” on THORChain, preventing all users, including DPRK’s money launderers, from using cross-chain swaps.
Until the service comes back online, neither the hackers nor anyone else can swap through a majority of THORChain’s bridges.
this is cool.
有人发现了一种非常离谱的 JavaScript 混淆技术：通过在代码里塞入大量生物武器、毒药和致死剂量相关的函数名，让 AI 因为安全限制而拒绝分析代码。
安全研究员 veritas 在分析浏览器指纹脚本时，发现一段来自 ServicePipe 相关域名的 JavaScript，里面充斥着极其诡异的函数名：
computeRicinLethalDosageForWaterSupply
calculateDeadlyLethalSuicideAspirinDosage
光看名字，仿佛在研究生化武器。但实际上，这段代码只是在收集浏览器指纹，包括 User-Agent、GPU、Canvas 等信息。
研究员怀疑，这些名字是开发者故意设置的陷阱，专门用来阻止 AI 辅助逆向工程。
于是他做了个实验：
• Astra：拒绝分析
• Astra + Daybreak Blue：仍然拒绝
• GLM-5.3：成功分析
这就有意思了。
传统代码混淆是通过复杂控制流、加密字符串、无意义变量名，让人类和工具难以理解代码。
而这种新思路，直接瞄准了大模型的安全机制。
代码本身可能完全无害，但只要在函数名里塞进足够多的危险词汇，部分 AI 就可能主动放弃分析。
更讽刺的是，模型越严格遵守某些安全规则，面对这种混淆时反而可能越容易受到干扰。
当然，目前还没有证据证明开发者一定是故意这样设计，也不能说明这种方法对所有模型都有效。
但它揭示了一个值得关注的趋势：随着越来越多的逆向工程依赖 AI，未来的代码混淆技术可能不再只针对编译器、反编译器和人类研究员，还会专门针对大模型的安全判断机制。
以前是让代码难以理解，现在甚至可以尝试让 AI 不敢理解。
good samaritans unite. if we all buy some $ETH and get it above $2,365 the hacker will be liquidated.
you know what to do
see this post: teapaw.org/mzeller/status/2115940…
unavailable
The 29th of every month should be Kim Jong Sucks Day.
All employees/contractors of your company need to publicly post: Kim Jong sucks.
If they don't, they are NK infiltrators.
Aave and other collateral platforms ought to have a waiting period before you're able to withdraw funds.
If we let users deposit stolen funds as collateral, and immediately withdraw ETH, there isn't much recourse.
Need a window of a few hours before ability to wd is unlocked
these hacks are bullish for Bitcoin.
1) bitcoin wasn't hacked
2) hacker doesn't have anywhere to keep the funds. can't cash out. usdt/usdc get frozen. it's btc or eth.
Less EIPs
This is why Bitcoin is best.
Keep it simple. The less you can do, the less attack surface there is.
ok safe to say binance is pwned
The biggest assets were approximately:
• 1.615B USDT
• 62.6M RLUSD
• 29,934 wBETH
• 11,372 XAUt
• 6.19M UNI
• 198.4M ENA
• 7.175T PEPE
• 27.8M USD1
• 2.26M LINK
• 127,853 AAVE
about $2.5B at the explorer’s prices.
1. The attacker deployed a delegate contract, 0x63e842b0...73e9c6fAe, from the relayer at 15:59 UTC.
2. About four hours later, the relayer submitted an EIP-7702 authorization transaction
That installed the malicious 7702 delegate code on Binance 14.
3. The relayer then made a call to Binance 14, executing the delegate code, and transferring Binance 14’s ETH and token balances.
4. The calldata is labeled sweep(address, address[]). The second argument is a list of recognizable token-contract addresses plus the zero-address (eth).
5. The same relayer later used the same delegate to move 780M USDC from Binance Hot Wallet 34 into the same pool
• The relayer 0x7C... was funded by three Tornado Cash 0.1 ETH withdrawals over the prior two days.
• The same pool also received 28,005.6 ETH from Binance 16 24 seconds after the main sweep.
• The pool then forwarded 195,460 ETH to a fresh, unlabeled EOA
• That EOA has only one transaction and currently holds the 195,460 ETH.
What's a bit unclear is why the relayer was needed in the first place. If they had the private keys to the Binance wallets, they could have just sent the assets out in an atomic bundle without needing a relayer at all.
So it seems likely that they didn't have the private keys and instead just had the 7702 auth signatures, which must have been phished somehow.
I think we can go ahead and assume DPRK.
🧑🏼💻
1/4 EIP-7702 can turn the wallet itself into the sweeper.
A Binance-labeled wallet moved 167,464 ETH + 63 tokens, including 1.615B USDT, just 3m24s after a signed delegation was installed.
This is bad.
etherscan.io/tx/0x53bd38b816…
