金錢樂高遊戲的潛在風險 : 從 DeFi 攻擊事件簿中可以學到什麼？

Transcript

1. 金錢樂高遊戲的潛在風險 : 從 DeFi 攻擊事件簿中 可以學到什麼？ Charles Jhong August 2, 2020
2. About Me ➔ Hey I’m Charles Jhong ➔ A golang developer ➔ Working at AMIS ◆ Wallet service backend ◆ Bitcoin/Omni integration ➔ Motorcycle enthusiast 2
3. Agenda ➔ DeFi x Security ➔ Type 1 : Re-entrancy Attack ➔ Type 2 : Flash Loan ➔ Type 3 : Front Running ➔ Summary 3
4. DeFi x Security ➔ 近兩年 DeFi 火熱，吸引更多人加入幣圈，也導致了多起駭客攻擊事件， 安全議題是 DeFi 走下去必須克服的挑戰 ➔ 探究漏洞產生原因，才有機會防止再次發生 ➔ 其他類型 Dapp 也有可能面臨同樣的威脅 4
5. Type 1 : Re-entrancy Attack 5
6. What’s Re-entrancy attack? ➔ 智能合約的 race condition 問題 6 Figure source : Sereum: Protecting Existing Smart Contracts Against Re-Entrancy Attacks (Michael Rodler, Wenting Li, Ghassan O. Karame, Lucas Davi) https://arxiv.org/abs/1812.05934
7. Attack Events ➔ The DAO (June 2016, 3.6M ETH, rollbacked and forked) ➔ Uniswap ◆ 4/18 2020 UTC ◆ $ 300K ➔ Lendf.me ◆ 4/19 2020 UTC ◆ $ 25M 7
8. Lendf.me hack ➔ 類似 Compound 的借貸應用 ➔ 駭客利用 imBTC (ERC-777) 和 lendf.me 組合下所形成的漏洞，進行 imBTC 假入金 ➔ 利用洗出來的假餘額，把平台中其他的資產全部借走 8
9. Lendf.me 9 Tx Execution Trace from bloxy.info https://bloxy.info/tx/0xe49304cd3edccf32069dc bbb5df7ac3b8678daad34d0ad1927aa725a896 6d52a
10. Lendf.me Code analysis by Peckshield https://medium.com/@peckshield/uniswap-lendf-me-hacks-root-cause-and-loss-analysis-50f3263dcc09 10
11. Lendf.me 駭客疑似身份被暴露，最後還錢了 11
12. Class of re-entrancy attack ➔ Cross-function Re-Entrancy ➔ Delegated Re-Entrancy ➔ Create-Based Re-Entrancy 12 參考論文 Sereum: Protecting Existing Smart Contracts Against Re-Entrancy Attacks (Michael Rodler, Wenting Li, Ghassan O. Karame, Lucas Davi) https://arxiv.org/abs/1812.05934
13. How to defense? ➔ Reentrancy Guard (OpenZeppelin) ◆ nonReentrant function modifier ➔ Variable mutex ➔ Enhanced EVM ➔ Analysis tools ➔ 確保對接合約內容 ◆ 萬惡 ERC-777 ? 13
14. Type 2 : Flash Loan 14
15. How flash loan works? ➔ 無抵押貸款 ➔ 利息低，額度高 ➔ 借款及還款必須在同一筆 tx 完成，否則 revert() 15 Image source https://hackingdistributed.com/2020/03/11/fla sh-loans/
16. Flash lender and borrower ➔ 提供 liquidator & arbitrager 利用 ➔ 洗交易量、collateral swapping ➔ AAVE, dydx, Uniswap V2 (flash swap), bZx... 16
17. AAVE Source https://github.com/aave/aave-protocol/blob/master/contracts/lendin gpool/LendingPool.sol#L880-L891 17
18. Attack Events ➔ bZx ◆ 2/15 & 2/18 2020 UTC ◆ $350K + $650K ➔ Balancer ◆ 6/28 2020 UTC ◆ $500K 18
19. bZx hack ➔ Flash loan attack 首例 ➔ Capital intensive attack ➔ 短期內連續被攻擊兩次 ◆ 2/15, pump then arbitrage ◆ 2/18, oracle manipulation (price manipulation) 19
20. bZx hack 20 bzx first attack analysis by peckshield https://medium.com/@peckshield/bzx-hack-full-disclosure-with-detailed-profit-analysis-e6b1fa9b18fc
21. bZx hack 21 bzx second attack analysis by peckshield https://medium.com/@peckshield/bzx-hack-ii-full-disclosure-with-detailed-profit-analysis-8126eecc1360
22. Flash attackers will be the new normal ➔ 貸款方風險低，競爭之下會使得借款成本極低 ➔ 降低發動此類型攻擊成本，沒錢的駭客有了方便的資金來源 ➔ 攻擊風險低，萬一失敗了，不怕本金被染色、標籤 節錄自 https://medium.com/dragonfly-research/flash-loans-why-flash-attacks-will-be-the-new-normal-5144e23ac75a 22
23. How to defense? ➔ 邏輯上應避免單一 tx 操縱 ◆ 治理投票 ◆ Oracle price feed (time-weighted average) ◆ 檢查餘額，必須存在一個區塊以上 ◆ DeFi/Dapp 在模擬時，必須考量到大量資金的 corner case，以及相 依性造成的 transitive risk 23
24. Type 3 : Front Running 24
25. What’s Front Running? ➔ 搶跑交易 ➔ 源自於底層 protocol 特性 (廣播交易、15 秒區塊時間) ➔ Transaction order DOES matter ➔ 較難偵測與區分惡意行為 ◆ 偵測需紀錄 mempool 資料，並與鏈上數據交叉比對 ◆ 良性競爭 ? Priority Gas Auction (PGA) 25
26. Class of front running attack ➔ Displacement attack ➔ Insertion attack ➔ Suppression attack 26 參考論文 SoK: Transparent Dishonesty: front-running attacks on Blockchain https://arxiv.org/abs/1902.05164
27. Priority Gas Auction 27 Source http://frontrun.me/
28. Arbitrage market size (10~100 ETH / day) 28 參考論文 Flash Boys 2.0: Frontrunning, Transaction Reordering, and Consensus Instability in Decentralized Exchanges https://arxiv.org/abs/1904.05234
29. Bot profit (10K USD / day) 29 參考論文 Flash Boys 2.0: Frontrunning, Transaction Reordering, and Consensus Instability in Decentralized Exchanges https://arxiv.org/abs/1904.05234
30. MEV ➔ Miner-extractable value ◆ 區塊獎勵以外的潛在收益 ◆ 來源之一 : 透過操縱交易順序，礦工可以得到的報酬 ➔ What if MEV > block reward? ◆ 上層應用危及底層 protocol ◆ 礦工會嘗試 fork • Undercutting attacks • Time-bandit attacks 30 參考論文 Flash Boys 2.0: Frontrunning, Transaction Reordering, and Consensus Instability in Decentralized Exchanges https://arxiv.org/abs/1904.05234
31. How to defense? ➔ Commit-reveal ◆ LibSubmarine ➔ MEV auction ➔ POS 31 參考論文 SoK: Transparent Dishonesty: front-running attacks on Blockchain https://arxiv.org/abs/1902.05164 Source https://libsubmarine.org/
32. Summary 32
33. Summary ➔ 安全的挑戰日新月異，新型應用甚至會影響底層協議安全 ➔ Dapp 間的相依，同時繼承風險 ➔ 無許可制系統，金融應用不再是特許事業 ◆ Dapp Gate Keeper? ◆ 整體生態仰賴社群力量 ➔ 開發環境/工具仍有相當大的需求缺口 33
34. References ➔ https://medium.com/@slowmist/slowmist-details-of-lendf-me-reentrancy-attack-3e168ab5f2b1 ➔ https://www.youtube.com/watch?v=QfFtnW46uwI ➔ https://github.com/OpenZeppelin/openzeppelin-contracts/blob/master/contracts/utils/ReentrancyGuard.sol ➔ https://hackingdistributed.com/2020/03/11/flash-loans/ ➔ https://medium.com/@peckshield/bzx-hack-full-disclosure-with-detailed-profit-analysis-e6b1fa9b18fc ➔ https://medium.com/@peckshield/bzx-hack-ii-full-disclosure-with-detailed-profit-analysis-8126eecc1360 ➔ https://uniswap.org/docs/v2/core-concepts/flash-swaps/ ➔ https://medium.com/dragonfly-research/flash-loans-why-flash-attacks-will-be-the-new-normal-5144e23ac75a ➔ https://medium.com/offchainlabs/mev-auctions-considered-harmful-fa72f61a40ea ➔ https://medium.com/offchainlabs/front-running-as-a-service-334c929c945a ➔ https://medium.com/offchainlabs/meva-what-is-it-good-for-de8a96c0e67c ➔ https://themerkle.com/f2pool-allegedly-prevented-users-from-investing-in-status-ico/ ➔ https://ethresear.ch/t/mev-auction-auctioning-transaction-ordering-rights-as-a-solution-to-miner-extractable-value/6788 ➔ https://libsubmarine.org/ ➔ https://github.com/lorenzb/libsubmarine 34
35. Slides & Handles 35 @shad0356jc @shad0356jc @charles_jhong
36. Thank you 36