Thread by @moo9000

1/ Is #Solana going down to a hole?

Let's look at the causes of the Wormhole hack and what we can learn from this.

Let's discuss Solana security.

A thread.

👇👇👇

2/ One of Solana's bridges, Wormhole, got hacked for ~120k ETH earlier this week.

coindesk.com/markets/2022/0…

The amount of US dollar value at risk, or lost, at this point, is $200-300M.

3/ A bridge is a blockchain application that bridges value between two blockchains together.

In this case, Wormhole was bridging ETH from Ethereum Mainnet to Solana.

4/ A good overview on bridges is in this @_prestwich's presentation "Building bridges, not wallet gardens"

5/ I believe this sets the hack the second most successful DeFi hack over history. Note that centralised exchange hacks like MtGox, Thodex, BitFinex, BitMart might be bigger.

6/ To establish the root cause, let's look at technical analysis first.

👇👇👇

7/ The hack focuses on Solana smart contract programming concept called Sysvar Instructions

docs.rs/solana-program…

8/ @KudelskiSec has a detailed and friendly run down in their blog. Even software developers that do not possess Solana or #rustlang experience can follow the line of thought

research.kudelskisecurity.com/2022/02/03/qui…

9/ The Solana #rustlang toolkit function load_instructions_at was deprecated October 2021 in this commit:

10/ github.com/solana-labs/so…

11/ The commit replaces the with the safe load_instruction_at_checked() variant. because the issue was already well known by October 2021

Here is also the relevant Solana programming documentation

docs.solana.com/developing/pro…

12/ (Note that when you deprecate something, also rename the old function as unsafe_() per secure defaults rule)

See PyYAML incident

blog.ankursundara.com/pyyaml-cve/

13/ Thus, the root cause of the incident be broken down to

- Designing an unsafe API (Solana "developer experience" mistake)
- Using this unsafe API (Wormhole team)
- Not notifying the Wormhole team about the potential upcoming incident (Solana security team)

14/ Aftermath.

Is Solana insecure?

👇👇👇

A risk registry is a tool that TradFi fintechs use to classify risks and create risk matrix for regulators about their business.

15/ In your business there exist

- Inherited risk, due to the nature of business
- Mitigations
- Residual risk after mitigations

16/ In Solana's case inherited risk comes due to use of novel blockchain technology and Linus Tolvard's law of eyeballs:

"Given enough eyeballs, all bugs are shallow"

(Linus from #Linux fame).

17/ This means that more developers you have, over time, all bugs will be found in an #opensource code base and none are left.

18/ Solana has not had enough time and developers have yet to become mature e.g. Ethereum, which had a headstart since 2014.

There exists more code that needs more eyeballs to read it.

19/ This is the nature of the software: only time can produce mature software, nothing else. Solana devs cannot speed up time. The inherited risk will be there for a few more years at least: Solana is new - it will have issues with a new product.

20/ This leaves us mitigations to bring down the inherited risk.

Mitigations are happening, we can see load_instruction_at() -> load_instruction_at_checked() commit was done in October 2021.

21/ But are mitigations enough and how can they be improved?

I do not know the details of the Solana security researcher scene, I have only familiarity with Bitcoin/Ethereum auditors and security research.

22/ Ethereum has private chat rooms where security researchers can discuss and alert each other with some implied assumed privacy (though this assumption is simply so called gentlemen's agreement.)

23/ Such a group could have alerted Wormhole team of vulnerable code in Nov-Dec 2021 and prevented the issue.

If such a Solana security chat group exists please invite me in.

24/ Solana will definitely learn from this incident.

25/ Aftermath #2.

Shitting on Solana and Jump Trading.

👇👇👇

26/ Jump Trading, a very rich Chicago based HFT firm, owns Certus One, the developer of Wormhole (I did not verify this.)

They gapped the missing 120k ETH from their own balance sheet.

27/ What else could they have done? This is the only move to signal that they are committed and adults who clean after their own mess.

28/ Coinbase did a similar move in 2017 and reimbursed traders in a flash crash.

techcrunch.com/2017/06/24/coi…

This ensured the traders can trust Coinbase (though I disagree giving people back money of losing trades does not set a good precedence.)

29/ Certus One did not do anything especially bad from the engineering standpoint. Solana devs are good as Ethereum devs.

30/ While I am all about shitting on borderline scam projects like Wonderland and their criminal founders, Wormhole or Solana are not such a project.

31/ Risks cannot be totally got rid of. Risks will realise. In the contemporary blockchain business, unlike in 2017 when Ethereum was going through similar issues, stakes are higher.

High stakes mean there will be higher losses, now and then.

32/ Also, because the issue was a smart contract bug, a similar incident can happen to L2s, though the inherited risk is lower. I bet my left kidney at least one L2 will blow up in 2022-2024 in a similar fashion.

33/ FIN

Now up to the mountain hiking trail to clean it up from trash.

Ps. On secure defaults: #Ethereum ERC-20 approve() / transferFrom() should be renamed to

unsafeApprove() and unsafeTransferFrom()

https://twitter.com/moo9000/status/1299647432817606656