Summary: There is a non-zero chance of a MyAlgo wallet software compromise leading to the theft of at least $7.2m worth of assets on the Algorand blockchain. We recommend rekeying MyAlgo accounts to fresh private keys, or simply moving funds where possible. This precautionary remediation of further risk should not have usability impact and, if done carefully, may have a significant security benefit.\
We have been day 1 responders to these attacks: reporting the potential compromises, organizing affected users, systematically collecting as much information as possible in order to find a possible common vector, suggesting actions to be taken, preparing transaction trail documents for authorities. This case quickly grew with multiple confirmations and more discoveries daily.
Impact, so far
Of the 13 addresses we identified on the day of the attacks as "suspicious/highly suspicious", 12 have now been confirmed, and a further 5 new addresses have been confirmed so far by impacted users and organizations coming forward.
A total of 17 addresses have been confirmed compromised with at least $7.2MM stolen in $ALGO, $USDC and other assets. A further $1.4MM is suspected compromised in 4 more addresses.
4 more addresses were identified by Rand Labs a few hours ago for a total of 25. We have not had time to look into those yet.
After establishing some potential areas of investigation we created a 48 question form for all affected users to answer, aiming to systematically find commonalities in affected users' environments.
The threads identified in all compromises were 1) use of the MyAlgo wallet on desktop, as well as 2) a recent login (unlocking) into MyAlgo
No other software, dApps, wallets or other kind of vectors could be identified to be shared between every affected user (or even most users).
Some base facts:
- To start from the very basics: public-key cryptography, and specifically Edwards25519 elliptic curves protecting Algorand accounts, are secure and not guessable or breakable.
- The initial fraudulent transactions are single-signature Payment or Asset-Transfer transactions authorized by the proper keys. The base Algorand protocol was not exploited.
- Delegated logic signatures have not been involved.
- Rekeying has not been involved.
- The attacks took place in close temporal proximity, clustering around a time window of less than 6 hours for the initial transactions.
From this point on we present our (D13) interpretations of the situation with two disclaimers:
- The situation is still unfolding; we want to present a checkpoint, and check out.
- Some of our interpretations may not be shared by all involved parties.
Some possibilities we have considered and eliminated:
- Weak key generation and entropy related questions have been examined and ruled out.
- Gard came forward at some point in the investigation as an affected organization. As part of the ecosystem, a compromise of their frontend had to be considered as a common point of compromise of other users. However, this was quickly ruled out based on affected users' reports.
- Recent mac/iOS vulnerabilities have been eliminated as a vector: Multiple users not using Mac or iOS at all. While the timing of Apple's recent Remote Code Execution vulnerability patch was conspicuous, it turned out not to be a factor.
- Browser-wide or operating-system-wide malware is considered unlikely: Metamask was present in many cases and it should have been the first target for theft. Cryptocurrency and NFTs valued in the millions in USD has been left untouched in the same environments where MyAlgo keys were compromised. Metamask and MyAlgo’s storage locations would be technically different (extension storage vs website local storage/indexed DB) and both would be encrypted at rest (while locked) but overall, malware that could access one should also be able to access the other; and effort expended into exploiting Metamask would have multiplied dividends compared to MyAlgo. Metamask stealers or injectors may even be available off-the-shelf in the right corners of the Dark Web.
At this point in the investigation we are considering these two scenarios as leading/most probable:
- Each individual has had their seed phrase compromised through social engineering/phishing.
- A MyAlgo.com compromise leading to targeted exfiltration of unencrypted private keys.
It is always user fault
It is possible that all of these users and organizations were compromised via phishing; however their responses, profiles & history as users and organizations cast reasonable doubt that this is not the only possible cause. Aside from the user reports, the behaviors we saw while interacting with them showed vigilance, awareness of phishing, proactive reporting.
Anecdotally, one affected user cleared his browser storage after he was made aware of the attack; when he tried to log into MyAlgo again at a later date, he was prompted to create a new wallet entirely. Rather than proceeding, he messaged us reporting that he may have been phished after all and asking “does that mean i was *never* connecting to the real MyAlgo Wallet and it was a fake all along?”. Some clarifications later we let them know that it was related to clearing browser data.
Another compromised user is a visible owner of a high-value-target NFT - the variety that is very frequently exposed to all sorts of targeting, scams and phishing.
Another compromised user reported in our affected users' group that one of the investigators reached out to him over email unexpectedly. He reported one of us to all of us, wanting to double check the email contents. (The contents were checked and were not malicious.)
Aside from behaviors and impressions, the questionnaire has a section on seed phrase management with a pointed question about the last time they had restored a seed phrase into any wallet. More than 50% of collected responses were either "Never" or "more than 1 year ago".
...except when it isn't
All in all, it is hard to definitively dismiss this attack exclusively as user error. While user error is a common cause, it isn't always the case:
The aforementioned Solana wallet hack impact, while numbering in thousands of wallets, resulted in fewer total funds lost ($4.1m per article.)
Mass exploitation is not the only way a supply-chain attack can be fulfilled.
"If MyAlgo were compromised, my $100 would surely be missing"
A reasonable rebuttal to "why are more addresses not compromised" would be "because it would be detrimental to the attacker end goal" - assuming the attack is financially motivated.
More can be less: if thousands of accounts were compromised simultaneously, the market would panic, the $ALGO token could crash.
More can be harder: thousands of compromises would mean a lot of more attention would be drawn to this case, making it harder to funnel the stolen funds out of the ecosystem.
The assumption that an attacker would always try to maximize the amount gained in any attack scenario does not always hold - there are tradeoffs involved in everything and real-world crime does not come with a high score board.
It is reasonable to us that an attacker could compromise MyAlgo and "only" choose to steal ~$8m from ~20 accounts. If anything, we are surprised that it wasn't executed in a more drawn out fashion, forcing a conclusion of user error and flying under the radar for longer.
A interesting commonality to note it that certain users reported multiple accounts compromised of medium-large value (say, 250K ALGO to 1M ALGO) all from the same wallet, whereas smaller but-still-large accounts (one case of 80K ALGO, another distinct case of ~$30K worth of LP tokens) were also available on the same wallet. The least minimum amount attacked so far is about 275K ALGO.
The investigation is still ongoing. Law enforcement contact has been made. We have not shared any details or leads that would compromise the investigation.
So while we can not prove that there has been a MyAlgo compromise, we have enough reasonable doubt to strongly recommend that MyAlgo users rekey their MyAlgo accounts using Pera Web or Defly wallets, or if their old addresses are not significant (governance, NFT minting addresses), simply moving to a freshly created wallet on different wallet software.
(EDIT: While this advisory was being drafted, MyAlgo issued a similar recommendation)
The fact that we have not detected any movements from the attackers in a week is not a guarrantee of continued silence and safety for accounts of any size. Not your keys, not your crypto, remember? Since the blockchain doesn't issue refunds, either move or rekey to be safe.
Rekeying is a powerful feature of the Algorand blockchain that is akin to "changing your password". You can create a new address on Pera web or Defly, then import your MyAlgo address, and issue a rekeying transaction from the MyAlgo address that will instruct the Algorand network to disregard the MyAlgo private key/seed phrase and instead delegate authority to the newly created address (and corresponding private key/seed phrase.)
Your potentially-compromised MyAlgo account will now gain the benefit of a brand new private key securing it, without having to change your address, lose your governance commitment, or have to mint your next NFT collection with a different address.
If you plan to rekey, make sure to try rekeying in TestNet first, or with a dummy account on MainNet. Your account rekeying can be tested in TestNet without compromising your MainNet access in case things go wrong. Each network's rekeying status is, as one would expect, completely independent of the others.
Pera iOS is known to have some issues signing at the moment which are being worked on for release as soon as possible.
Pera Web Rekeying instructions: https://support.perawallet.app/en/article/how-to-rekey-an-algorand-account-with-pera-web-wallet-9alza3/?bust=1677517191020
Defly Rekeying instructions: https://docs.defly.app/app/rekey-an-account
And we're out (mostly)
After this report we feel like it is a good time to conclude out leadership part in this and formally hand over the investigation to our friends and co-investigators at Nimble Insurance for continued impacted user support and collaborating with Law Enforcement.
We lack the bandwidth to continue working on this as intensely as we have done thus far. We have every confidence in both Nimble and VantagePoint Blockchain to get the best possible result for impacted users.
We will remain available on an ad-hoc basis to the individuals affected and in a supporting role in the investigation, if needed.