Comment about TUSD in fixMediatorBalance

[k1rill-fedoseev]

Tokens that have more than one address, through which they can be called, can be stolen when
they are bridged. An example for such a token is TUSD. The attack would work as follows:

1. The token is already bridged using the first token address. An amount X has been transferred across the bridge.
2. An attacker bridges the token using another token address. The attacker also bridges X tokens. Now the mediator balance on the native side is X for both token addresses. However, the actual balance, when queried from balanceOf is 2*X for both of them.
3. The attacker colludes with the administrators, which trigger a call to fixMediatorBalance on the native side and withdraw X amount of tokens.
4. Then, the attacker can withdraw X tokens, by sending back the bridged tokens.
   Overall, turned X tokens into 2*X tokens, when ignoring fees. The attacker managed to withdraw the full amount of bridged tokens.

[akolotov]

The issue is even more general: how the system should behave if it was discovered during the very first transfer of the token that the OB contract already owns some tokens. Does it make sense to set a flag that the mediator balance is not zero at the moment of the first token transfer as so this flag will require special handling of the token when fixMediatorBalance is going to be used?

[k1rill-fedoseev]

The issue is even more general: how the system should behave if it was discovered during the very first transfer of the token that the OB contract already owns some tokens. Does it make sense to set a flag that the mediator balance is not zero at the moment of the first token transfer as so this flag will require special handling of the token when fixMediatorBalance is going to be used?

1. Do you know any other tokens, except TUSD, that have more than 1 address?
2. What do you mean by "special handling"? Introduce a fixMediatorBalance(address token, address receiver, bool force) function?

[akolotov]

1. Do you know any other tokens, except TUSD, that have more than 1 address?

I haven't known even about the second address of TUSD. But I think that such tokens exist or at least could appear at any moment.

2. What do you mean by "special handling"? Introduce a fixMediatorBalance(address token, address receiver, bool force) function?

Ok, I thought about this a bit more and it seems that there is no simple way to pay additional attention of admins that the discussed situation happens. E.g. this imbalance will occur every time when users use the second token address, not only when the very first transfer happens. So the only way to notify admins about a suspicious actions is to run an offchain code which will analyse every transfer made through OB to identify if there are several addresses serving the same token.

Don't you mind to change the comment to "Before the call of this his method, it must be carefully investigated how imbalance happens in order to avoid attempt to stole the funds from a token with double addresses (e.g. TUSD accessible at both 0x8dd5fbCe2F6a956C3022bA3663759011Dd51e73E and 0x0000000000085d4780B73119b644AE5ecd22b376)"