As Hash and Time Lock Contracts become more popular, they begin to be seen by the end user. For instance, wallets are now being built with this functionality. And that exposes the user to some unscrupulous actors that could cause financial harm.
This is intended to be a document aimed at software developers. What an HTLC is and how an atomic swap is conducted will not be discussed. But interested end-users may also gain some knowledge within.
Hashlock attacks
A message digest looks very random, and it is. An important item to remember is that it is a one-way hash. The hash itself tells you nothing about the original preimage (a.k.a seed). We do not know what it was, how large it was, or when the hash was generated, or even if it was generated.
Only if we have the preimage and algorithm can we verify that the hash was generated by that preimage. In other words: if we apply a message digest algorithm to a given preimage, it will always generate the same output.
Note in the statement above that there are no guarantees about a different preimage not generating the same output as well. Therefore, using a strong algorithm is important. At the time of this writing, SHA256 is the most common, and is considered secure.
Most hashing algorithms have a fixed size output. For instance, SHA256 generates 256 bits (32 bytes) of output. No matter the input, the output will be at least and at most 32 bytes.
An Oversized Preimage Attack
There are limits to the size of a preimage that can be used. That limit is set by the blockchain itself. But when dealing with atomic swaps across blockchains, an attacker could use a preimage that is too large for one side of the 2 sided transaction. That means that an HTLC can be created on both chains, but only redeemed on one chain, and not the other.
This can be mitigated by specifying the preimage size as part of the contract. This feature is available on many bitcoin-based chains among others, but is not standard. Should the preimage size be used with an atomic swap, both sides of the transaction should include the preimage size in their contract.
Timelock Attacks
It is standard practice for HTLCs that the timelock should be long enough so that the block that contains the contract can be considered irreversible. This makes it possible for the redeemer to expose the preimage and still be guaranteed that the contract itself will not be reversed.
This becomes even more important with atomic swaps. The shorter duration (a.k.a “inner”) contract should allow time to achieve its own irreversibility. And the longer duration (a.k.a. “outer”) contract must allow time for irreversibility of both.
In addition, with an atomic swap both sides must consider the redemption time necessary. The creator of the inner contract must decide that should the redeemer redeem at the last moment, is there enough time to redeem the outer contract before the timelock expires.
One final consideration of the timelock portion of the contract is the use of capital. Should the other party not accept, the funds in the contract are held until the timelock expires. Long expiration times could result in missed opportunity costs.
The Technical Hurdle for Wallets and Users
Hash and Time Lock Contracts are implemented on many chains. Following the general guidelines of BIP199 is not the hard part. Providing an interface where an end-user is as protected as possible will require effort. With the current wallet user base, such protections for HTLC creation, verification, and execution will be the responsibility of the wallet developer.
Other notes on HTLCs
It is generally recommended that the “outer” contract be on the slower chain. This can minimize the length of the timelock, should there be the desire to use the shortest possible time.
While it may be obvious, the “inner” contract must expire before the “outer” contract.
While the Bitcoin blockchain provides the functionality for specifying the preimage length, BIP199 does not include that logic in its proposal. It is my opinion that if both chains support it, it should be used. Most chains that derive from Bitcoin should support it. BitShares also supports it.