Security Analysis¶
This section provides an overview of security considerations related to BTC-Relay. We refer the reader to this paper (Section 7) for more details.
Security Parameter k¶
Blockchains using Nakamoto consensus as underlying agreement protocol (i.e., leveraging PoW for random leader election in a dynamically changing set of consensus participants) exhibit so called stabilizing consensus. Specifically, finality of transactions included in the blockchain converges with a security parameter k, measured in confirmations (i.e., blocks mined on top of a block containing the observed transaction). That is, the probability of a transaction being reverted in a blockchain reorganization decreases exponentially in k. We refer the reader to this paper for more details on Nakamoto consensus.
In Bitcoin, this security parameter is often set to k = 6, i.e., transactions are considered “final” after 6 blocks have been mined on top. However, there is no mathematical reasoning behind this, nor is there a proof that 6 confirmations are sufficient.
In fact, research has shown that when estimating the necessary confirmations before accepting a transaction, the transaction value itself must also be considered: the higher the value, the more confirmations are necessary to maintain the same level of security. However, recent analysis suggests that it is insufficient to consider the value of a single transaction - instead, to estimate the necessary k one must study the value of the entire block. The existence of bribing attacks, which can even be executed cross-chain, makes the situation worse: in theory, it is impossible to estimate k reliably, as there can always be a large transaction that is being attacked by a reorg in an older block.
What does this mean for BTC-Relay?
BTC-Relay does not specify a recommended value for k. This task lies with the applications which interact with the relay. BTC-Relay itself only mirrors the state of Bitcoin to Harmony, including all forks and failures which may occur.
Liveness Failures¶
The correct operation of BTC-Relay relies on receiving a steady stream of Bitcoin block headers as input. A high delay between block generation in Bitcoin and submission to BTC-Relay yields the system susceptible to attacks: an adversary can attempt to poison the relay by submitting a fork, even if the fork was not submitted to Bitcoin itself (see Relay Poisoning below).
While by design, any user can submit Bitcoin block headers to BTC-Relay, it is recommended to introduce an explicit set of participants for this task. These can be Staked Relayers, which already run Bitcoin full nodes for validation purposes, or Vaults which are used for the creation of Bitcoin-backed assets in the ONEBTC component.
Safety Failures¶
51% Attack on Bitcoin¶
One of the major questions that arises in cross-chain communication is: what to do if one of the interlinked chains fails?
In the case of BTC-Relay, a major chain reorganization in Bitcoin would be accepted, if the new chain exceeds the tracked Chains in BTC-Relay.
If the length of the fork exceeds the security parameter k relied upon by applications using BTC-Relay, this can have sever impacts, beyond that of users loosing BTC.
However, as BTC-Relay acts only as mirror of the Bitcoin blockchain, the only possible mitigation of a 51% attack on Bitcoin halting BTC-Relay via manual intervention of Staked Relayers or the Governance Mechanism. See Failure Handling for more details on BTC-Relay failure modes and recovery procedures.
A major challenge thereby is to ensure the potential financial loss of Staked Relayers and/or participants of the Governance Mechanism exceeds the potential gains from colluding with an adversary on Bitcoin.
Relay Poisoning¶
BTC-Relay poisoning is a more subtle way of interfering with correct operation of the system: an adversary submits a Bitcoin fork to BTC-Relay, but does not broadcast it to the actual Bitcoin network.
If Liveness of BTC-Relay is breached, e.g. Staked Relayers are unavailable, BTC-Relay can be tricked into accepting an alternate Chains than actually maintained in Bitcoin.
However, as long as a single honest participant is online and capable of submitting Bitcoin block headers from the Bitcoin main chain to BTC-Relay within k blocks, poisoning attacks can be mitigated.
Replay Attacks¶
Since BTC-Relay does not store Bitcoin transactions, nor can it be aware of all possible applications using verifyTransactionInclusion, duplicate submission of transaction inclusion proofs cannot be easily detected by BTC-Relay.
As such, it lies in the responsibility of each application interacting with BTC-Relay to introduce necessary replay protection mechanisms (e.g. nonces stored in OP_RETURN outputs of verified transactions) and to check the latter using the Functions: Parser component of BTC-Relay.
Hard and Soft forks¶
Permanent chain splits or hard forks occur where consensus rules are “loosened” or new conflicting rules are introduced. As a result, multiple instances of the same blockchain are created, e.g. as in the case of Bitcoin and Bitcoin Cash.
BTC-Relay by default will follow the old consensus rules, and must be updated accordingly if it is to follow the new version of the system.
Thereby, is it for the Governance Mechanism to determine (i) whether an update will be executed and (ii) if two parallel blockchains result from the hard fork, whether an additional new instance of BTC-Relay is to be deployed (and how).
Note: to differentiate between the two resulting chains after a hard fork, replay protection is necessary for secure operation. While typically accounted for by the developers of the verified blockchain, the absence of replay protection can lead to undesirable behavior. Specifically, payments made on one fork may be accepted as valid on the other as well - and propagated to BTC-Relay. To this end, if a fork lacks replay protection, halting of the relay may be necessary until the matter is resolved.