,
11 tweets,
2 min read
Read on Twitter
1/n An interesting game theoretic issue in @iotatoken that I haven't seen discussed much before, which shows the complexities of orchestrating incentives of different parties.
2/n Honest parties in IOTA are expected to perform two random walks on the transactions DAG to find two tips and make their transaction approve those two tips. The random walk favors heavier subdags.
2/n Thus, to have more transactions approve my transactions I'm incentivized to be where future random walks are likely to end, which is heavier subdags, which is exactly where the random walk would likely take me, so intuitively the incentives are aligned.
3/n IOTA has a paper that analyzes the equilibrium here: assets.ctfassets.net/r1dr6vzfxhev/2…
The paper can be summarized by this quote:
"Our results show that the studied selfish strategy outperform the non-selfish ones
by a reasonable order of magnitude."
The paper can be summarized by this quote:
"Our results show that the studied selfish strategy outperform the non-selfish ones
by a reasonable order of magnitude."
4/n Since there's no gas fees, what would be an incentive to behave differently? One such incentive is resources. IOTA is designed to run on low end devices, and storing the entire DAG or a large subgraph of it is very expensive
5/n Here's a possible alteration to one's behavior: instead of performing random walks, do not store the DAG at all, and instead monitor the network and remember the last 10 transactions gossiped. Approve two of them when posting your transaction.
6/n If the majority of other participants follow the random walks, the last 10 transactions overheard are the tips of some random walks, and thus approving them will provide a very good chance of being approved in the future.
7/n Thus, individual participants have a rather high incentive to switch to this approach, since it saves space without trading finality.
8/n Even if majority of participants switch to this approach, transactions keep being posted and approved, and for an external observed it would be extremely hard to spot that it's happening.
9/n But the security of IOTA goes to zero, since now to create a major fork in the DAG it is sufficient to post 10 transactions in a different subdag, and all the lazy participants will start approving them.
10/n The moral is: Designing proper game theoretic incentives is hard.
