How DeFi cannibalizes PoS security

Piggybacking off recent research paper by Tarun Chitra from Gauntlet, Haseeb Qureshi suggests that on-chain lending markets and staking directly compete with each other leading to interesting implications on network security. He claims that POS networks are only secure if participants are incentivized to stake and that participants are only incentivized to stake if the reward is high enough. Theoretically, economically rational actors would stop staking if they could get a more attractive yield elsewhere such as on-chain lending markets. If participants do stop staking in pursuit of a higher return elsewhere, the network would be less secure and more vulnerable to 51 percent attacks.

Haseeb models out such a scenario using a technique known as agent-based simulation. What the results showed between a simulation of ETH in Compound and ETH staked is that while most ETH holders stake their ETH initially, as the block reward fell and the stake rate became less attractive relative to Compound lending rates, participants rebalanced their staked ETH over to Compound. The conclusion reached is that if a networks POS block rewards decrease over time, then its long-run equilibrium will be for almost all assets to be lent, not staked. Therefore POS must have adaptive monetary policies to respond to stake demand fluctuations.

Haseeb suggests that such an attack could be performed by subsidizing on-chain lending markets (borrowing a ton of the asset in order to raise rates) in order to drive stakers away from staking and towards lending. Once the amount staked is depleted the attacker could attack the chain more cheaply. “This could lead to a snowball as onlookers see the total stake shrinking, they now want to go short ETH, further increasing the borrow demand on Compound.” All this could be done while eliminating ETH price exposure by collateralizing the loans with other crypto assets such as USDC or tokenized Bitcoin.

Why it matters:

• Touted for their increased efficiency and environmental friendliness, POS chains have been heralded as the “next step” in the evolution of blockchain consensus mechanisms. Nearly every single network that has launched this year or will launch next year (“The Ethereum Killers”) uses some form of POS consensus, including high profile projects such as Cosmos, Algorand, Polkadot, Dfinity, and Hedera Hashgraph. Although we now have some data points on how these systems perform in the wild with some of these POS networks having launched, the above analysis shows just how little we may understand about POS security in the real world.
• Blockchains do not exist in a vacuum and their security may depend on many factors exogenous to the protocol. Furthermore, turing complete blockchains theoretically introduce an infinite amount of complexity to a network that may have unpredictable consequences on the security of the underlying chain.