Your Loss is My Gain: Low Stake Attacks on Liquid Staking Pools

A team of researchers led by Sen Yang, Aviv Yaish, Arthur Gervais, and Fan Zhang has published a paper revealing a new attack surface in permissionless Proof-of-Stake (PoS) blockchains that use liquid staking. Their study, supported by empirical analysis of Ethereum data, demonstrates that a low-stake adversary can manipulate the consensus layer to degrade a specific liquid staking pool's operational performance. This manipulation is then exploited on the application layer: the attacker takes short positions on the pool's liquid staking token (LST), profiting when the market reprices the token to reflect the worsened performance. They developed a deep reinforcement learning (DRL) framework that automatically discovers near-optimal attack strategies, which can significantly impair target pool performance.

The paper includes detailed Monte Carlo simulations showing that leveraged shorting of LSTs from major staking pools can be profitable with over 50% probability. The attack does not require a majority stake—only a modest amount of ETH is needed to trigger validator-level disruptions. The findings highlight a fundamental gap between the economic security assumptions of PoS and the practical risks introduced by liquid staking, where performance metrics directly correlate with token returns. This cross-layer vulnerability could impact major liquid staking protocols like Lido or Rocket Pool, potentially affecting billions in staked value. The researchers recommend re-evaluating the security models of liquid staking systems to account for these low-cost manipulation vectors.