This is a segment from the Lightspeed newsletter. To read full editions, subscribe.
It seems like conversations about sandwich attacks on Solana — where advanced traders front-run and back-run trades to extract value at the expense of unsophisticated traders — come and go every few weeks with little resolution. Most people agree sandwiching is bad, but there’s not yet consensus on how the problem can be solved.
The discourse flared up again this week after the Solana researcher Ben Coverston posted on social media that the network’s top sandwich bot, dubbed arsc, was pocketing millions of dollars-worth of tokens per day from sandwich attacks. Coverston predicted that if current trends hold, arsc would become Solana’s largest staker within 1-2 years. Interestingly, one of the researcher’s proposed fixes included the reintroduction of a “mempool” staging area for transactions — which Jito got rid of earlier this year in hopes of preventing sandwiching.
Sandwich attackers exploit slippage — the gap between a trade’s expected and executed price. They front-run the target transaction by placing their own trade first, which shifts the price in their favor. Then, they back-run the same transaction by placing another trade immediately after, locking in a financial win off of the manipulated price movement. As a result, the original trader often ends up with a worse execution price, effectively subsidizing the attacker’s profit.
This is a form of maximal extractible value (MEV) where sandwichers extract value from Solana blocks at the expense of less sophisticated users. Solana has taken a number of actions meant to prevent sandwiching in recent months. In March, Solana infrastructure company Jito Labs got rid of its public mempool, which was a staging area where transactions could be arranged to create MEV revenue for validators.
Sandwiching validators then set up their own private mempools, and the attacks continued. Two months later, the Solana Foundation stopped offering financial help to validators who were participating in private mempools. In September, Jito blacklisted a group of private mempool validators from participating in its stake pool, which is a large pile of staked SOL Jito’s software delegates.
The criticism of the Solana Foundation and Jito’s blacklist-the-bad-guys strategy is that it’s essentially playing whack-a-mole: Blacklisted validators can always spin up a new validator and resume sandwiching.
That’s what’s interesting about Coverston’s proposal, which he expanded upon in a blog post. The privatization of mempools has reduced the number of validators executing sandwich attacks, but the few who remain, like arsc, are staking their substantial profits back into their own operations. This has led to a concerning centralization of staked Solana within a handful of nefarious validators.
To counter this, Coverston suggests re-enabling a public mempool. While this move might make sandwiching more widespread, it could also foster competition among validators executing these strategies. The increased competition would help distribute MEV profits more evenly, preventing a single group of validators from disproportionately accumulating stake — and power — on the network.
This would essentially be a reversal from Solana’s anti-sandwiching strategy over the past nine months.
To put this all in perspective though, sandwiching may not really matter in the long run. High-frequency trading strategies that are ascendent in the US equities market are basically meant to sandwich attack other traders, Blockworks Research’s Ryan Connor pointed out. That hasn’t stopped equities trading volumes from climbing to all-time highs.
