The Hidden Costs of Sandwich Attacks: How Much Retail Investors Overpay for 'Free Trading'?

MEV sandwich attacks are not order flow payments, but rather front-running practices prohibited in regulated markets.

Written by: Yuval Rooz, Co-founder and CEO of Digital Asset

Compiled by: AididiaoJP, Foresight News

Note: The author is the co-founder and CEO of Digital Asset, which developed the Canton Network.

MEV priority fees are just the lower limit of losses; front-running is the hidden bill.

Describing MEV as "the crypto version of order flow payments" is a common saying in the industry. The author believes this analogy does not hold.

Order flow payments come with best execution, price protection, and disclosure requirements, while MEV sandwich attacks do not. Retail investors pay worse execution prices on public chains, which are invisible and not itemized.

The article has three core judgments.

First, order flow payments come with protections, while MEV does not. Second, MEV priority fees underestimate the harm. What front-runners pay to validators is only a portion of what they take from traders. Third, MEV is a design choice, not a natural law. Sandwich attacks require two prerequisites: visibility of pending orders and the ability to pay to insert themselves ahead. Both prerequisites can be removed.

The author also reminds that meme coins are the most obvious harm at present, and tokenized stocks are next.

I. The Overlooked 'Market Integrity'

A recent guideline tells financial institutions that permissionless chains meet financial integrity standards. However, this guideline discusses anti-money laundering and sanctions compliance, not market integrity.

"Permissionless Networks and Financial Integrity" was written by Rebecca Rettig (COO and Chief Legal Officer of Jito Labs), Omid Malekan (Columbia Business School), and Michael Mosier (former Acting Director of FinCEN). The full text is 27 pages long, and MEV, front-running, and sandwich attacks are not mentioned once.

The paper itself delineates this issue out of scope. Footnote 6 lists "transaction ordering determinism" as outside the discussion and claims existing work has "substantially mitigated" it, but provides no evidence. Transaction ordering is precisely where retail investors lose money.

The paper's own wording exposes the issues it skips over.

It does not fully describe the fees. The paper states that users pay "fees determined by the protocol-selected validators." However, on Solana, users and bots also pay off-protocol priority fees to control transaction landing spots. This market is operated by Jito, which is the company of the first author.

It lists rent extraction as a risk. The paper states that institutions can avoid risks such as "network operator rent extraction" on permissionless networks. MEV is precisely the rent extraction conducted by network operators.

It promises confidentiality from competitors. The paper states that institutions can "provide no information to competitors" when trading. However, on a chain where pending orders can be seen and front-run, institutions effectively reveal their next moves to competitors.

After reading this guideline, institutions might believe that integrity issues have been resolved. However, the issue of client execution prices has not been addressed at all.

II. Why MEV Cannot Be Compared to Order Flow Payments

A common defense is that MEV is just the crypto version of order flow payments. In traditional finance, market makers also profit from retail orders, which is why Robinhood offers free trading.

This comparison overlooks all the protections that allow order flow payments to stand.

Order flow payments require brokers to still achieve best execution, with execution prices meeting or exceeding the national best bid and ask prices, and to disclose relevant arrangements. Traders who are subject to sandwich attacks receive none of these protections.

Even with these protections, order flow payments remain controversial. In 2014, UK regulators required brokers to stop accepting order flow payments (FCA TR14/13). The European Securities and Markets Authority also stated in 2021 that it is unlikely to comply with EU rules.

Defending MEV with order flow payments is akin to borrowing the legitimacy of a regulated practice while refusing to accept the rules that grant it legitimacy.

Not all MEV is equally harmful. Cross-market arbitrage can lead to price convergence, and clearing can maintain the solvency of lending markets, similar to traditional high-frequency trading: racing to close price gaps using co-location.

However, competition in traditional high-frequency trading is fierce. Co-location is a public product, and any company can rent racks in exchange data centers at published prices. Exchanges also make the cables from each co-located server to the matching engine equal in length, ensuring that racks at the end of the hall are as fast as those next to them. Physical location is deliberately made irrelevant. Many companies compete on the same starting line for the same opportunities, and the profit from individual price differences is driven close to the cost of capture by competition.

Arbitrage MEV is not like this in practice. Whoever controls the ordering decides the winner, and winners are few.

On Ethereum, as of November 10, 2026, a single builder, Titan, constructed 53% of blocks through MEV-Boost, with the top three builders accounting for 85% (relayscan.io). Theoretically, this is benign MEV; in practice, it is neither as competitive as the markets it is compared to nor as open.

The industry often claims that public chains are fair because they are decentralized. The ordering layer tells another story: thousands of validators, but ultimately a few companies decide who goes first. This is "decentralized performance."

Traditional markets are often described as closed systems, yet they achieve equal access to the point that even cable lengths must be consistent. Chains that promise openness, on the other hand, concentrate the most critical advantage.

Even if the ordering rules are intended to protect liquidity, once positions can be bought, the rules can fail. Hyperliquid gives priority to cancel orders, allowing market makers to withdraw expired quotes before the taker executes. However, a trading firm analyzing its data found that under heavy load, cancellations can be pushed to later blocks, while immediate or cancel orders, especially those paying 8 basis points in priority fees, are hardly affected. When there are over 40,000 transactions per second, 31% of failed cancellations are signed before they consume their paid orders. On November 9, 2026, in one second, $1.27 million in AVAX sell orders were consumed, causing market makers to lose about $13,000 within 10 seconds (analysis on X). Market makers whose quotes are swept will widen the spreads, and all traders will pay for it.

Sandwich attacks are not a matter of degree but of nature. The industry knows this. Jito's closure of public mempools to curb sandwich attacks is an acknowledgment that this practice harms the users it is supposed to serve.

III. Priority Fees Are Just the Lower Limit of Losses

The money that front-runners pay to validators is always less than what they take from traders. No one will pay $8,000 for a trade that earns $7,000. Therefore, MEV priority fee data measures the minimum harm, not the total harm.

For example, a stock is quoted at $99 to buy and $100 to sell. A buyer places a limit order for 10,000 shares at $100. The market has enough liquidity to execute all orders at $99.30, which is within the quoted range, and best execution requires brokers to find it. The front-runner sees the order first, buys the liquidity at $99.30, and then sells it to the buyer at a limit price of $100.

On-chain records show $2,000 in MEV. The buyer actually loses $7,000.

Real data also shows the same gap. From December 7, 2024, to January 5, 2025, a Solana sandwich program executed 1.55 million sandwich trades, earning 65,880 SOL, approximately $13.43 million in profit, while paying 22,760 SOL, about $4.63 million in Jito priority fees (Helius). Looking only at priority fees, one can see only about a quarter of the value taken from traders. Helius also points out this limitation: Jito data "does not include the profits of searchers."

The retail scenario is similar. A meme coin buyer spends $1,000 with a slippage set at 10%. The bot jumps in front, pushing the price to the tolerance edge, and then sells afterward. The buyer receives tokens worth about $100 less than if they had not been sandwiched. If the bot pays a $30 priority fee to secure the position, the data only shows $30; the trader loses $100. (Example numbers.)

There are two metrics to maintain. The total priority fee includes benign MEV such as arbitrage, so when discussing the lower limit, one should use sandwich-specific priority fees. The trader's loss equals the priority fee plus the bot's profit; the two cannot be added again, or else sandwich losses will be double-counted.

IV. Solana Data: Large Scale, Concentrated in Meme Coins, Still Ongoing

The scale of sandwich attacks on Solana is large, concentrated in meme coins, and has not stopped.

From January 2024 to November 8, 2026, Jito priority fees totaled 8.08 million SOL, approximately $1.42 billion (Dune query of 8 public priority fee accounts from Jito).

The share of Solana staked running the Jito validator client increased from 48% at the beginning of 2024 to 92% at the beginning of 2025 (Helius).

A sandwich program executed 1.55 million sandwiches in the 30 days ending January 5, 2025, with profits of $13.43 million and priority fees of $4.63 million (Helius).

Among the 20 tokens most affected by sandwiching, 16 were created on pump.fun (Helius).

In a recent report window, 606 attackers launched 224,964 sandwiches, with 52,728 victims, with data access on November 7, 2026 (sandwiched.me).

Priority fees peaked at 1.29 million SOL during the meme coin frenzy in January 2025, then fell back to about 40,000 to 50,000 SOL per month. The total for 2024 was 3.7 million SOL, which differs by less than 2% from the 3.75 million SOL reported by Helius. Priority fees are the lower limit of trader losses, not the total.

The core number is: since January 2024, Solana users have paid about $1.42 billion for trading positions. This includes $661 million in 2024, $705 million in 2025, and $49 million so far in 2026; in January 2025 alone, there was $285 million. The dollar values are converted based on the average SOL price per month from Dune.

Not every priority fee is used for sandwiches. Priority fees also buy fast packaging and fund arbitrage. However, the proportion of sandwiches itself is the lower limit, as bots take most of the proceeds. In the above Helius case, priority fees account for about a quarter of the extracted value.

V. Validator Economics: Who Pays for 'Free Trading'

On large chains, ordering revenue has become the core of validator income.

On Solana, users and bots paid 3.75 million SOL in Jito priority fees in 2024; the share of validators running the Jito client increased from 48% to 92% (Helius). Estimates suggest that Solana's annual revenue is about $1.5 billion, of which only about $100 million comes from base fees (RYT, March 2026). Base fees are close to zero, with money mainly coming from positions.

This raises incentive issues. As validator income increases with competitive ordering, the party protecting the network earns more when user transaction prices worsen. No one needs to act maliciously; the income structure is simply like that.

Removing MEV would significantly change validator economics. Income would need to come from elsewhere:

• More token issuance. Validators receive newly minted tokens, diluting all holders.
• Higher explicit fees. Users pay visible fees, and "free trading" is no longer free.
• Fewer validators. Operators unable to cover costs exit, weakening decentralization.
• Another model. Solutions like destroying MEV, auctioning order flow, and returning profits to users attempt to recapture value. Canton Network rewards validators for operating the network, rather than rewarding them for ordering user transactions.

The uncomfortable conclusion is that public chains appear cheap partly because users pay through transaction prices rather than through visible fees. This is similar to criticisms aimed at zero-commission brokers, except broker clients still have best execution and disclosure protections.

VI. Removing Two Preconditions

A sandwich attack requires two things: the attacker can see pending orders and can pay to insert themselves in front of them. Remove one, and the attack becomes harder; remove both, and the attack cannot occur.

Jito closed public memory pools, removing part of the first condition. Initially, it reduced harmful MEV, but private memory pools emerged, and Jito's own analysis found that more and more sandwich attacks occurred outside its auctions. Sandwich attacks continue today: a tracker recently reported over 220,000 attacks and more than 52,000 victims within a recent window. No matter how good the intentions, a design that profits from selling positions cannot be fixed.

The author states that, as described, Canton Network attempts to remove the two preconditions for sandwich attacks:

• No public memory pool. Pending transactions do not enter a publicly visible queue, making it difficult to observe pending orders.
• No paid ordering market. Users cannot pay priority fees or tips to change transaction placement.

If these two points hold, frontrunners see neither orders nor can they buy a front position, making sandwich attacks lack the conditions for implementation under this network design. This refers to mechanical conditions, not external validation of actual operational effects.

According to Canton Network documentation, transactions are ordered by the Global Synchronizer, using a Byzantine fault-tolerant protocol run by independent Super Validators, with no single entity controlling the ordering. The documentation also states that these operators cannot see transaction data, only the limited metadata needed for ordering and confirmation. Network fees are paid in Canton Coin and destroyed; validators earn newly minted Canton Coin for providing services, rather than priority fees paid by users competing for positions.

SupaNova is a meme coin trading application recently launched on Canton Network, which the author uses as an observational case. Meme coin trading is one of the scenarios where other chains extract the most MEV from retail investors. According to the author, on Canton, this type of protection is determined by network layer mechanisms, not relying on how individual applications are built. However, this judgment is still based on project documentation and the author's stance, and cannot replace independent evaluation.

VII. Questions for Issuers

Meme coins are speculative corners, and some might say traders accept the risks. But the same infrastructure now begins to carry stocks.

xStocks offers over 100 tokenized U.S. stocks and ETFs, reporting over $47 billion in trading volume, and trades on Solana venues like Jupiter (xStocks). Ondo Stocks offers over 470 tokenized stocks and ETFs on Ethereum, BNB Chain, and Solana (Ondo). bStocks and Robinhood's stock tokens for international clients also extend the same model.

When assets are tokenized stocks, sandwich attacks are no longer a gray area. Trading ahead of customer orders in securities trading is called frontrunning, which has been prohibited in regulated markets for decades. A chain that rewards validators for ordering trades has built this practice into its settlement layer.

The issuer's question is simple: would you list stocks on a venue that allows others to frontrun every one of your shareholders' orders? When you tokenize on a chain with MEV, you are choosing that venue.

This article's data comes from Solana because its data is public, and the meme coin market is also the largest. But Solana is not an exception. Any chain where pending orders are visible and positions can be bought allows sandwich attacks; mainstream public chains share this design.

On Avalanche's C-Chain, documentation states that transactions are first sorted by priority fee and then by timestamp. Priority fees will be destroyed, and validators cannot directly earn income from them, but frontrunners can still compete for positions by bidding higher than victims. Thus, destroying fees eliminates the incentive for validators but does not eliminate the attack itself.

This article's topic arises from two aspects: first, SupaNova's launch on Canton Network, and second, the MEV issues exposed by meme coin trading. The impact of this issue is not limited to meme coins. As real-world assets go on-chain, supporters often describe tokenization as bringing faster settlement, broader access, and lower costs. If the infrastructure carrying these assets instead makes execution quality worse than it is now, these expectations become hard to fulfill. If on-chain finance provides a 24/7 market and instant settlement but extracts corresponding benefits from every transaction, the result will undermine the value proposition of on-chain finance. MEV is not unchangeable; it is a design choice. Chains without memory pools and that do not offer paid ordering have chosen different mechanisms.

References

• Rettig, Malekan and Mosier, The Compatibility of Permissionless Networks and Financial Integrity: A Practical Guide for Financial Institutions, SSRN abstract 7343938 (2026)
• RYT, "The Architecture Nobody's Building" (March 30, 2026)
• Canton Network, FAQ
• Canton Network, The Global Synchronizer
• FCA, TR14/13: Best execution and payment for order flow (2014)
• ESMA, warning on payment for order flow (2021)
• xStocks
• Ondo Stocks
• Helius, "Solana MEV Report"
• sandwiched.me, Solana sandwich tracker
• Avalanche Builder Hub, "Transaction Fees"
• relayscan.io, MEV-Boost builder statistics
• Analysis of Hyperliquid cancel priority under load, X (November 2026)
• FINRA Rule 5310 (Best Execution); FINRA Rule 5320 (Prohibition Against Trading Ahead of Customer Orders); SEC Rules 605 and 606 of Regulation NMS