Democratizing MEV: How Marinade uses SAM to beat sandwiching and MEV Abuse

Key Takeaways:

• On average, less than 10% of staked SOL with Marinade is assigned to validators participating in sandwich attacks and other MEV strategies.
• However, in cases where SOL is staked to MEV validators, higher staking yields are being paid out to stakers (redistributing MEV profits).
• The problem of Maximum Extractable Value (unchecked validator profit at the expense of users) can be mitigated by turning it into a mechanism for staking rewards.

A Note on Maximum Extractable Value (MEV)

The economics of validating a blockchain means prioritizing maximum extractable value, or MEV. Validators can ‘extract’ or earn more from their operations in how they order transactions or decide which transactions to include or exclude from a block. Maximum extractable value harms a blockchain, however, when it imposes costs on its users.

Solana is under increasing pressure from malicious validators and sandwich attacks

With its high throughput and low fees, Solana's design makes it particularly susceptible to MEV attacks. Validators can exploit transaction sequencing at a scale not possible on other blockchains. Recent reports estimate that a single Solana validator extracted over $60 million in a single month through sandwich attacks.

The centralization of Jito, Solana’s leading MEV infrastructure, has only increased concerns about Solana’s resilience to sandwich attacks and other malicious validator behavior. Despite initiatives like Jito’s attempt to create a “fair” MEV market, validators engaging in sandwiching still benefit disproportionately. Because delegation remains largely opaque, knowing which validators are complicit is difficult.

This trend concerns the Marinade Labs team because it threatens its users, who rely on the Marinade protocol to connect them with high-performing validators with whom to stake their Solana. Their SOL is automatically staked with Marinade Staking Auction Marketplace's highest-performing and highest-bidding validators.

We took it upon ourselves to study which validators in Marinade’s marketplace regularly engaged in this malicious behavior. This study includes a brief summary of our findings and the team’s proposed next steps to further strengthen Solana and its staking ecosystem.

Methodology to Identify Malicious Validators

Leveraging open-source tools like the a-guard malicious validators repository, the Marinade Labs team analyzed recent epochs for sandwiching validators, cross-checking if they were listed in the Marinade Staking Auction Marketplace.

The team:

• Identified known sandwiching programs and their related transactions.
• Located pairs of transactions in which the token amount increased in a single block, a hallmark of sandwiching attacks.
• Identified paired transactions in the same block that did not send a Jito tip, which indicates that only the block producer could have facilitated the attack.
• Mapped these transactions back to the validators responsible for producing the blocks.
• Cross-referenced these validators with Staking Auction Marketplace participants to determine their impact on our users.

Research Findings:

The Marinade Staking Auction Marketplace automatically delegates SOL to the highest-performing and highest-bidding validators. Our intent was to discover whether the validators executing sandwich attacks were using our marketplace.

By reviewing recent epochs, we found that:

• Between 3.7% and 9.6% of staked SOL was earmarked for validators participating in sandwich attacks. These validators clearly use their profits to make higher bids in our marketplace for staked SOL.
• These validators are using more than just Marinade’s staking auction marketplace to make their attacks - in some epochs, up to 770,000 SOL (8.8% of all staked SOL) was allocated to these validators. Epoch 734 was particularly impacted, with 1.48M SOL (~17.1% of total stake target) for blocklisted validators.
• The most recent SFDP blocklist does not include all malicious validators - when applied to the Staking Auction Marketplace, up to 17% of staked SOL would have been eligible to go to these validators.

In summary: Malicious validators are using staking platforms, including the Marinade Staking Auction Marketplace, to capture delegated SOL for staking. In the marketplace, they are bidding higher for this SOL, likely due to their increased profits from MEV.

This is, however, good news for Solana stakers. The Stake Auction Marketplace forces malicious validators to return the value they earn from these attacks to the stakers themselves through higher staking yields.

How the Stake Auction Marketplace Protects Stakers from Sandwich Attacks

The Stake Auction Marketplace, in simple terms, makes sandwiching attacks more difficult for malicious validators.

First, validators must set their staking yield transparently to bid for this SOL. They then need to bid highly to be delegated SOL for staking—this ensures that Solana holders earn yield, no matter what. Even if these validators later benefit from MEV, they are, in effect, returning this profit to the stakers through these bids.

But even if this malicious validator regularly bids the highest for the SOL, they will never have the majority of the stake - the marketplace redistributes staked SOL dynamically to prevent centralization. This is unique to Marinade, as other delegation models are static and can more easily be exploited.

While our marketplace does redistribute extracted value back to stakers, we are still concerned about malicious validators misusing it to reinforce their MEV operations. Here’s what we intend to do about it.

How Marinade is Combatting Malicious Validators and MEV

First, the Marinade DAO is proposing the implementation of a blocklist mechanism within our marketplace that allows the protocol to ban recurrent malicious validators. Validators with a clear, repeated, and harmful pattern of behavior that disrupts fair and open market dynamics will be banned. You can weigh in on this proposal here.

Simplify Marinade’s delegation strategy to give high-performing validators more room for growth and allow more good actor validators to participate - this was proposed in February here.

Another initiative focuses on collaborating with top-tier, trusted, and compliant validators to develop a new protocol — currently in beta — that enables secure staking for large SOL holders while granting white-listed validators access to institutional stake.

The team at Marinade Labs is also in the process of setting up a grant for other initiatives and developments to combat MEV and will share more news about this soon.

What Solana Needs to Fight MEV

Marinade has built the Stake Auction Marketplace to ensure Solana staking could remain attractive and stable - for now, it is the only staking solution that democratizes MEV by redistributing its winnings to stakers.

Marinade’s marketplace does not eliminate MEV but instead creates a mechanism where validators performing negative MEV must compensate stakers.

This is a fundamental shift. With traditional staking, malicious validators extract value from the network unchecked. With Marinade, they must bid competitively, giving a portion of extracted MEV back to stakers.

The goal for Solana should be similar - to transform MEV from a validator-centric profit center into a staker-aligned, market-driven redistribution system.

In the future, protocol-level changes to the Solana blockchain could help make this possible, including:

• Concurrent Block Producers — Reducing the ability of a single validator to control transaction ordering.
• Shorter Block Times — To limit the window of opportunity for MEV extraction.
• Transparent Order Flow — Encouraging open mempool designs that minimize asymmetry.

Until then, we will keep refining the Marinade Stake Auction Marketplace to ensure that Solana staking remains fair, transparent, and rewarding for all participants.

‍