Encrypted Lending on Ethereum: How FHE Enables Private, MEV-Safe DeFi

How Fhenix's FHE coprocessor (CoFHE) enables MEV-resistant, confidential lending protocols with full Ethereum composability

TL;DR

• FHE supports private lending on Ethereum by computing over encrypted collateral and debt balances. CoFHE runs today on Ethereum Sepolia, Arbitrum Sepolia, and Base Sepolia, with mainnet planned to follow.
• Borrower positions stay encrypted during computation. Selective permissions let a borrower, auditor, or regulator decrypt specific values without exposing anyone else.
• Build with @cofhe/sdk for client-side encryption and decryption. Solidity contracts import FHE.sol for encrypted types, operations, and access permissions.
• Production gas costs for encrypted lending operations are not yet published. The gas and benchmarks documentation is in progress, so benchmark your own contracts on testnet.

The Missing Primitive in Decentralized Finance

Decentralized finance has evolved through distinct technological breakthroughs. Automated market makers enabled DeFi Summer in 2020. Liquid staking tokens powered the 2023-2024 growth cycle. Each era introduced capabilities that expanded who could participate and what could be built onchain.

The next evolution centers on confidentiality. As institutional capital evaluates blockchain infrastructure, the transparent-by-default nature of current lending protocols creates fundamental barriers to adoption. Every position, every collateral amount, every liquidation threshold remains visible to competitors, arbitrageurs, and the broader market.

Fully Homomorphic Encryption (FHE) is a cryptographic scheme that allows operations on encrypted data without ever decrypting it, and can be used as the technological foundation for private decentralized finance. In the first half of 2025, Fhenix demonstrated that FHE-based encrypted computation operates at practical scale on Ethereum today.

FHE has historically faced performance limitations that restricted its practical applications. Fhenix addresses these challenges through a comprehensive approach combining multiple optimization techniques.

We build CoFHE, an FHE-based coprocessor. In simple terms it works like an engine that handles requests for private computations submitted by different dApps. CoFHE's architecture provides a dedicated encrypted execution infrastructure purpose-built for high-throughput workloads. Key technical approaches include:

• Parallelized threshold decryption enables party members to jointly decrypt results with low latency and distributed trust
• Optimized ciphertext structures reduce memory overhead and accelerate common FHE operations across the system
• An async model that allows calculations to run without delaying the host chain operation, helping to bypass computation latency

Our CCS 2025 research paper, "High-Throughput Universally Composable Threshold FHE Decryption" (ePrint 2025/1781), presents empirical measurements showing the viability of encrypted smart contracts for real-world applications:

• 64,319 decryptions per second in throughput, and 8.48 ms single-operation latency
• Approximately 20,000× throughput improvement over prior threshold FHE baselines
• Up to 37× improvement in latency over prior threshold FHE baselines

These measurements were taken under controlled research conditions (online phase, four parties, 1 ms ping) and demonstrate technical feasibility rather than current production performance. The protocol is not yet integrated into CoFHE's production system. Production results will vary with implementation details, network conditions, and workload characteristics.

Guy Itzhaki, CEO of Fhenix, explains the shift:

"The next wave of DeFi needs privacy by default. In H1 2025, we showed that encrypted compute is no longer theoretical. It's live, scalable, and composable. Developers now have the foundation to build truly Confidential DeFi."

What Encrypted Lending on Ethereum Actually Means

Public Ethereum exposes lending activity through transaction calldata, sender addresses, and contract state. A confidential lending protocol encrypts collateral amounts, debt balances, and health calculations while preserving the rules that govern borrowing and liquidation. Transaction timing, sender addresses, gas usage, and contract interactions may remain visible.

Fully homomorphic encryption lets a contract compute directly on ciphertext. With Fhenix and CoFHE, Solidity contracts compare encrypted collateral values, update encrypted debt, and evaluate lending conditions without the execution nodes seeing plaintext. Authorized decryption requires an explicit on-chain permission. Independent nodes then combine key shares through threshold computation, so no single node can decrypt a position alone.

Zero-knowledge proofs and trusted execution environments use different privacy models. A zero-knowledge proof shows that a computation followed specified rules without revealing its private inputs to the verifier, though the prover generally handles those inputs. A trusted execution environment processes plaintext inside a hardware-protected enclave and relies on the chip vendor's security model.

For encrypted lending, the practical FHE model separates private values from public metadata. Borrower positions stay encrypted during computation, while selective permissions let a borrower, auditor, or regulator decrypt specific values. Evaluate any privacy tool by where plaintext exists, who controls decryption, and which transaction details remain observable.

Why Lending Protocols Need Confidentiality

Current Ethereum lending infrastructure, including protocols like Aave, Compound, Morpho, and Spark, operates with complete transparency. While transparency serves important verification purposes, it exposes critical vulnerabilities that prevent institutional adoption:

• Wallet balances and debt positions. Collateral amounts are visible to all market participants, enabling targeted liquidation strategies and position monitoring
• Borrowing demand. Borrower positions can be tracked across protocols or tokens, revealing wallet-level risk profiles and capital deployment strategies
• Liquidation thresholds are public information, allowing sophisticated actors to time market movements for maximum extraction
• Competitive alpha is exposed, signaling the trader's strategy to the market
• Yield strategies deployed by institutions can be reverse-engineered and front-run by competitors

This transparency enables several harmful dynamics:

• Liquidation sniping occurs when observers monitor positions approaching liquidation thresholds and execute transactions at precisely calibrated moments
• MEV extraction systematically captures value from borrowers through transaction ordering
• Market surveillance allows competitors to profile strategies and adjust their own positions accordingly
• Capital inefficiency forces sophisticated participants to account for information leakage in their position sizing

As institutions evaluate onchain lending, these transparency issues create compliance and competitive concerns that override technical benefits. Dr. Guy Zyskind, Founder of Fhenix, emphasizes the fundamental nature of this challenge:

"If DeFi remains transparent by default, it will never serve institutional scale. Confidentiality is not optional. It's fundamental."

Fhenix addresses these limitations by enabling developers to build fully encrypted lending protocols that preserve Ethereum's core strengths while adding confidential computation.

FHE vs ZK vs TEE for a Confidential Lending Protocol

Match the privacy architecture to where plaintext can appear, how contracts share private state, and which trust assumptions you can accept.

FHE fits lending contracts that repeatedly calculate over shared encrypted balances and need selective disclosure. ZK fits proof-oriented workflows with user-controlled private state. TEE designs favor low-latency general computation when hardware trust is acceptable.

Building Encrypted Lending Protocols on Ethereum and EVM Chains

Fhenix provides infrastructure for deploying confidential lending protocols directly on Ethereum and any other EVM chain without compromising user experience or composability. Developers can launch encrypted applications while maintaining compatibility with:

• Standard wallet infrastructure including MetaMask, Ledger, and Rabby. Users interact with encrypted protocols through familiar interfaces without additional setup or specialized software
• EVM equivalence ensures encrypted contracts integrate seamlessly with existing Ethereum tooling, development workflows, and deployment infrastructure
• Composability enables running confidential logic within smart contracts, so builders can build any dApp they want with privacy built in
• Open-source development tooling provides familiar frameworks and libraries, reducing the learning curve

With that in mind, lending protocols can integrate with Fhenix's stack to decide which components or data should remain private, such as:

• Keep borrower balances encrypted throughout their lifecycle to prevent liquidation sniping
• Encrypt collateral values so they never appear as plaintext onchain
• Keep the LTV (Loan-to-Value) ratio confidential and encrypt the liquidation logic using FHE functions to compute over the encrypted data
• Proof-of-reserves and protocol-level solvency remain publicly verifiable through threshold decryption mechanisms that selectively reveal aggregate data without compromising individual positions

CoFHE provides encrypted state, encrypted execution, and encrypted transactions while maintaining EVM equivalence.

This design allows lending protocols to operate with institutional-grade confidentiality while preserving the verification properties that make public blockchains trustworthy.

Encrypted Lending Architecture on CoFHE

Fhenix enables smart contracts to compute on encrypted data without ever revealing it. This preserves privacy while maintaining Ethereum compatibility, and changes what is possible in lending and credit.

Encrypted Lending Use Cases

1. Confidential Collateral Positions. All sensitive protocol data exists as FHE ciphertexts rather than plaintext. Users can borrow without revealing how much they have posted or withdrawn as collateral. This encrypted state remains opaque to observers while remaining computationally accessible to the protocol itself.
2. Shielded Loan Terms. Lenders and borrowers negotiate terms (rates, durations, caps) privately, avoiding front-running or copycat strategies.
3. User credit score. Calculating a credit score that reflects a user's financial activity using encrypted logic, while keeping the score private, lets a lending platform offer better rates based on that score, giving the platform a competitive edge.
4. Private Liquidation Thresholds. Liquidations can be triggered by encrypted logic, minimizing attack vectors and stress visibility.
5. DAO or Treasury Lending with Hidden Exposure. Institutions can lend without disclosing how much risk they are taking, or in which markets.
6. Sealed Bids for liquidation. Replace mempool races with sealed commitments (commit-reveal, private relays) or builder/solver auctions that accept encrypted bids and settle atomically. That prevents the gas wars and MEV dynamics seen in Maker's Black Thursday. (Maker moved away from the 2020 parameters after the incident, but the lesson stands.)
7. Selective Transparency via Threshold Decryption. While individual positions remain private, protocol-level metrics can be selectively revealed through threshold decryption: proof-of-reserves becomes publicly verifiable, total value locked can be disclosed for transparency, and aggregate utilization rates provide market insight, while individual user data remains encrypted. This satisfies regulatory and transparency requirements while preserving user privacy.

High-Demand Encrypted Lending Use Cases (Based on Market Research)

Fhenix product-market research shows that developers, institutions, and risk teams consistently request privacy-preserving DeFi primitives. These represent the most in-demand applications for encrypted lending on Ethereum and EVM chains.

1. Encrypted Per-Account Collateral Balances (Private Collateral Positions). Confidential collateral accounting is the top requirement for private lending protocols. With encrypted collateral values, FHE prevents public visibility into who is close to liquidation, removing the "liquidation hit list" that MEV bots scrape today. Solvency remains verifiable through proofs of Health Factor (HF ≥ 1) without revealing raw numbers. This is the foundation for MEV-resistant lending and institutional DeFi privacy.
2. Encrypted Per-Account Debt Balances (Private Borrowing Exposure). Debt values encrypted under FHE prevent adversaries from inferring leverage, monitoring repayment patterns, or manipulating markets to force liquidation. Instead, protocols reveal only proofs such as "debt ≤ max borrow under risk parameters."
3. Encrypted Per-Account Health Factor (Private Liquidation State). Rather than publishing exact Health Factors, protocols can provide binary proofs: HF ≥ 1 means safe, HF < 1 means liquidatable. This removes precision targeting and eliminates MEV bots timing liquidations.
4. Sealed Liquidation Bids and Pricing Intent (Encrypted Auctions for Fair Liquidations). Liquidations operate via sealed bids, commit-reveal flows, or builder-routed encrypted bundles, preventing front-running, gas wars, copycat bidding, and ordering manipulation.
5. Private Rescue Orderflow (Encrypted Top-Ups and Repayments). "Save my position" actions route through private orderflow so distressed positions are not exposed. This prevents sandwiching, griefing, and MEV sniping during top-ups or repayments.
6. Per-Account Collateral Composition (Private Asset Mix). Even if collateral amounts are hidden, revealing which assets a user has posted exposes oracle-manipulation vectors. If an attacker sees a wallet heavily collateralized by a thin LP token or a low-liquidity ERC-4626 asset, they can manipulate inputs or oracles to push that asset into liquidation range. Fhenix enables collateral composition to remain fully encrypted while still publishing proofs that a user meets all basket-level requirements, raising the attacker's cost and complexity.
7. Pending Liquidation Queue and State Transitions (Encrypted Lifecycle Management). DeFi liquidations often rely on highly visible queueing: who is next, when eligibility triggers, timestamps, block ordering. That transparency enables last-block sniping, ordering games, and premature front-running of at-risk borrowers. With FHE, protocols can hide per-address queue positions and timing, publishing only liquidation eligibility and auction state. Pairing this with Dutch-style price discovery replaces speed races with fair liquidation mechanics while keeping distressed users protected.

Getting Started: Build a Confidential Lending Protocol on CoFHE

Despite the underlying cryptographic complexity, the user and developer experience stays familiar. End users interact through standard Ethereum wallets without modification. Developers continue writing Solidity using familiar patterns and tooling. No circuit design or specialized cryptographic knowledge is required.

Implementation requires only three changes:

1. Replace sensitive fields with encrypted datatypes such as euint64 instead of standard Solidity integers
2. Use library functions such as FHE.add, FHE.gt, and FHE.mul for operations on encrypted values
3. Deploy and test using standard Ethereum development tools

CoFHE currently supports Ethereum Sepolia, Arbitrum Sepolia, and Base Sepolia, with mainnet planned to follow.

Set up the project. Follow the CoFHE quick start: install Node.js 20 or later and pnpm, then clone the CoFHE Hardhat starter:

git clone https://github.com/fhenixprotocol/cofhe-hardhat-starter.git
cd cofhe-hardhat-starter
pnpm install

Write the contract. The skeleton below stores collateral and debt as encrypted 64-bit integers. It omits token transfers, collateralization checks, interest accrual, and liquidation logic, which a production protocol must add.‍

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.25;

import "@fhenixprotocol/cofhe-contracts/FHE.sol";

contract EncryptedLendingPool {
  mapping(address => euint64) private collateral;
  mapping(address => euint64) private debt;
  mapping(address => bool) private opened;
  euint64 private poolLiquidity;

  constructor(uint64 initialLiquidity) {
    poolLiquidity = FHE.asEuint64(initialLiquidity);
    FHE.allowThis(poolLiquidity);
  }

  // Initialize encrypted values on first use.
  function _open(address user) private {
    if (!opened[user]) {
      collateral[user] = FHE.asEuint64(0);
      debt[user] = FHE.asEuint64(0);

      FHE.allowThis(collateral[user]);
      FHE.allowThis(debt[user]);

      opened[user] = true;
    }
  }

  function deposit(InEuint64 memory encryptedAmount) external {
    _open(msg.sender);

    euint64 amount = FHE.asEuint64(encryptedAmount);

    collateral[msg.sender] = FHE.add(
      collateral[msg.sender],
      amount
    );

    FHE.allowThis(collateral[msg.sender]);
    FHE.allowSender(collateral[msg.sender]);
  }

  function borrow(InEuint64 memory encryptedAmount) external {
    _open(msg.sender);

    euint64 amount = FHE.asEuint64(encryptedAmount);

    debt[msg.sender] = FHE.add(
      debt[msg.sender],
      amount
    );

    poolLiquidity = FHE.sub(
      poolLiquidity,
      amount
    );

    FHE.allowThis(debt[msg.sender]);
    FHE.allowSender(debt[msg.sender]);
    FHE.allowThis(poolLiquidity);
  }
}

FHE.allowThis lets the contract reuse each ciphertext in later calculations.
FHE.allowSender lets the borrower request private decryption through @cofhe/sdk without exposing the balance publicly.

Test and deploy. Run pnpm test against the local mock contracts first, then add PRIVATE_KEY and the relevant RPC URL to .env and deploy with a starter task. For a complete walkthrough see Your First FHE Contract, and for permission patterns see ACL usage examples. The CoFHE documentation covers network configuration and client-side encryption with @cofhe/sdk.

Gas Costs and Benchmarks for FHE Lending

Fhenix has not published production gas figures for encrypted ERC-20 transfers, collateral deposits, borrows, repayments, or liquidations. The gas and benchmarks documentation is still in progress, and CoFHE remains testnet-stage, so current measurements are not stable mainnet pricing.

The available CoFHE result measures threshold-decryption performance rather than Ethereum gas. In a four-party research environment with 1 ms network latency, Fhenix measured roughly 20,000× higher throughput and 37× lower latency than prior threshold-FHE work, reaching 64,319 decryptions per second in throughput, and 8.48 ms single-operation latency. Those figures do not predict the gas cost or end-to-end latency of a lending transaction.

Aztec provides a competitive reference point. Its Alpha V5 benchmarks report an average fully private token-transfer cost below $0.05 with native proving time near 2.5 seconds, as of July 2026. Aztec measures a ZK rollup transaction while Fhenix measures threshold decryption, so these are not equivalent operations. Aztec's figures also come from alpha software under active development and may change in later releases.

Cost is only one axis. FHE's distinct advantage is computing on shared encrypted state with native EVM composability and selective disclosure, without a per-application circuit.

Benchmark your own encrypted lending contract on Ethereum Sepolia, Arbitrum Sepolia, or Base Sepolia. Measure each encrypted arithmetic operation, access grant, decryption request, and host-chain transaction under realistic collateral and liquidation flows. Generic FHE cost claims cannot replace contract-specific testnet data.

When to Choose FHE, ZK, or TEE

• Choose FHE for regulated or institutional lending that needs selective disclosure. CoFHE computes over encrypted balances and positions, while threshold decryption lets a protocol grant specific auditors access without publishing borrower data. Fhenix's Access Delegation demo demonstrates this audit-key pattern. FHE does not conceal transaction metadata such as sender addresses, timing, or gas use.
• Choose ZK for anonymous transfers and lending flows that fit defined circuits. ZK proofs work well when a prover must demonstrate transaction validity without exposing inputs to the verifier. Fixed or batched transaction logic can support high throughput, but changing private state logic may require new circuits and proving infrastructure.
• Choose a TEE for low-latency, general-purpose private computation. TEEs run ordinary application code with modest compute overhead, which suits credit models or risk engines that are hard to express with encrypted Solidity types or ZK circuits. Confidentiality depends on the processor vendor, attestation chain, and resistance to hardware attacks.

For compliance-sensitive lending, FHE offers the clearest trust model of the three. As Dr. Guy Zyskind puts it:

"TEEs introduce hardware trust assumptions that institutional compliance teams struggle to accept. FHE provides mathematical certainty."

That refers to cryptographic protection of plaintext during computation. Your protocol still needs access controls, key governance, and policies for authorized disclosure.

Encrypted DeFi Applications Beyond Lending

While lending represents the first major use case for confidential DeFi, the same FHE infrastructure enables privacy across other DeFi primitives.

Encrypted ERC-20 Tokens

Privacy-preserving token implementations that maintain fungibility and composability:

• Balances remain hidden from all parties
• Transfer amounts stay encrypted onchain
• Users verify their own balance through client-side decryption
• Observers see only encrypted state

MEV-Proof Liquidation Systems

Confidential position monitoring without information leakage:

• Monitor encrypted health factors continuously
• Trigger liquidations based on encrypted thresholds
• Execute without revealing position details until completion
• Eliminate liquidation sniping that extracts value in transparent protocols

Encrypted Yield Strategies

Institutional-grade strategy deployment with competitive protection:

• Deploy capital according to proprietary strategies
• Conceal positions and targets from competitors
• Hide rebalancing logic and timing
• Maintain verification properties of public blockchains

Private Vaults and Collateral Management

Treasury management with traditional finance confidentiality standards:

• Manage positions with enterprise-grade privacy
• Execute complex collateral strategies without exposure
• Meet compliance requirements through selective transparency
• Preserve verification through threshold decryption

These applications share a common foundation: encrypted state, encrypted execution, and selective transparency where needed for verification or compliance.

The Infrastructure Opportunity for DeFi Developers

The performance milestones presented in "High-Throughput Universally Composable Threshold FHE Decryption" (CCS 2025) show that the core technical barriers to Confidential DeFi are being addressed. FHE computation operates at speeds that support:

• Real-time liquidation systems
• High-frequency risk monitoring
• Interactive user experiences
• Production-scale throughput requirements

This creates immediate opportunities for developers to build confidential primitives that expand DeFi's addressable market.

Private Lending Markets

Offer institutional participants the confidentiality required for compliance while maintaining the transparency needed for protocol security and solvency verification.

Encrypted Automated Market Makers (AMMs)

Eliminate MEV extraction from liquidity provision and protect traders from sandwich attacks and information leakage.

Confidential Yield Aggregators

Execute sophisticated strategies without revealing positions, enabling competitive advantage through privacy rather than information asymmetry.

Privacy-Preserving Risk Management

Allow protocols to assess systemic risk without exposing individual user positions.

Institutional-Grade Settlement Layers

Bring traditional finance capital onchain without compromising the confidentiality standards expected in regulated markets.

The developers who build these confidential primitives will define the infrastructure layer for the next evolution of decentralized finance. Fhenix provides:

• The cryptographic foundation for encrypted computation via CoFHE
• Developer-friendly tooling that abstracts complexity
• Full Ethereum compatibility for seamless integration
• Testnet-ready infrastructure available today

Confidential DeFi as Ethereum's Privacy Layer

Public blockchains used for regulated assets will need encrypted state with selective, auditable access. Privacy requirements limit unnecessary disclosure, while financial oversight requires authorized inspection. Transparent ledgers expose data too broadly. Systems that prevent later disclosure can leave auditors unable to verify.

Fhenix delivers the infrastructure for encrypted lending on Ethereum. The platform provides:

• Encrypted EVM for confidential smart contracts
• A dedicated FHE coprocessor for blockchain applications
• Research measurements showing approximately 20,000× throughput improvement over prior threshold FHE baselines
• Real encrypted DeFi demonstrations including ERC-20 payments, AMMs, and lending prototypes
• Full compatibility with Ethereum's user experience, development tooling, and composability model

Developers can build privacy-preserving protocols without fragmenting liquidity, forcing wallet migrations, or abandoning the Ethereum ecosystem's established network effects.

Confidential lending is the beginning of a broader shift toward privacy-preserving decentralized finance. As institutional capital evaluates blockchain infrastructure, confidentiality moves from optional feature to fundamental requirement. The protocols that provide transparency where it is needed for verification and privacy where it is needed for competition will capture the institutional market.

Fhenix is at the testnet stage, so validate costs and latency before planning a production deployment. Use the CoFHE Hardhat starter to prototype encrypted collateral, private debt balances, and auditor permissions on Sepolia.

Frequently Asked Questions

Q: How do borrower balances stay private?

A: CoFHE stores collateral and debt balances as ciphertext rather than public values. The contract uses FHE.allow or FHE.allowSender to grant specific addresses permission to view a balance. An authorized user then requests threshold decryption off-chain, while unauthorized users continue to see only ciphertext.

Q: Can regulators audit FHE lending positions?

A: Yes. A confidential lending protocol can grant an auditor access to selected positions without publishing every borrower's data. The contract assigns ciphertext-specific permission to the auditor's address, which functions like an audit key. Fhenix's Access Delegation demo demonstrates this selective-disclosure pattern.

Q: Is FHE lending composable with existing DeFi?

A: Mostly. CoFHE contracts run in the EVM and can call existing Solidity contracts. Plaintext protocols cannot directly interpret encrypted types, so integrations may require adapters or carefully defined decryption boundaries. Ethereum still exposes metadata such as contract addresses, transaction timing, and interaction patterns even when lending values remain encrypted.

Q: How does FHE compare to ZK for private lending?

A: FHE lets contracts compute directly on encrypted balances and reuse them across EVM logic, which fits lending that repeatedly evaluates shared encrypted state and needs selective disclosure to auditors. ZK proves a computation followed the rules without revealing inputs to the verifier, which fits proof-oriented flows with user-controlled private state. Many ZK systems also rely on elliptic-curve cryptography that is not quantum-safe, while FHE is lattice-based.

Q: Is FHE quantum-safe?

A: Yes. FHE is built on lattice-based cryptography, the same family as NIST's post-quantum standards, so FHE-encrypted data resists quantum attacks by construction, including data encrypted today.