PaperlayerView exact revision on arXiv

Source-linked AI summary

A Systematization of Knowledge on DeFi Vaults: Architectures, Curation Mechanisms, and Strategy Design

Davide Mancino, Luca Pennella

arXiv:2610.01300v1cs.CRcs.CEcs.CY

TL;DR

DeFi vaults have gained complex strategies and curator control, but the field lacks a unified framework for their architectures, accountability, and risks. This paper develops a system model and three taxonomies, then maps production protocols to them. Its analysis highlights adapter-boundary attack surfaces, uneven curator accountability, and operational dependencies as recurring risk factors.

  • Problem

    Vault complexity and curator control create principal–agent dynamics, while the lack of a shared taxonomy leaves users and designers without a common framework for reasoning about security, composability, incentives, and measurable risk.

    Source: §1 · p. 1

  • Method

    The paper provides a unified system model and three taxonomies for vault exposures, curator systems, and strategy design and failure modes, then maps a representative set of production protocols.

    Source: §1 · p. 2

  • Results

    The analysis identifies adapter-boundary attack-surface expansion, heterogeneous and often weakly enforceable curator accountability, and operational dependencies as recurring features of vault risk.

    Source: §10 · p. 11

  • Takeaways & Limitations

    The paper organizes vault failure modes into mitigations and on-chain monitoring proxies across accounting, oracle, MEV, keeper, and privileged-role risks.

    Source: §10 · p. 10

    Reproducible measurement of depositor-level MEV externalities remains an open challenge given the diversity of harvest, rebalance, and bridge operations.

    Source: §10 · p. 11

Abstract

from arXiv · show

Decentralized finance (DeFi) vaults are smart-contract-based asset management systems that pool deposits, execute programmable strategies, and mint tokenized shares representing claims on underlying assets and strategy performance. As vault designs have evolved from early yield aggregators to modular, actively managed systems, a new control layer, curation, has emerged to select strategies, configure risk parameters, and coordinate operational execution, introducing principal-agent dynamics and new failure modes. This paper systematizes DeFi vault architectures and curator-mediated control planes through (i) a unified system model and formal definitions for share accounting, roles, and operational dependencies, and (ii) three complementary taxonomies covering vault exposures and objectives, curator governance and accountability mechanisms, and strategy execution patterns together with their failure modes. We further map a representative set of production protocols to the proposed dimensions. The frameworks in this work aim to support rigorous analysis and safer design of blockchain-based financial applications.

1 Introduction

DeFi vaults automate pooled asset management, but increasingly complex strategies and curator control introduce trust, liveness, and principal–agent concerns. The paper addresses the lack of a unified framework with a system model, three taxonomies, and a representative protocol mapping.

1 Introduction

  • Vaults automate capital allocation and compounding, while privileged strategy configuration and execution introduce additional trust and liveness assumptions.Curation adds principal–agent dynamics: curators control deployment decisions while depositors bear downside risk and may have limited exposure transparency.Source: §1 · p. 1
  • Earlier work covers yield aggregators, broad DeFi, derivatives, bridges, and smart-contract vulnerabilities, but does not provide this vault-specific unified framework.The cited yield-aggregator study predates curator systems, ERC-4626 standardization, and newer vault types; other surveys and systematizations focus on broader DeFi or different primitives [16].Source: §1 · p. 2
  • The systematization organizes its analysis around vault exposure, curator governance, and strategy execution and failure modes.These questions address classification by exposure, leverage, and outcome; curator scope, fees, and accountability; and strategy dependencies, failures, and mitigations.Source: §1 · p. 2
  • The paper formalizes vault architectures and develops taxonomies for exposures, curator systems, and strategies, then maps representative production protocols to those dimensions.The frameworks identify recurring patterns, design tradeoffs, observable on-chain indicators, and vault-relevant failure modes.Source: §1 · p. 2
  • The scope is on-chain vault systems with at least partially observable deposits, share accounting, and strategy interactions, excluding purely off-chain products and centralized yield programs.Partially off-chain keepers, oracle operations, and governance remain important threat-modeling concerns.Source: §1 · p. 2

2 Methodology

The study combines literature review, protocol-documentation analysis, and security-incident evidence. It builds a representative, rather than exhaustive, protocol corpus and derives its taxonomies through iterative coding.

2 Methodology

  • The study reviews documentation, whitepapers, and audit materials for more than 20 production vault systems meeting pooled-capital, deposit/withdrawal-interface, and yield-strategy criteria.The corpus spans yield aggregators, lending and credit, liquidity management, options and structured products, liquid staking, and modular platforms; it is representative rather than exhaustive.Source: §2 · p. 2
  • The taxonomies derive from bottom-up feature coding followed by a validation pass, while the security analysis synthesizes audits, incident post-mortems, and prior security literature.The work also surveys prior research on yield aggregation, automated market makers, tokenized vault standards, and smart-contract security [16] [38].Source: §2 · pp. 2–3

3 Background and Definitions

The background defines vault systems and their accounting, strategies, roles, and risk concepts. It distinguishes the ERC-4626 interface from broader vault designs and describes how deposits, withdrawals, and fees are handled.

3 Background and Definitions

  • A vault system pools deposits, tracks proportional ownership through shares, allocates assets to strategies, and lets users redeem under a defined conversion rule.The term includes ERC-4626 vaults and multi-asset or modular architectures outside that specification.Source: §3 · p. 3
  • ERC-4626 standardizes single-asset deposit and withdrawal entrypoints and asset-share conversion queries, but not fees, strategy disclosure, or risk parameters.Deposits mint shares; withdrawals can be immediate, queued for position unwinding, or subject to lockups or cooldowns, with stress gating possible.Source: §3 · p. 3
  • Canonical share conversion makes deposits and redemptions proportional to vault assets and outstanding shares, while price per share can diverge temporarily from economic net asset value.Reporting latency or unrealized profit and loss can account for that temporary difference.Source: §3 · p. 3
  • Strategies deploy capital through external protocols to generate yield, manage risk, or provide liquidity, and may be embedded, adapter-based, or triggered by off-chain agents.Curators select strategies and configure risk parameters; allocators distribute capital across strategies; keepers submit transactions for actions such as rebalancing and compounding.Source: §3 · p. 3
  • Management fees are typically annualized against assets under management and often dilute holders through share inflation; performance fees apply to gains, often above a high-water mark.Performance fees may be realized through asset transfers or share minting, depending on implementation.Source: §3 · p. 3
  • Vault risks include contract, market, liquidity, oracle, governance, operational, cross-chain, and counterparty or legal risks.Examples include delayed keeper execution, manipulated price feeds, bridge liveness failures, and off-chain custodians or legal structures failing [16].Source: §3 · p. 3

4 System Model and Architectural Patterns

The system model represents a vault through its contract, users, assets, strategies, curators, and keepers, and specifies how deposits, redemptions, strategy actions, and fees interact. Architectural choices trade simplicity and composability against diversification, attribution complexity, and added boundaries.

4 System Model and Architectural Patterns

  • The system model represents a vault system as the vault contract, users, supported assets, strategy modules, curators, and keepers or automation agents.These components organize the account of vault operations and dependencies.Source: §4 · p. 4
  • Deposits mint shares in proportion to total assets and shares, while redemptions return the proportional assets immediately when liquid funds suffice or after unwinding or queuing.The model assigns strategy selection and parameter configuration to curators, transaction-triggered actions to keepers, and periodic or high-water-mark-based accrual to fees.Source: §4 · p. 4
  • Modular vault adapters allow strategies to be added, removed, or upgraded without redeploying the vault, but do not guarantee loss isolation and can widen adapter and allowance attack surfaces.Monolithic designs embed strategy logic in the vault, though proxy patterns may still allow evolution.Source: §4 · p. 4
  • Multi-strategy vaults can diversify exposure when correlations are controlled, but add allocator risk and more complex attribution; poor yield-only allocation can concentrate risk.Single-strategy vaults offer more straightforward attribution, with price per share primarily driven by that strategy.Source: §4 · p. 4
  • Global share accounting gives depositors uniform strategy exposure and supports simplicity and composability, whereas per-user accounting permits heterogeneous entry prices but sacrifices share fungibility.Global accounting is the ERC-4626 default; tranched accounting uses separate share classes or vault instances for different risk/return profiles.Source: §4 · p. 4

5 Taxonomy 1: Vault Types by Exposure

Vaults are classified by primary exposure, yield source, and target risk–return profile, but production vaults may combine multiple exposures.

Lending Vaults

  • Lending vaults earn borrower interest, while facing protocol vulnerabilities, utilization-related liquidity risk, and oracle-driven bad-debt risks.The passage cites the March 2023 Euler Finance exploit, with approximately $197 million in losses.Source: §5 · p. 4

Liquidity Provider (LP) Vaults

  • LP vaults earn trading fees and liquidity-mining rewards; concentrated positions can yield more than full-range positions but require active management.Risks include impermanent loss, MEV losses during rebalances, and inventory risk from range selection.Source: §5 · p. 4

Delta-Neutral and Options Vaults

  • Delta-neutral vaults combine positions such as long spot and short perpetuals to seek funding-rate yield, but volatility can trigger liquidation.Funding can turn negative, requiring reserve buffers; options vaults instead earn premiums but have asymmetric risks and may face illiquid markets.Source: §5 · p. 5

Liquid Staking and Restaking Vaults

  • Liquid-staking and restaking vaults layer rewards through LSTs and AVSs, compounding exposure to slashing, depegging, and multiple protocol dependencies.Restaking extends liquid staking by restaking LSTs through layers such as EigenLayer.Source: §5 · p. 5

Real-World Asset Vaults

  • RWA vaults invest in tokenized off-chain assets, exposing depositors to counterparty, legal, custody, valuation, and redemption risks.Custodian insolvency or fraud can result in total loss with limited recourse for on-chain depositors.Source: §5 · p. 5

Stablecoin Yield-Arbitrage Vaults

  • Stablecoin yield-arbitrage vaults exploit rate differences across lending markets, liquidity pools, or yield protocols for typically modest but stable yields.Depegs and gas costs can erode thin spreads.Source: §5 · p. 5

6 Taxonomy 2: Curator Systems and Curation Markets

The curator taxonomy distinguishes who selects or configures strategies, what they control, and how fees and accountability constrain their decisions.

Curator Types and Strategy Inclusion

  • Curator systems range from protocol-governance voting or multisig decisions to delegated, third-party, and permissionless curation, each with distinct tradeoffs and risks.Delegation can speed iteration but introduces cartel and collusion risks; permissionless curation relies on robust risk-scoring infrastructure.Source: §6 · p. 6

Strategy Inclusion

  • Strategy-inclusion policies may permit open proposals, require curator whitelisting, or combine open submission with curator vetoes or risk-adjusted caps.The inclusion policy directly shapes the vault’s risk surface and trust assumptions.Source: §6 · p. 6

Scope of Curation

  • Curators may control strategy and venue selection, collateral and LTV settings, risk limits, and approved oracle feeds.These decision rights determine the scope of curation.Source: §6 · p. 6

Fee Structures and Economic Incentives

  • Fee mechanisms include management and performance fees, curator take rates, referral fees, and token emissions, with incentives and risks varying by structure.Performance fees can align incentives with user returns but may amplify short-horizon risk-taking; emissions can support advertised APYs that compress as emissions decline.Source: §6 · p. 6

Custody and Emergency Controls

  • Privileged-control designs range from minimal-control vaults to admin, timelock, and guardian models, which trade incident response powers against user exit options.Guardian powers can be limited to pausing the vault or halting strategies, while timelocks give users time to exit before changes take effect.Source: §6 · p. 6

Accountability and Principal-Agent Dynamics

  • Principal-agent risks include volatile fee-seeking strategies, affiliated allocations, delayed disclosure, and curator concentration, with on-chain proxies for monitoring.Proxies include PPS volatility, drawdowns, affiliated-strategy share, disclosure lag, and HHI or top-k TVL shares; accountability tools include reputation, slashing, and insurance.Source: §6 · p. 6

7 Taxonomy 3: Strategy Design and Execution

The strategy taxonomy organizes designs by execution model, trigger, capital allocation, reinvestment, hedging, and liquidity management.

Strategy Execution Models

  • Execution models range from static holdings and deterministic rule-based actions to adaptive strategies, bounded keeper discretion, and MEV-aware execution.Adaptive strategies may require off-chain computation and multiple oracle feeds; MEV-aware execution can augment any model with private submission, batching, and slippage bounds.Source: §7 · p. 7

Rebalancing Triggers

  • Rebalancing can be time-based, threshold-based, oracle-conditioned, keeper-driven, or governance-triggered, with tradeoffs in timing, monitoring, liveness, and latency.Threshold triggers require continuous keeper monitoring, while governance-based changes introduce latency and human-error risk.Source: §7 · p. 7

Capital Allocation Patterns

  • Capital allocation ranges from single-venue concentration and fixed multi-venue splits to dynamic multi-venue and cross-chain deployment.Dynamic and cross-chain allocation can access opportunities or capture higher returns, but transaction costs, slippage, MEV, bridge risk, and operational complexity can offset gains.Source: §7 · pp. 7–8

Reinvestment, Hedging, and Liquidity Management

  • Strategies may auto-compound, rely on manual harvesting, or distribute rewards without reinvestment, while hedges and redemption mechanisms shape exposure and liquidity.Liquidity options include withdrawal queues, cash buffers, partial withdrawals, and redemption in kind; buffers enable immediate withdrawals at the cost of capital efficiency.Source: §7 · p. 8

8 Security, Risk, and Failure Modes

Vault risks span vulnerabilities in share accounting and user actions, strategy-level exposures, and curator or governance behavior; the section maps these risks to mitigations and indicators.

Risk Framework

  • The risk framework separates vault-level vulnerabilities affecting accounting and user funds from strategy-level risks affecting deployed capital.The section maps failure modes to mitigation controls and observable on-chain indicators.Source: §8 · p. 8

Share Accounting Vulnerabilities

  • Reentrancy can exploit external calls made before accounting updates, while direct asset donations can manipulate PPS, especially when share supply is small.Mitigations include checks-effects-interactions, reentrancy guards, initial seeding or virtual share offsets, and ERC-4626-aligned conversion logic.Source: §8 · pp. 8–9

Rounding and Execution Attacks

  • Integer-rounding leakage can arise through repeated small operations, while price-sensitive deposits or withdrawals can be sandwiched by MEV bots.Suggested controls include consistent high-precision rounding, slippage bounds, batching or delayed execution, and private transaction submission.Source: §8 · p. 9

PPS Accounting Drift and Insolvency

  • Unrealized strategy gains or losses can make PPS diverge from true NAV, and extreme losses can leave assets below liabilities and prevent full redemption.Mitigations include regular NAV updates, conservative mark-to-market, strategy risk limits, stop-losses, tranche structures, and insurance.Source: §8 · p. 9

Strategy-Level Risks

  • Strategy-level risks include liquidation cascades, depegs, bridge exploits, MEV extraction, keeper failure, and admin-key compromise.Controls cited include conservative LTV, peg monitoring, bridge exposure limits, private submission, keeper redundancy, multisigs, and timelocks.Source: §8 · p. 9

Governance and Curator Risks

  • Governance risks include harmful parameter changes, self-serving curator allocations, and delayed incident disclosure, with timelocks, bounded parameters, reputation, bonds, and reporting as mitigations.The lag between public incident reports and vault exit actions can quantify information asymmetry.Source: §8 · p. 9

9 Protocol Landscape

The protocol landscape maps 22 production vault systems across taxonomy dimensions, showing varied control roles and growing reliance on off-chain coordination and composability. A separate table maps vault failure modes to mitigations and on-chain indicators.

9 Protocol Landscape

  • The 22-protocol snapshot shows modular architectures, heterogeneous control roles, and stronger off-chain and cross-protocol dependencies shaping the ecosystem.The mapping spans category, control model, chain coverage, and salient design features, blurring boundaries between asset management, market-neutral structuring, and protocol infrastructure.Source: §9 · p. 9
  • Table 4 maps vault failure modes to mitigation controls and observable on-chain indicators.Source: Table 4 · p. 10
  • Table 5 catalogs protocols by taxonomy classifications; its control-model labels include governance, strategist, manager, delegate, backer, operator, and configurable-parameter roles.The caption identifies the catalog as a representative snapshot dated 01/2026.Source: Table 5 · p. 10

10 Conclusion

The conclusion presents a unified view of vault architectures, three taxonomies, a 22-protocol mapping, and an analysis of failure modes. It highlights design and accountability patterns while identifying measurement, disclosure, and monitoring needs for future work.

10 Conclusion

  • The systematization defines vault architectures, presents three taxonomies, maps 22 production protocols, and organizes failure modes into mitigations and monitoring proxies.The failure modes include accounting exploits, oracle-conditioned execution, MEV-sensitive operations, keeper liveness, and privileged-role failures.Source: §10 · p. 10
  • The authors identify adapter-boundary attack surface, weakly enforceable curator accountability, and operational dependencies as recurring risk patterns in actively managed strategies.Source: §10 · p. 11
  • Open needs include verifiable curator disclosures, reproducible depositor-MEV measurement, cross-chain accountability, aligned fee incentives, and composability-aware systemic-risk monitoring.Source: §10 · p. 11
Loading 2610.01300v1…