# Price is Only the Symptom: Understanding and Analyzing Economic-State Manipulation Vulnerabilities in DeFi

Crypto.world

> Catalogues 276 DeFi attacks plus 91 audit findings into an economic-state-integrity taxonomy, pushing auditors beyond price checks toward accounting, AMM-state and timing invariants.

**arXiv**: 2611.00587

**Published**: 2026-10-31

**Authors**: Hao Wu, Haijun Wang, Xiapu Luo, Ming Fan, Ting Liu

## AI summary

## Overview

**Research area:** Blockchain and DeFi security — specifically vulnerability analysis and detection for economically motivated attacks on smart contracts.

**Technical level:** Intermediate (the paper's taxonomy and empirical results are accessible, but the formal definition of ESM uses trace-level notation and the detection-tool discussion assumes familiarity with taint analysis and transaction-trace reconstruction).

**Scope:** A large-scale empirical study of Economic-State Manipulation (ESM) vulnerabilities in DeFi, built on 276 real-world attacks and 91 Code4rena audit findings, that produces a cause-oriented taxonomy and benchmarks seven existing detection tools against it.

## What This Paper Is About

DeFi attacks are usually labeled "price manipulation," but the authors argue that price distortion is often just the visible symptom of a deeper failure: a protocol trusted an economically meaningful state (a reserve balance, an exchange rate, an accounting ratio, a settlement timing assumption) that an attacker could influence. The paper redefines this broader class as Economic-State Manipulation (ESM), builds a taxonomy of its root causes from real incidents, and then uses that taxonomy to test whether existing detection tools and their evaluation datasets actually capture the diversity of the problem.

## Key Contributions

1. **A curated ESM dataset.** 276 real-world attack incidents (177 collected from DeFiHackLabs plus 99 added from six existing PM/ESM datasets after deduplication and manual validation, spanning October 2020 to March 2026) and 91 ESM-related findings drawn from 1,283 high-severity Code4rena audit findings. The combined case set is described as 367 real-world ESM cases.
2. **A cause-oriented taxonomy.** Four primary categories and eleven subcategories describing which economic state is manipulated and which integrity assumption is violated, rather than categorizing by observable symptom or attack technique.
3. **A representativeness audit of prior benchmarks.** A mapping of the evaluation datasets of six existing PM/ESM studies onto the new taxonomy, showing that they cover 6–10 of the 11 subcategories but concentrate heavily on a small number of manipulation mechanisms.
4. **An empirical evaluation of seven detection approaches.** A taxonomy-guided false-negative and false-positive analysis identifying why detection fails: reliance on predefined attack patterns, insufficient reasoning about economic-state semantics, and incomplete exploitability validation.

## Main Findings

- **Price manipulation is a subset, not the whole class.** The taxonomy's four categories — Price-Source Trust, Protocol-Internal Accounting, AMM and Liquidity State Integrity, and Execution-Time Protection — show that ESM vulnerabilities frequently involve violations of protocol-internal accounting, AMM state consistency, and execution-time constraints rather than explicit price feeds.
- **Economic losses are large and persistent.** The 276 collected real-world incidents represent approximately $524 million USD in aggregate losses. The paper separately cites about $611.3 million in smart-contract-related losses during 2025 (Certik, 2025), and its top-10 incident table for 2025–2026 includes GMX at 42M on Arbitrum (2025-07-09) and ResupplyFi at 9.6M on ETH (2025-06-26).
- **The eleven subcategories are:** A1 Unsafe Direct Market-State Pricing, A2 Oracle Construction or Validation Defect, A3 Static or Stale Reference Dependence; B1 Claim-to-Backing Conversion Violation, B2 Composite or Position Misvaluation, B3 Entitlement or Obligation Misattribution; C1 Unsafe Token-side AMM State Mutation, C2 LP-State Integrity Violation, C3 Unsafe Balance-Reserve Settlement; D1 Ineffective Execution Output Constraint, D2 Premature Consumption of Unsettled State.
- **Existing benchmarks look diverse but are not.** Datasets from six prior studies cover 6–10 of the 11 subcategories, yet their case distributions are dominated by a limited set of manipulation mechanisms. The authors note that having diverse instances in a benchmark does not mean the corresponding detector can recognize their underlying root causes.
- **Detection coverage is poor.** Across seven representative approaches, the best-performing tool identifies only 50.4% of the real-world attacks evaluated.
- **Three detection paradigms each have structural blind spots.** Transaction-dependent methods (DeFiRanger, DeFort, DeFiScope) depend on observable execution evidence and known attack behaviors; bytecode analysis methods (DeFiTainter, SmartCat) are constrained by predefined patterns, taint abstraction, and candidate-generation mechanisms; LLM-assisted source code methods (GPTScan, PMDetector) still lack systematic reasoning about which state is manipulated, which assumption is violated, and how the violation propagates to a security-critical operation.
- **False positives stem mainly from incomplete exploitability validation**, such as failing to determine whether a suspicious state is actually attacker-controllable.
- **The XDK incident (February 16, 2026) illustrates the reframing.** A `recycle` hook removed XDK from a live XDK/GPC pool without burning LP tokens or proportionally withdrawing GPC, letting an attacker cycle `transfer`–`recycle`–`sync`–`skim` to drain reserves. The price distortion was a downstream consequence of a broken balance–reserve relationship, not a manipulated oracle.
- **Prior work had a narrow base.** The closest related taxonomy effort, DeFiRanger, analyzed 41 price manipulation attacks collected before March 2023; DeFiScope's dataset contains 95 PM attacks.

## Methodology in Plain English

The authors combined two complementary evidence sources. First, they took all 754 publicly documented incidents in DeFiHackLabs up to March 10, 2026, filtered out ones whose root cause clearly belonged to other vulnerability classes (such as reentrancy or access control), and had two researchers independently review descriptions and proof-of-concept scripts, inspecting transaction traces with Phalcon where needed. They merged in incidents from six published datasets and deduplicated, ending with 276 real-world incidents. Second, they reviewed 1,283 high-severity Code4rena audit findings (using findings curated by Zhang et al. for reports before 2023, then reviewing reports from January 2023 to March 2026) to capture vulnerabilities caught before exploitation, yielding 91 ESM-related findings.

To build the taxonomy, they used open card sorting. Each case got a card containing basic incident information, a report-derived description of the vulnerable behavior, and a root-cause analysis identifying the affected economic state, the violated integrity assumption, and the manipulation mechanism. Two researchers independently grouped the cards and labeled the groups, then reconciled their groupings iteratively; a third researcher with industry experience had to confirm any new category and adjudicated disagreements.

For tool selection, they reviewed top-tier security and software engineering venues and applied two filters: the tool had to be explicitly designed for PM detection or PM-related analysis (excluding exploit-generation tools like FlashSyn), and it had to have public source code (excluding general analyzers like Slither, Oyente, and Mythril). Seven tools qualified. The authors then mapped prior datasets onto their taxonomy and ran the seven tools against their own dataset, analyzing failures through the lens of the taxonomy. Formally, they define an ESM vulnerability as existing when an adversarial trace causes an economically sensitive operation to commit while an economic-integrity condition is violated, and the outcome materially deviates from an integrity-preserving counterfactual execution.

## Why This Matters

**Impact on research.** The paper challenges the framing that has organized a whole subfield. By showing that benchmark coverage is uneven and that the best detector reaches only 50.4% of real-world attacks, it argues that progress requires reasoning about economic-state semantics and violated assumptions rather than pattern matching on symptoms. The formal definition and the ⟨V, A, S, M⟩ tuple give follow-up work a vocabulary for expressing root causes.

**Real-world applications:**

- **Protocol design review:** developers can check their contracts against the eleven subcategories, particularly the non-price ones (B1–B3, C1–C3, D1–D2) that price-focused reviews tend to miss.
- **Audit scoping:** auditors can use the taxonomy to structure Code4rena-style reviews and to justify findings that do not manifest as obvious price anomalies.
- **Detection tool engineering:** builders of static analyzers and LLM-assisted scanners can target the identified failure modes, especially exploitability validation to cut false positives.
- **Incident classification:** post-mortem analysts can distinguish reserve-state manipulation from oracle tampering instead of filing both under "price manipulation."

**Industry relevance.** The incidents catalogued run through March 2026 and include losses from 105.5K to 42M across BSC, BASE, ETH, and Arbitrum, showing that ESM attacks affect multiple chains and protocol types. The released dataset and evaluation code (referenced as the paper's artifact) give practitioners a shared benchmark.

## Future Directions

- **Root-cause-level detection.** Designing techniques that explicitly model which economic state is consumed by which security-critical sink and which integrity assumption is required, instead of matching predefined attack patterns.
- **Exploitability validation.** Adding reasoning about attacker-controllability of suspicious states to reduce the false positives that dominate proactive approaches.
- **Broader, balanced benchmarks.** Expanding evaluation datasets beyond the mechanisms that currently dominate, so that coverage of 6–10 subcategories does not mask heavy concentration in a few.
- **Extending beyond price.** Systematically testing detection on the Execution-Time Protection and Protocol-Internal Accounting categories, where the paper shows price-centric tools have the least grounding.

## Target Audience

Smart contract security researchers and detector-tool developers will get the most from this paper, followed by DeFi protocol engineers and auditors who want a structured checklist of economic-state failure modes, and by graduate students entering blockchain security who need a current empirical picture of the attack landscape and the limits of existing defenses. Readers should be comfortable with AMM mechanics and basic static-analysis concepts; the taxonomy itself is readable without the formal definitions.

## Authors’ abstract

Economic-state manipulation has emerged as a significant security threat to decentralized finance (DeFi), causing substantial financial losses. Existing studies predominantly characterize these attacks as price manipulation, focusing on observable price distortions while overlooking underlying violations of economic-state integrity. We conduct a comprehensive empirical study of Economic-State Manipulation (ESM) vulnerabilities in DeFi. We first curate a diverse dataset…

[The full abstract on arXiv](https://arxiv.org/abs/2611.00587)

[Read the original paper](https://arxiv.org/abs/2611.00587)

---

Canonical: https://crypto.world/paper/price-is-only-the-symptom-understanding-and-analyzing-economic-state-2611.00587
