Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM-Powered Detection of Price Manipulation in DeFi

Lu Liu, Wuqi Zhang, Lili Wei, Hao Guan, Yongqiang Tian, Yepang Liu

TL;DR

PMDetector addresses the proactive detection of price manipulation in DeFi by fusing static inter-procedural taint analysis with a two-stage, domain-aware LLM reasoning pipeline and a final semantic sanity checker. Guided by a formal attack model, it identifies high-risk code paths, prunes infeasible ones, and validates exploitable scenarios before reporting. On a benchmark of 73 vulnerable and 288 benign contracts, PMDetector achieves up to 88% recall and 100% precision (depending on the LLM backend) while reducing audit costs to about $0.03 per vulnerability and processing times around 4 seconds per audit. The framework significantly outperforms state-of-the-art static and LLM-based baselines, offering a scalable, cost-effective solution for securing DeFi contracts against complex economic exploits.

Abstract

Decentralized Finance (DeFi) smart contracts manage billions of dollars, making them a prime target for exploits. Price manipulation vulnerabilities, often via flash loans, are a devastating class of attacks causing significant financial losses. Existing detection methods are limited. Reactive approaches analyze attacks only after they occur, while proactive static analysis tools rely on rigid, predefined heuristics, limiting adaptability. Both depend on known attack patterns, failing to identify novel variants or comprehend complex economic logic. We propose PMDetector, a hybrid framework combining static analysis with Large Language Model (LLM)-based reasoning to proactively detect price manipulation vulnerabilities. Our approach uses a formal attack model and a three-stage pipeline. First, static taint analysis identifies potentially vulnerable code paths. Second, a two-stage LLM process filters paths by analyzing defenses and then simulates attacks to evaluate exploitability. Finally, a static analysis checker validates LLM results, retaining only high-risk paths and generating comprehensive vulnerability reports. To evaluate its effectiveness, we built a dataset of 73 real-world vulnerable and 288 benign DeFi protocols. Results show PMDetector achieves 88% precision and 90% recall with Gemini 2.5-flash, significantly outperforming state-of-the-art static analysis and LLM-based approaches. Auditing a vulnerability with PMDetector costs just $0.03 and takes 4.0 seconds with GPT-4.1, offering an efficient and cost-effective alternative to manual audits.

LLM-Powered Detection of Price Manipulation in DeFi

TL;DR

PMDetector addresses the proactive detection of price manipulation in DeFi by fusing static inter-procedural taint analysis with a two-stage, domain-aware LLM reasoning pipeline and a final semantic sanity checker. Guided by a formal attack model, it identifies high-risk code paths, prunes infeasible ones, and validates exploitable scenarios before reporting. On a benchmark of 73 vulnerable and 288 benign contracts, PMDetector achieves up to 88% recall and 100% precision (depending on the LLM backend) while reducing audit costs to about $0.03 per vulnerability and processing times around 4 seconds per audit. The framework significantly outperforms state-of-the-art static and LLM-based baselines, offering a scalable, cost-effective solution for securing DeFi contracts against complex economic exploits.

Abstract

Decentralized Finance (DeFi) smart contracts manage billions of dollars, making them a prime target for exploits. Price manipulation vulnerabilities, often via flash loans, are a devastating class of attacks causing significant financial losses. Existing detection methods are limited. Reactive approaches analyze attacks only after they occur, while proactive static analysis tools rely on rigid, predefined heuristics, limiting adaptability. Both depend on known attack patterns, failing to identify novel variants or comprehend complex economic logic. We propose PMDetector, a hybrid framework combining static analysis with Large Language Model (LLM)-based reasoning to proactively detect price manipulation vulnerabilities. Our approach uses a formal attack model and a three-stage pipeline. First, static taint analysis identifies potentially vulnerable code paths. Second, a two-stage LLM process filters paths by analyzing defenses and then simulates attacks to evaluate exploitability. Finally, a static analysis checker validates LLM results, retaining only high-risk paths and generating comprehensive vulnerability reports. To evaluate its effectiveness, we built a dataset of 73 real-world vulnerable and 288 benign DeFi protocols. Results show PMDetector achieves 88% precision and 90% recall with Gemini 2.5-flash, significantly outperforming state-of-the-art static analysis and LLM-based approaches. Auditing a vulnerability with PMDetector costs just $0.03 and takes 4.0 seconds with GPT-4.1, offering an efficient and cost-effective alternative to manual audits.
Paper Structure (31 sections, 5 equations, 7 figures, 4 tables, 1 algorithm)

This paper contains 31 sections, 5 equations, 7 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Automated Market Makers and Price Discovery in DeFi Ecosystem
  4. Price Manipulation Vulnerabilities in DeFi Protocols
  5. Motivation and Attack Model
  6. Illustrating Example
  7. Attack Model
  8. Methodology
  9. Static Taint Analysis
  10. Taint Sources
  11. Taint Propagation Rules
  12. Taint Sinks
  13. Taint Analysis Algorithm
  14. Path Grouping and Abstraction
  15. LLM-based Reasoning
  16. ...and 16 more sections

Figures (7)

  • Figure 1: Vulnerable logic in the ZZF protocol.
  • Figure 2: Demonstration of the Attack on the ZZF Protocol
  • Figure 3: The workflow of PMDetector.
  • Figure 4: Definition of taint sinks.
  • Figure 5: Prompt Templates of the Path Filtering Stage (left) and the Attack Simulation Stage (right).
  • ...and 2 more figures