Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Contemporary Survey of Large Language Model Assisted Program Analysis

Jiayimei Wang, Tao Ni, Wei-Bin Lee, Qingchuan Zhao

TL;DR

The paper surveys how Large Language Models are applied to program analysis, tackling static, dynamic, and hybrid approaches to vulnerability and malware detection, program verification, and test generation. It highlights representative methods, such as LLift, IRIS, CoqPilot, LaM4Inv, CHEMFuzz, PentestGPT, and AutoSafeCoder, to illustrate how LLMs augment traditional static/dynamic analyses and enable more scalable, explainable reasoning. Major contributions include a comprehensive taxonomy, cross-domain synthesis of LLM-assisted techniques, and a discussion of challenges like token limits, non-determinism, and hallucinations, along with future directions emphasizing deeper integration with formal methods and hybrid analysis frameworks. The work aims to guide security researchers and developers in deploying domain-specific LLMs to improve vulnerability detection, verification, and automated testing in real-world software systems, while outlining practical constraints and research opportunities in next-generation program-analysis tools.

Abstract

The increasing complexity of software systems has driven significant advancements in program analysis, as traditional methods unable to meet the demands of modern software development. To address these limitations, deep learning techniques, particularly Large Language Models (LLMs), have gained attention due to their context-aware capabilities in code comprehension. Recognizing the potential of LLMs, researchers have extensively explored their application in program analysis since their introduction. Despite existing surveys on LLM applications in cybersecurity, comprehensive reviews specifically addressing their role in program analysis remain scarce. In this survey, we systematically review the application of LLMs in program analysis, categorizing the existing work into static analysis, dynamic analysis, and hybrid approaches. Moreover, by examining and synthesizing recent studies, we identify future directions and challenges in the field. This survey aims to demonstrate the potential of LLMs in advancing program analysis practices and offer actionable insights for security researchers seeking to enhance detection frameworks or develop domain-specific models.

A Contemporary Survey of Large Language Model Assisted Program Analysis

TL;DR

The paper surveys how Large Language Models are applied to program analysis, tackling static, dynamic, and hybrid approaches to vulnerability and malware detection, program verification, and test generation. It highlights representative methods, such as LLift, IRIS, CoqPilot, LaM4Inv, CHEMFuzz, PentestGPT, and AutoSafeCoder, to illustrate how LLMs augment traditional static/dynamic analyses and enable more scalable, explainable reasoning. Major contributions include a comprehensive taxonomy, cross-domain synthesis of LLM-assisted techniques, and a discussion of challenges like token limits, non-determinism, and hallucinations, along with future directions emphasizing deeper integration with formal methods and hybrid analysis frameworks. The work aims to guide security researchers and developers in deploying domain-specific LLMs to improve vulnerability detection, verification, and automated testing in real-world software systems, while outlining practical constraints and research opportunities in next-generation program-analysis tools.

Abstract

The increasing complexity of software systems has driven significant advancements in program analysis, as traditional methods unable to meet the demands of modern software development. To address these limitations, deep learning techniques, particularly Large Language Models (LLMs), have gained attention due to their context-aware capabilities in code comprehension. Recognizing the potential of LLMs, researchers have extensively explored their application in program analysis since their introduction. Despite existing surveys on LLM applications in cybersecurity, comprehensive reviews specifically addressing their role in program analysis remain scarce. In this survey, we systematically review the application of LLMs in program analysis, categorizing the existing work into static analysis, dynamic analysis, and hybrid approaches. Moreover, by examining and synthesizing recent studies, we identify future directions and challenges in the field. This survey aims to demonstrate the potential of LLMs in advancing program analysis practices and offer actionable insights for security researchers seeking to enhance detection frameworks or develop domain-specific models.

Paper Structure

This paper contains 20 sections, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Program Analysis
  4. Large Language Models
  5. LLM for Static Analysis
  6. LLM for Vulnerability Detection
  7. LLM for Malware Detection
  8. LLM for Program Verification
  9. LLM for Static Analysis Enhancement
  10. LLM for Dynamic Analysis
  11. LLM for Malware Detection
  12. LLM for Fuzzing
  13. LLM for Penetration Testing
  14. LLM for Hybrid Approach
  15. LLM for Unit Test Generation
  16. ...and 5 more sections

Figures (7)

  • Figure 1: Taxonomy of the survey.
  • Figure 2: Static analysis workflow.
  • Figure 3: Dynamic analysis workflow.
  • Figure 4: A diagram of LLMs’ application in vulnerability detection.
  • Figure 5: Fuzzing classifications.
  • ...and 2 more figures