Conference paper · 2024 · Open access

LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks

Saad Ullah, Mingji Han, Saurabh Pujar, Hammond A. Pearce, Ayse Kivilcim Coskun, Gianluca Stringhini

IEEE Symposium on Security and Privacy (S&P), 862–880 · Published 19 May 2024 · DOI 10.1109/sp54263.2024.00210

Open PDFView at the publisherCite this paper

In brief

“LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks” is a 2024 conference paper by Saad Ullah and 5 co-authors, published in IEEE Symposium on Security and Privacy (S&P). OpenAlex counts 365 citations as of 19 November 2026, which puts it in the top 8% of works published in 2024 by citations. It is open access: the full text is free to read at arXiv (Cornell University).

Citations
365
Most cited in
2026 (205)
Field-weighted impact
60.59
References
42
Authors
6

Abstract

Large Language Models (LLMs) have been suggested for use in automated vulnerability repair, but benchmarks showing they can consistently identify security-related bugs are lacking. We thus develop SecLLMHolmes, a fully automated evaluation framework that performs the most detailed investigation to date on whether LLMs can reliably identify and reason about security-related bugs. We construct a set of 228 code scenarios and analyze eight of the most capable LLMs across eight different investigative dimensions using our framework. Our evaluation shows LLMs provide non-deterministic responses, incorrect and unfaithful reasoning, and perform poorly in real-world scenarios. Most importantly, our findings reveal significant non-robustness in even the most advanced models like ‘PaLM2’ and ‘GPT-4’: by merely changing function or variable names, or by the addition of library functions in the source code, these models can yield incorrect answers in 26% and 17% of cases, respectively. These findings demonstrate that further LLM advances are needed before LLMs can be used as general purpose security assistants.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Ullah, S., Han, M., Pujar, S., Pearce, H. A., Coskun, A. K., & Stringhini, G. (2024). LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks. IEEE Symposium on Security and Privacy (S&P), 862–880. https://doi.org/10.1109/sp54263.2024.00210

MLA 9

Ullah, Saad, et al. “LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks.” IEEE Symposium on Security and Privacy (S&P), 2024, pp. 862–880. https://doi.org/10.1109/sp54263.2024.00210.

Chicago

Ullah, Saad, Mingji Han, Saurabh Pujar, Hammond A. Pearce, Ayse Kivilcim Coskun, and Gianluca Stringhini. 2024. “LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks.” IEEE Symposium on Security and Privacy (S&P): 862–880. https://doi.org/10.1109/sp54263.2024.00210.

Harvard

Ullah, S. et al. (2024) ‘LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks’, IEEE Symposium on Security and Privacy (S&P), pp. 862–880. doi:10.1109/sp54263.2024.00210.

IEEE

S. Ullah, M. Han, S. Pujar, H. A. Pearce, A. K. Coskun, and G. Stringhini, “LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks,” IEEE Symposium on Security and Privacy (S&P), pp. 862–880, 2024, doi: 10.1109/sp54263.2024.00210.

BibTeX

@inproceedings{ullah2024llms,
  title     = {{LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks}},
  author    = {Ullah, Saad and Han, Mingji and Pujar, Saurabh and Pearce, Hammond A. and Coskun, Ayse Kivilcim and Stringhini, Gianluca},
  booktitle = {IEEE Symposium on Security and Privacy (S\&P)},
  year      = {2024},
  pages     = {862--880},
  doi       = {10.1109/sp54263.2024.00210},
  url       = {https://doi.org/10.1109/sp54263.2024.00210}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Citations by year

YearCitationsShare of the best year
202322
20243434
2025124124
2026205205

Its field-weighted citation impact is 60.59: it has been cited 60.59 times as often as the average work of the same field, type and year (1.00 is the world average).

Topic and keywords

Physical Sciences › Computer Science › Software › Software Testing and Debugging Techniques

  • large language models
  • security vulnerabilities
  • vulnerability detection
  • automated vulnerability repair
  • evaluation framework
  • benchmarking
  • reasoning errors
  • model robustness

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Ullah, S., Han, M., Pujar, S., Pearce, H. A., Coskun, A. K., & Stringhini, G. (2024). LLMs Cannot Reliably Identify and Reason About Security Vulnerabilities (Yet?): A Comprehensive Evaluation, Framework, and Benchmarks. IEEE Symposium on Security and Privacy (S&P), 862–880. https://doi.org/10.1109/sp54263.2024.00210

How many times has this paper been cited?

365 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at arXiv (Cornell University): sign in and open the PDF from this page.

Who wrote this paper?

Saad Ullah, Mingji Han, Saurabh Pujar, Hammond A. Pearce, Ayse Kivilcim Coskun, Gianluca Stringhini.

What is the DOI of this paper?

10.1109/sp54263.2024.00210, permanently at https://doi.org/10.1109/sp54263.2024.00210. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.