haomin-yu/cve-2024-13176-timing-analysisInstrumented analysis and reproducibility materials for ECDSA timing leakage in OpenSSL CVE-2024-13176
This repository contains the research report, website, data, source material, and reproducibility files for an independent study of CVE-2024-13176.
The project looks at whether the nonce-related timing condition reported for OpenSSL ECDSA signing can be detected in a controlled ARM64 virtualized Linux environment.
The confirmatory experiment compared OpenSSL 3.4.0 with OpenSSL 3.4.1 using ECDSA P-521 and source-level instrumentation.
The final dataset contains 600,000 signing observations across 30 paired collection blocks.
At the primary internal timing point:
p = 1.86265 × 10^-9The strongest result is localized to the downstream Montgomery multiplication
measured by kinv_mul_ns. Broader nonce inversion and full signature timing
measurements did not show the same stable separation.
The study focuses on detectability at this internal measurement point. It does not claim practical private key recovery or remote visibility of the timing effect.
.
├── README.md
├── docs/
├── report/
└── reproducibility/
├── README.md
├── data/
├── derived/
├── figures/
├── plans/
├── records/
├── scripts/
└── source/
docs/Contains the GitHub Pages version of the research report.
report/Contains the PDF version of the research report.
reproducibility/Contains the public research materials, including the raw confirmatory data, analysis scripts, source changes, derived outputs, figures, plans, and supporting records.
See reproducibility/README.md for instructions on
rerunning the analysis and understanding the data layout.
The raw confirmatory CSV files are included in the repository, so reproducing the statistical analysis does not require collecting new timing measurements.
Start with:
cd reproducibility
Then follow the instructions in
reproducibility/README.md.
The publication package includes:
10.5281/zenodo.22036547)The measurements were collected in one ARM64 Linux virtual machine using OpenSSL 3.4.0 and OpenSSL 3.4.1 with NIST P-521.
Results from another processor, operating system, compiler, or virtualization environment may differ. A useful replication should focus on whether the same condition can be classified and whether the vulnerable and patched releases show a similar difference in timing behavior.