npy-oom-dos-poc

ericblackgachara/npy-oom-dos-poc

npy-oom-dos-poc is a machine learning model from ericblackgachara. Use it for the machine learning task on the model card, and read the license before you ship it in a product.

Status: PoC CONFIRMED ✅ — Ready for huntr submission

Downloads · 30 days

Access

Public

Updated May 9, 2026

Repo size

135 KB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.pdf135 KB · 94%

At a glance

Access: Public
Created: May 9, 2026
Updated: May 9, 2026
SHA: 5715bb61

Model card

From the Hugging Face model README

NumPy — OOM DoS via Oversized Shape in .npy Header

Status: PoC CONFIRMED ✅ — Ready for huntr submission

Field	Value
Target	numpy (huntr.com — numpy/numpy)
Format	NumPy `.npy` / `.npz` array files
Vuln	CWE-789 — Unbounded Memory Allocation → MemoryError DoS
CVSS	7.5 High
Trigger	`numpy.load("bomb.npy")` — no flags required
Effect	`MemoryError: Unable to allocate 8.00 TiB for array with shape (1099511627776,)`

Quick Repro

python3 poc/make_poc.py    # creates bomb.npy (128 bytes)
python3 poc/trigger.py     # MemoryError → crash

Files

poc/make_poc.py — builds bomb.npy (128 bytes)
poc/trigger.py — triggers crash
poc/bomb.npy — malicious array file
report.md — huntr submission report
evidence/poc-run.html — self-contained HTML evidence

Key Source Location

numpy/lib/_format_impl.py — read_array() ~line 1089
count = numpy.multiply.reduce(shape, dtype=numpy.int64) — no upper bound on count
array = numpy.ndarray(count, dtype=dtype) — allocates before reading any data

Attack Summary

128-byte .npy file with header declaring shape (2^40,) float64. numpy.load() computes count=1,099,511,627,776, allocates before reading data → 8 TiB → MemoryError. Same path applies to .npz files (ZIP of .npy arrays).

Created: May 9, 2026
Updated: May 9, 2026