Gated security proof-of-concept
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This repository contains a coordinated-disclosure security proof-of-concept: a crafted Darknet model configuration that demonstrates a memory-safety defect in the framework's convolutional weight allocation / forward path. It contains no user data, no weaponized exploit, and no network activity. Access is restricted to the affected maintainers and the huntr / ProtectAI triage team for verification and remediation only.
Log in or Sign Up to review the conditions and access this model content.
Darknet .cfg convolutional path — crafted-file memory-safety PoC (gated)
This is a security research proof-of-concept submitted to the huntr Model File Formats
program (format darknet). It contains a crafted Darknet .cfg that triggers a heap
out-of-bounds read in the maintained Darknet/YOLO fork (hank-ai/darknet) during a normal
network load + forward pass.
The payload only crashes the process to demonstrate the defect. There is no exploit weaponization, no user data, and no network callback.
- Class: integer overflow → under-allocation → heap out-of-bounds read (CWE-190 / CWE-125)
- Affected:
hank-ai/darknetconvolutional layer weight allocation vs gemm forward - Disclosure: coordinated via huntr MFF
Full technical details, root cause, ASAN + stock-build crash logs, the crafted .cfg, and the
reproducible harness notes are in the gated files. Please request access if you are triaging
this report.
Contact: security researcher mrw0r57 (huntr: ssjcorpsec).