CVE-2026-72852
NVD analysis in progress
hank-ai/darknet sizes a convolutional layer's weight and output heap buffers by multiplying configuration fields taken from a .cfg file in unchecked 32-bit int arithmetic. In src-lib/convolutional_layer.cpp, l.nweights is computed as (c / groups) * n * size * size and l.outputs as l.out_h * l.out_w * l.out_c, and both feed xcalloc directly. A .cfg whose true dimension product exceeds INT_MAX wraps to a small or zero value, so the allocation is undersized; for example width and height of 256 with filters of 65536 gives 2^32, which wraps to 0. forward_convolutional_layer then re-derives the GEMM dimensions with a different operand order, computing k as l.size*l.size*l.c / l.groups where the allocation divided before multiplying, and reads and writes through the undersized buffer. Loading the crafted .cfg for inference or training is sufficient and no valid .weights file is required. The reported proof of concept observed a heap buffer overflow read in gemm_nn_fast under AddressSanitizer and glibc allocator metadata corruption in a release build of the same input, indicating an out-of-bounds write.
What this means
- Exposure
- Exploitable with local access to the machine and without authentication — but only if a user opens booby-trapped content.
- Impact
- An attacker can read sensitive data, modify or destroy data and take the service offline.
- Weakness
- A computation exceeds the integer type’s range, producing a nonsensical size and a broken allocation.
- Likelihood
- Its EPSS score stays low: nothing points to imminent exploitation, which is no reason to leave it unpatched.
What to doFold into the next patch cycle.
Read automatically from the CVSS vector, the weakness type (CWE) and the EPSS score. The technical description above remains the one published by NIST.