Listen to this Post
How the CVE works:
MLX is a framework for machine learning on Apple silicon.
The function mlx::core::load() parses NumPy .npy files.
A vulnerability exists in the parsing logic.
Malicious .npy files trigger a heap buffer overflow.
The issue is in bounds checking during file reading.
It causes a 13-byte out-of-bounds read.
This read accesses memory outside the allocated buffer.
The overflow occurs in the heap memory region.
Attackers can craft .npy files to exploit this.
The out-of-bounds read can lead to a crash.
It may also result in information disclosure.
Memory contents could be leaked to attackers.
The vulnerability is in versions before 0.29.4.
The patch adds proper validation of file data.
CVSS score is 5.5, indicating medium severity.
Attack vector is network-based with low complexity.
No user interaction is required for exploitation.
It affects confidentiality, integrity, and availability.
The issue is fixed in version 0.29.4.
Platform: MLX framework
Version: Pre-0.29.4
Vulnerability: Heap buffer overflow
Severity: Medium
Date: 2025-11-21
Prediction: Patched in 0.29.4
What Undercode Say:
Analytics:
pip show mlx
python -c "import mlx; print(mlx.<strong>version</strong>)"
cat > test_load.py << 'EOF'
import mlx.core as mx
try:
data = mx.load("malicious.npy")
except Exception as e:
print(e)
EOF
pip install mlx==0.29.4
How Exploit:
Craft malicious .npy file.
Trigger out-of-bounds read.
Crash or disclose memory.
Protection from this CVE:
Update to version 0.29.4.
Validate .npy files.
Apply bounds checking.
Impact:
Application crash.
Information disclosure.
Moderate severity.
🎯Let’s Practice Exploiting & Learn Patching For Free:
Sources:
Reported By: nvd.nist.gov
Extra Source Hub:
Undercode
