Model-output manipulation via unrestored quantization parameters
Published Mar 10, 2025 · Updated Mar 10, 2025
Improper initialization in PyTorch 2.6.0+cu124 allows local users to alter quantized Sigmoid output through crafted model state. When state_dict is loaded into a newly initialized nnq_Sigmoid instance, output_scale and output_zero_point retain constructor values instead of the saved scale and zero_point. Exploitation requires local low-privileged access and control of model state, and the demonstrated consequence is incorrect inference data rather than code execution or service interruption.
Summary
What happened
Improper initialization in PyTorch 2.6.0+cu124 allows local users to alter quantized Sigmoid output through crafted model state. When state_dict is loaded into a newly initialized nnq_Sigmoid instance, output_scale and output_zero_point retain constructor values instead of the saved scale and zero_point. Exploitation requires local low-privileged access and control of model state, and the demonstrated consequence is incorrect inference data rather than code execution or service interruption.
The record
- CVE
- CVE-2025-2149
- Published
- Mar 10, 2025
- Updated
- Mar 10, 2025
- Vendor
- PyTorch
- Product
- PyTorch
- Classifications
- CWE-665, T1565.001
- Attack vector
- local
- Privileges
- authenticated
Timeline
How it unfolded
- Mar 10, 2025CVE publishedPublication date reported by the CVE source.
- Mar 10, 2025Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=2.6.0+cu124
What conditions does exploitation require?
What is affected?
Published CVSS scores
CVSS describes severity. EPSS estimates exploitation probability.
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Public exploit references
- Quantized Sigmoid state-restoration reproducerproof of concept · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo