Denial of Service Vulnerability in PyTorch by Meta
CVE-2025-55560

7.5HIGH

Key Information:

Vendor

Meta

Status
Vendor
CVE Published:
25 September 2025

What is CVE-2025-55560?

A vulnerability in PyTorch version 2.7.0 can potentially lead to a Denial of Service (DoS) attack. This issue arises when models utilize the torch.Tensor.to_sparse() and torch.Tensor.to_dense() functions while being compiled by Inductor. The misuse of these tensor methods can cause significant performance degradation, resulting in service unavailability. It is crucial for developers using this version of PyTorch to be aware of this flaw and take necessary precautions to mitigate the associated risks.

References

CVSS V3.1

Score:
7.5
Severity:
HIGH
Confidentiality:
None
Integrity:
None
Availability:
None
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
None
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.