CVE-2026-115913: Python Pytorch Attention Probability Retained On Module
An attention probability tensor (the softmax output over the O(T^2) score positions) computed during a forward pass is stored on the module itself (`self.$ATTR = ...`). nn.Module instances are long-lived and are shared across requests in an inference server, so this input-sized tensor is retained after the request finishes and sits outside per-request memory
Pythongreprules fetch cve-2026-115913-python-pytorch-attention-probability-retained-on-module --engine opengrepDescription
An attention probability tensor (the softmax output over the O(T^2) score positions) computed during a forward pass is stored on the module itself (`self.$ATTR = ...`). nn.Module instances are long-lived and are shared across requests in an inference server, so this input-sized tensor is retained after the request finishes and sits outside per-request memory
Community feedback
0 rule-level signals from signed-in users.
- Useful reports
- 0
- Context false positives
- 0
- Metadata suggestions
- 0
Contextual precision
Aggregated from approved scan feedback. False-positive reports describe observed scan contexts, not a global rule verdict.
- Findings observed
- 0
- Projects
- 0
- Reporters
- 0
Precision details are collecting more signed-in reports before verdict counts or ratios are emphasized.
No approved finding verdicts yet.
No scan diagnostics reported yet.