Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
neuralmagic/deepsparse
main @ eed7797
3
Fixes
0
CVEs
MEDIUM
Peak severity
Highlights
Denial of Service: 3 prior fixes. Scrutinize any change in this area.
src/deepsparse/transformers/pipelines/text_generation/autoregressive_preprocess_operator.py: most-fixed (1 issue). Treat as high-risk during review.
0 high-severity fixes in this history; regressions here are high-impact.
Recurring patterns
The bug types that recur here, drawn from past fixes, not open vulnerabilities.
Denial of Service: Initializing the KV cache with a dimension size of zero triggers a runtime segmentation fault, rendering the system unstable during empty cache states.
Denial of Service: Failing to check prompt size against remaining KV cache capacity during prefill operations leads to resource exhaustion and out-of-bounds cache access crashes.
Denial of Service: Absence of capacity validation checks during autoregressive token generation allows memory exhaustion or invalid state writes within the active KV session.