Security context

What an agent needs to avoid regressing past fixes and find the next vuln in this repo.

marqo-ai/marqo
mainline @ 37a7283
6
Fixes
0
CVEs
HIGH
Peak severity
Highlights
Denial of Service: 5 prior fixes. Scrutinize any change in this area.
src/marqo/s2_inference/clip_utils.py: most-fixed (2 issues). Treat as high-risk during review.
1 high-severity fix in this history; regressions here are high-impact.
Recurring patterns

The bug types that recur here, drawn from past fixes, not open vulnerabilities.

Auth Bypass: Custom models downloaded from private Hugging Face or S3 repositories bypassed proper authorization. Securing this requires stateless model authentication (model_auth verification) to prevent unauthorized retrieval of private artifacts.
Denial of Service: Unvalidated input payloads allowed excessively large document sizes or unbounded searchable tensor attributes, triggering resource exhaustion. Mitigations require strict enforcement of MARQO_MAX_DOC_BYTES and maximum searchable attribute lists.
Denial of Service: Failing to properly release connections and file descriptors during remote resource fetches (such as image streaming via PIL and requests) causes socket and file handle exhaustion, leading to service unresponsiveness.