Security context

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

modsetter/surfsense

              main @ ee241e0
            

Fixes

CVEs

HIGH

Peak severity

40.0%

Coverage

Highlights

Auth Bypass: 15 prior fixes. Scrutinize any change in this area.

surfsense_backend/app/app.py: most-fixed (7 issues). Treat as high-risk during review.

23 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.

Auth Bypass: The application historically failed to enforce strict tenant boundary and permission isolation checks across document editor API routes, multi-agent runtimes, search connectors, and chat endpoints. This allowed unauthorized users to view, edit, or traverse resources belonging to other tenants. Developers must enforce granular RBAC and tenant matching criteria for every query session.

Sandbox Escape: The LLM sandbox execution environment faced escape vectors via vulnerable code templating engines (Jinja2) and weak sentinel boundaries in python execution heredocs. Mitigation requires using randomized secure EOF sentinels, strict network egress isolation, and restrictive global allowlists on execution contexts.

Auth Bypass: Authentication configurations were circumvented due to routers remaining mounted and reachable under alternative configurations (such as password-based endpoints persisting while in OAuth-only mode) and missing CSRF mitigations for cross-origin callbacks on specific browsers.