Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
giskard-ai/giskard-oss
main @ 962ae81
42
Fixes
3
CVEs
HIGH
Peak severity
88.9%
Coverage
Highlights
Auth Bypass: 24 prior fixes. Scrutinize any change in this area.
backend/src/main/java/ai/giskard/config/SecurityConfiguration.java: most-fixed (4 issues). Treat as high-risk during review.
35 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: Critical missing or flawed authorization checks (such as using reference identity instead of string value equality for user comparisons) historically allowed attackers to bypass project access controls and manipulate datasets, models, or configurations.
Server-Side Template Injection: Rendering raw user-supplied rule parameters through standard Jinja2 template constructors permitted Server-Side Template Injection (SSTI), leading to remote code execution (RCE) on the server.
Server-Side Template Injection: Directly rendering user messages as template sources in agent workflows permitted remote code execution. This was mitigated by requiring explicit opt-in parsing flag gates.