Review threats in context
Investigate IP reputation, impersonation and suspicious content alongside your Rspamd signals. Use Vision LLM to review images and explore the beta URL reputation and introspection capabilities below.
IP reputation and URL detections
Use premium IP reputation and review URL-related detections alongside authentication and message content. Inspect the evidence behind a suspicious sender or link before deciding on a policy change.
Brand impersonation
Review suspected impersonation with the surrounding message evidence. Operators can assess whether a message's claimed identity and presentation fit its actual context, rather than relying on a single isolated signal.
Vision LLM and content analysis
Analyse images in captured messages with a vision-capable model and review the result alongside the message text and scan evidence. This helps investigate image-based lures and impersonation. Connect a local model or an approved hosted provider; Jev serves separate classification tasks.
URL introspection
Inspect suspicious links in an opt-in operator workbench. Previews do not load destinations, rewrite mail or change filtering. Additional classification is under evaluation and does not establish whether a destination is safe.
URL reputation over HTTPS
Look up URL reputation through the new HTTPS integration and review results alongside Console policy controls. Use the evidence to tune how reputation informs filtering. Beta availability depends on the service and plugin release.
Example workflow
Reviewing a suspicious message
- Inspect the sender, authentication results and reputation evidence.
- Review the message's links, content and available image analysis.
- Record a supported conclusion and compare related examples before changing policy.
Deployment considerations
Detection availability depends on your premium plugins, enabled services and plan. Hosted analysis and online reputation lookups require connectivity. For an isolated deployment, agree which models and reputation data can be provisioned locally and how they will be updated.