Security hardening
A full security review of BackupZit with new automated checks found and fixed these problems:
- Restore: a tampered snapshot that holds a link and a folder with the same name could make a restore write a file outside the restore target. Duplicate names are now skipped and reported.
- Repositories: a manipulated compressed blob in an unencrypted repository could make the console or an agent reserve a lot of memory. Blob sizes are now checked before they are unpacked.
- Instant recovery (NBD): a request at the very end of a disk could pass the bounds check. The check is now exact.
- Lists and reports: a filter with a very large id (for example in a crafted link) made the calendar, run list or reports fail with an internal error.
- Storage form: a recovery key pasted with line breaks made the request fail with an internal error. Such keys are now accepted.
- Windows paths: unusual paths (device paths, drive "." or a share named ".") are refused.
- TLS: the console, its certificate helper and the update client accept TLS 1.2 or newer only.
- Appliance: SSH allows no tunnels (port, agent or X11 forwarding) and three password tries per connection; hardened kernel settings; network protocols the appliance never uses (DCCP, SCTP, RDS, TIPC) are not loaded; the PostgreSQL settings are no longer readable by every user. Other servers are not changed.
How BackupZit is tested is now described in SECURITY.md (github.com): every change runs the test suite with the race detector, fuzzing of every parser of untrusted input, gosec, staticcheck and govulncheck; every release is checked with Semgrep, Trivy, Gitleaks, an authenticated OWASP ZAP scan, role checks of every page and a scan of the appliance image (Trivy, Lynis, testssl.sh, Nuclei).
Hyper-V failover clusters: clearer guide
The Hyper-V guide's section on failover clusters now walks through the setup step by step (all 14 languages):
- Install the agent on every node — a VM is always backed up by the agent of the node it runs on, so a VM on a node without the agent is skipped while it runs there.
- The console recognises the cluster by itself — about a minute after the agents report it, the Agents page marks each node and the cluster appears in the job form. Nothing has to be set up in BackupZit or Windows.
- Create the job — in Backup jobs, open the Agent list and choose the cluster at its end, under Hyper-V failover clusters; then what to do with earlier VM jobs of single nodes.
With a cluster chosen, the job form now says All clustered VMs instead of "All VMs (and containers) on the host".
Other
- Third-party notices now include the Feather icons (MIT).
- README: screenshots, quick start and project status.
Updating: consoles install this version from Settings → Updates; update agents from the Agents page. Existing jobs, repositories and settings stay as they are.
Releases are signed with the BackupZit release key; consoles verify them before installing.