CVE-2026-48797: Backpropagate · backprop ui --auth and backprop ui --share do not enforce authentication
Affects backpropagateSource GCVE Database
Not known to be exploited · CVSS 9.3 Critical : upgrade to 1.2.0 as a priority.
Data as of 27 Oct 2026, 17:42 UTC
- Patch
- Available – fixed in 1.2.0
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:NExploitation evidence
Disclosed: evidence found.
CVE published 16 Jun 2026
Detectable: no evidence on record.
No public nuclei template indexed
Weaponised: no evidence on record.
No public exploit code indexed in Metasploit or Exploit-DB
Exploited in the wildExploited: no evidence on record.
Not in CISA KEV and no source reports exploitation
Observed by OffSeq MirageSeen by Mirage: no evidence on record.
No attack traffic recorded by OffSeq Mirage sensors
A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 26 Oct 2026.
What to do
Upgrade backpropagate and @mcptoolshop/backpropagate to 1.2.0.
Affected software
| Product | Affected | Fixed in |
|---|---|---|
| PyPIbackpropagate | Affected
| Fixed in
|
| npm@mcptoolshop/backpropagate | Affected
| Fixed in
|
Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.
AI-suggested
A fixed version 1.2.0 is available and should be applied immediately. This release introduces proper authentication middleware enforcing HTTP Basic auth and WebSocket guards, along with defense-in-depth controls. Until upgrading, operators should avoid using the `--auth` and `--share` flags together. For remote access, use SSH port forwarding to securely tunnel the UI without exposing it publicly. Audit any deployments that used `--share` with `--auth` in versions 1.1.0 or 1.1.1, consider all exposed data and model paths compromised, and re-issue HuggingFace tokens used during those sessions. Note that standalone binary builds for 1.2.0 are not yet available; monitor for follow-up patches.
Description
Backpropagate versions 1.1.0 and 1.1.1 include a Reflex web UI that exposes training controls without enforcing authentication despite operator-facing flags claiming to do so. This allows any client able to reach the UI port to access sensitive functions such as reading uploaded datasets, triggering training runs, pushing to HuggingFace Hub repositories, and causing denial-of-service via disk fill. The vulnerability arises because the backend never reads the authentication environment variable and lacks middleware to enforce access control. The issue is fixed in version 1.2.0 with proper authentication middleware and multiple defense layers.
References
- github.com/mcp-tool-shop-org/backpropagate/security/advisories/GHSA-f65r-h4g3-3h9h (opens in a new tab)
- nvd.nist.gov/vuln/detail/CVE-2026-48797 (opens in a new tab)
- github.com/mcp-tool-shop-org/backpropagate (opens in a new tab)
- github.com/mcp-tool-shop-org/backpropagate/releases/tag/v1.2.0 (opens in a new tab)
- github.com/pypa/advisory-database/tree/main/vulns/backpropagate/PYSEC-2026-291.yaml (opens in a new tab)
- pypi.org/project/backpropagate (opens in a new tab)