CVE-2026-73034

CRITICAL

DB-GPT v0.8.1 Path Traversal Arbitrary File Write via user_id Header

Title source: CVE List
Search associated code STIX 2.1 vulnerability STIX 2.1 with PoCs

Record summary

CVE-2026-73034 has a selected CVSS score of 9.3 (critical); EIP currently links 1 repository PoC.

Description

DB-GPT v0.8.1 contains an unauthenticated path traversal vulnerability that allows remote attackers to write arbitrary files to any location on the server by injecting directory traversal sequences into the user_id HTTP header of the Python file-upload endpoint. Attackers can send a crafted multipart upload request with a traversal-poisoned user_id header to escape the intended upload directory and write attacker-controlled content to locations such as Python startup hooks, cron directories, or agent scripts, resulting in remote code execution.

Description source: CVE List

Exploitation context

Available material

Repository PoCs
1

CISA SSVC decision

ExploitationPoC
AutomatableYes
Technical impactTotal

CISA Coordinator · SSVC 2.0.3 · Evaluated Aug 13, 2026 · Source: CVE List

Affected products and versions

1
ProductSourceVersion range

DB-GPT

Browse eosphoros-ai / DB-GPT

Default status: affected

CVE ListThrough 0.8.1affected

Proofs of concept

1

Repository PoCs

GitHubBoreas37/CVE-2026-73034-PoCRepository PoCby Boreas371Exploit3 files

7.9 KiB

GitHub
PoC details
Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A bash script that exploits CVE-2026-73034, an unauthenticated path traversal vulnerability in DB-GPT v0.8.1, by sending crafted HTTP requests with directory traversal sequences in the 'user-id' header to write arbitrary files to the server.

Backdoor review

No backdoor observed in reviewed code

The supplied evidence consists of a README and a bash PoC script for CVE-2026-73034. Both files describe and demonstrate the path traversal vulnerability in DB-GPT v0.8.1. The script sends crafted HTTP requests to a target server to write files outside the intended directory, and then verifies the file placement via docker exec. No concealed executable behavior, persistence, credential theft, or unrelated payload delivery is present. The script's actions are consistent with a legitimate proof-of-concept for the stated CVE.

ClassificationExploit
Model confidence100%
AuthenticationNot required
Languagesbash
Target softwareDB-GPT
Attack typespath traversalarbitrary file write
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The script actively sends crafted HTTP requests to a vulnerable endpoint to write files to arbitrary locations on the server, which is the definition of an exploit. It includes verification steps to confirm successful exploitation.

poc_cve-2026-73034.sh:48-50poc_cve-2026-73034.sh:55-57poc_cve-2026-73034.sh:63-68

Requirements

  • A running, vulnerable instance of DB-GPT v0.8.1 or earlier.poc_cve-2026-73034.sh:6
  • Network access to the target's /api/v1/python/file/upload endpoint.poc_cve-2026-73034.sh:35
  • Docker must be installed and accessible for the verification steps.poc_cve-2026-73034.sh:63-65

Observed behavior

  • Sends a normal upload request with a safe 'user-id' header as a control test.poc_cve-2026-73034.sh:42-44
  • Sends an exploit request with a 'user-id' header containing directory traversal sequences to write a file to /tmp/pwned/.poc_cve-2026-73034.sh:48-50
  • Sends a second exploit request to write a file to /root/.poc_cve-2026-73034.sh:55-57
  • Verifies the exploit by executing commands inside the target Docker container to list and read the written files.poc_cve-2026-73034.sh:63-68
  • Tests a fixed build of the application to confirm the vulnerability is patched and the exploit is blocked.poc_cve-2026-73034.sh:72-76
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Vulnerability Exploit
Path traversal via user-id header in POST /api/v1/python/file/uploadThe PoC demonstrates the core vulnerability described in CVE-2026-73034 by injecting directory traversal sequences into the user-id header to write files to arbitrary locations.poc_cve-2026-73034.sh:48-50poc_cve-2026-73034.sh:55-57
Verification Command
docker exec commands to check file placementThe script uses docker exec to verify that the uploaded files were written to the traversal target directories, which is standard for a PoC that validates the exploit.poc_cve-2026-73034.sh:62-68
Review boundaries

What the analysis did not establish

  • The evidence packet reports complete_artifact_coverage as false, indicating one file (likely a license or similar) was omitted from the text content, but the two provided files (README.md and poc_cve-2026-73034.sh) are complete and sufficient for classification.
  • One file (unclassified, 2411 bytes) was omitted from text analysis per binary policy; its metadata was flagged but not inspected. No evidence suggests it contains harmful content, but it was not reviewed.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

References

5