Skip to content

Security: marcelorsdd/OpenPSApp

SECURITY.md

Security policy

Reporting a vulnerability

Please report vulnerabilities privately through GitHub's private vulnerability reporting:

https://github.com/marcelorsdd/OpenPSApp/security/advisories/new

If you cannot use GitHub, e-mail marcelorsdd@gmail.com instead.

Do not open a public issue, pull request or discussion for a vulnerability. Include what you found, how to reproduce it, the OpenPSApp version (the version field of http://<console-ip>:7777/api/v1/health, or the commit), and the impact you expect. Leave out real console ids, tokens and IP addresses.

OpenPSApp is maintained by one person in their spare time. You will get a first answer within 3 business days, and we will agree on a fix and a disclosure date in the advisory. Credit is given in the advisory and the release notes unless you prefer otherwise.

Supported versions

Only the latest release gets security fixes. The website and the cloud API are always the current version.

Scope

In scope:

  • The payload's LAN API: everything the ELF serves on port 7777 (/api/v1/... and the embedded files).
  • The console UI in ui/.
  • The website in web/.
  • The cloud API, including pairing, sync and the read API. Its source is private, but reports about its behaviour are welcome here.
  • The release and CI workflows in .github/.

Out of scope:

  • Anything that requires code execution on the jailbroken console. Such code can already read every file OpenPSApp keeps, including the token.
  • The jailbreak, payload loaders and ps5-payload-sdk. Report those to their projects.
  • Debug endpoints, which are off by default and documented as exposing console files to the network when "debug": true.
  • Findings that only show that the LAN API is readable without a password. That is by design and documented in the README.
  • Vulnerabilities in dependencies with no effect on OpenPSApp. Report them upstream; tell us if OpenPSApp is affected.

Threat model

What OpenPSApp protects: the user's synced gaming history (profile name, games, playtime, trophies), the console's device token, and the user's account on the website.

Who What we assume
Other devices on the LAN They can read the LAN API and hide games, but must not link, unlink or sync the console.
Websites open in a browser on the LAN They must not change anything on the console (CSRF) or reach the cloud actions (DNS rebinding).
The network between console and server It must not read or change the traffic, or impersonate the server.
Someone with a pairing link They must not get a console linked to their account without the owner seeing the code on the TV.
Someone with a copy of the database They must not be able to use the device tokens.
The server operator Can read the synced data. It is not end-to-end encrypted.

Existing protections

  • Loopback gating. Linking, unlinking, cancelling a code and Sync now only work when the request comes from 127.0.0.1 with a Host of 127.0.0.1:<port> or localhost:<port>. Other clients never see the pairing code.
  • Host allowlist against DNS rebinding. The payload answers only requests whose Host is localhost or an IP address; any host name gets 403 bad_host, so a website cannot rebind its own name to the console and read the API.
  • A custom header against CSRF. Every POST route requires X-OpenPSApp: 1. The payload sends no CORS headers, so a page on another origin cannot add that header.
  • Request limits. 16 KiB of headers, 64 KiB of body, a 10 s deadline per request and 32 concurrent connections.
  • Pinned roots. The payload's own HTTPS client (mbedTLS) refuses plain HTTP, checks the certificate and host name, and trusts only ISRG Root X1 and X2. Responses are capped at 8 MiB.
  • Hashed tokens. Device tokens are random 256-bit secrets; the server stores only their SHA-256 hash. A device token can only send data and is revocable from the website or the console. Pairing codes expire after 10 minutes and work once.
  • CSP. The server sends a Content Security Policy and other security headers with the website.
  • Read-only console access. The payload opens console data read-only and writes only to /data/openpsapp; a test checks that no route writes console data.
  • Fuzzing. The payload's parsers (HTTP requests and responses, URLs, trophy and console files, config.json, cloud.json) have libFuzzer targets run with ASan and UBSan (make -C payload fuzz, weekly in CI).
  • Supply chain. Dependencies are pinned and at least 14 days old, GitHub Actions are pinned by commit SHA, and releases carry SHA-256 sums and a signed build provenance attestation.

SYNC.md documents exactly what the console sends and how linking works. ARCHITECTURE.md shows the trust boundaries.

There aren't any published security advisories