SEZ9 commented on PR #11557: URL: https://github.com/apache/seatunnel/pull/11557#issuecomment-6889140130
Thanks for the update to `7fa315bac1d0fb30a3380f5b5eb8baebd1adb904`. Following up on the open item from the earlier review (PR11557-F1, HIGH / Security): the unsafe Java deserialization pattern in `seatunnel-engine/seatunnel-engine-server/src/main/java/org/apache/seatunnel/engine/server/task/operation/source/ManagedSourceRegisterOperation.java`. As noted on the thread, the net patch at this head has the same added and removed lines as the previously reviewed head `1f0ac9d495323a036a27bcca5320b547a9f9b0a0`, so I don't see anything indicating this finding has been addressed yet. If I've missed a change, please point me to it. Before I can approve, I'd ask for: 1. In `ManagedSourceRegisterOperation`, avoid plain Java object deserialization of the registration payload from untrusted operation bytes — either encode the payload field by field in a versioned envelope, or, if an object stream is unavoidable, restrict it to a strict allow-list of the exact types the registration envelope can carry. 2. A focused unit test for the read path that (a) round-trips a valid registration envelope and (b) asserts that a payload containing an unexpected or disallowed class is rejected rather than instantiated. 3. A short note in the PR description on which approach you chose, so the reasoning is on record alongside the protocol-version/compatibility story. Once that's pushed I'll re-check this item on the new head. Happy to discuss the encoding option first if that helps. <!-- streview-comment:2366 --> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
