tree 4f1783cf9605d8c305b4d29330eeebc1fef846b2
parent 85c233724020e27ca9e5ef69fec2fe6de1d592a1
author Steve Loughran <stevel@cloudera.com> 1783508419 +0100
committer GitHub <noreply@github.com> 1783508419 +0100
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJqTi3DCRC1aQ7uu5UhlAAAS5gQAGcThwuWSQAulNK4LWaYEqGj
 403aFZfbLGf3b2NN2thejO143JmpFjMwgEOi3cbhaRvZ0jEfA9zMBBJK9HeHFXFZ
 xXxP+NleVXKnagGxrzT7axc/nAHEiIQbNpuT0HWnWZvqdzQljZMGZs56rwtQCYwn
 gGY4H2JeSBF9GpaROg7+5Yj5VRzSobB8iRAcc6sj0Zr7kv8pK+Hkg5i0DAX5HvEu
 BX5o8wMWXDAtr3qG+ON5Yh2CwMXkTqnXMGQBW7DRTuNYp8O7wQyEm0NrReP2OetG
 Pf+sELw9NGXWNJ1rcNFtbV/RXf4gEGW8EibtFyAAPCMKV13R6KttgtT4D8IDTm7g
 C7oPaqZxvWq8b9qPsbCWfOjAl3UQWrw2ow0rOdx6wPOESHeQxmA6Z6Le684FytWr
 1HTIlB0esvWpnmhMRnqi54gaVePlUZX4OVpBTayHMlUul9VMpn0TG/3XXHKkSB0V
 nne14M9rmQL3OPuDVX6UuLiHgGQ/Ypysl/KfbedU1qaVK5CVe9ubAq47A5D++Kws
 j7HedH9Ei0thczl7WrM8OFHj3TFHB030WfSPF6vhcXg5ORrsjQDs0OW95n41KJqH
 /rFZllitDh6PeUSjeAWb/HwbImmONAIZZ/l7fopTsBfihyxPSKafAMtHPTXmWOgh
 pdmPfRBMYc2sTwBgbuuf
 =Km4M
 -----END PGP SIGNATURE-----
 

HADOOP-19925. Define the Hadoop Security Model in a SECURITY.md file (#8562)


Add a SECURITY.md file for humans and AI, with a link from AGENTS.md telling
the machines to read it.

The file tries to define the project policy as to what is:
- a CVE
- a bug
- expected behavior, especially in a cluster without kerberos
- job submission, which, rather than RCE with a score of 9, would be a
  support call if it didn't work.

The process for reporting a CVE is covered, and the pre-requisites: it must
be reproducable against trunk on a kerberized cluster, the human submitter
must actually understand the issue, and more.

The section on development and CI tries to explore what we have to trust
during development (javac, docker images), and how to escalate any issues in
CI (call asf central security).

Initial prose generated by Claude

Contributed by Steve Loughran