The paper doesn't cite specific commits used. It's possible that any of the commits in stable are actually good commits and not part of the experiment. I support the ban/revert, I'm just pointing out there's a 3rd option you didn't touch on.

Patches with built-in bugs made it to stable: https://lore.kernel.org/linux-nfs/YIAta3cRl8mk%2FRkH@unreal/.

Here's the commit specifically identified by Leon Romanovsky as having a "built-in bug"

https://github.com/torvalds/linux/commit/8e949363f017

That commit is from Aditya Pakki who I don't believe is affiliated with the paper in question, whose only authors are Qiushi Wu, and Kangjie Lu.

We have 4 people, with the students Quishu Wu and Aditya Pakki intruducing the faulty patches, and the 2 others, Prof Kangjie Lu and Ass.Prof Wengwen Wang patching vulnerabilities in the same area. Banning the leader seems ok to me, even if he produced some good fixes and SW to detect it. The only question is Wang who is now in Georgia, and was never caught. Maybe he left Lu at umn because of his questionable ethics.

At least one of Wang’s patches has been double reviewed and the reversion NACK’d - in other words it was a good patch.

I've looked at all of Wang's patches and they seemed to be all good.

The main culprit seems to be only Quishu Wu. He is also the one who wrote the paper.

Aditya Pakki is an RA under Kangjie Lu.

Also, they are talking of three cases. However, the list of patches to be reverted by gregkh is far longer than three, more than a hundred. Most of the first batch look sufficiently similar that I would guess all of them are part of this "research". So the difference in numbers alone points to them most probably lying.