Most of the 'attack' s are:
1. Plain old bugs in apt.
2. Involve disabling the very security features (GPG and checksum verification) designed to prevent that attack!
I'm the author of the article. We never suggest turning off GPG and checksum verification.
The bugs may be in APT, but they allow several attack vectors against APT, as explained throughout. Let me know if you have any specific questions and I'd be happy to help clear things up!
All of the attacks presented (replay attacks, freeze attacks, and downgrade attacks) affect GPG signed APT repositories.
Imagine a version of, say, libEXAMPLE has a vulnerability allowing remote code execution. The `Valid-Until` date is some time in the future, maybe a few days from now. The authors release a new version of libEXAMPLE to patch the vulnerability and the APT repository metadata is updated.
However, a malicious actor performing a MitM against your machine has saved the metadata with the vulnerable version. The malicious actor replays that metadata to your system, preventing your system from seeing the newly patched libEXAMPLE. This gives the attacker up until the `Valid-Until` date to attempt to launch an attack against you.
This option effectively disables package authentication. This is because it forces "yes" answer to all questions, including the question about installing unauthenticated packages.