On 13/11/2026 9:28 am, Elven Inquisition wrote:
He noted in his initial reply to me in this thread[2] that
the upgrade process during netinstall utilizes unattended-upgrades. That
utilizes APT internally.

I think, unattended-upgrades is not involved and apt is invoked directly. In Debian it is not installed by default, see apt list '(~D~nunattended-upgrades)|(~DRecommends:~nunattended-upgrades)'

Perhaps the point was that users having unattended-upgrades got fixed apt version rather early. In my opinion however, this case should be handled manually to upgrade apt safely.

it's not a *full* mitigation. The mirror itself can still ship malicious
packages;

I hope, no official mirrors are compromised. It is responsibility of users to evaluate reputation of mirrors they use. I expect that attacks by 3rd party is more probable.

By the way, may local mirror (e.g. apt-cacher-ng) be a workaround against the attack?

An incredibly similar vulnerability, CVE-2019-3462[6], likewise involving HTTP
redirection, likewise totally bypassing PGP signature verification, occurred in
2019. An emergency point release[7] was made one day after the vuln was
announced, with exactly one thing upgraded: APT:

Was there a simple workaround that time?

Currently due to high rate of discovered vulnerabilities, a release with single bug update may cause user confusion. Other updates may be rather important as well.

I am afraid, emergency releases are bearable only if images testing are fully automated.

As I noted in my initial email, the "mitigation" instructions in
this case seem to be, by virtue of their absence, "Upgrade as usual.

I agree that handling of this couple of issues was not optimal. It is the reason why I joined this thread.

I wasn't subscribed when I posted, and I hadn't expected to be
dropped. I'm subscribed now, though.
Perhaps I have seen a suggestion to explicitly request "Cc" in some docs related to Debian mailing lists.

My rule of thumb for mailing lists where direct copies are frequently used is that if a person joined an earlier started thread than their address may be skipped. An exception is the case when it is important to hear their opinion. For new threads I sometimes check if the person joined to other threads in the past.

On 13/11/2026 9:55 pm, monodev wrote:
On 11/12/26 6:19 PM, Max Nikulin wrote:
I do not mind that new system is not vulnerable as soon as security
updates are installed. So the window for attacks is narrow, but it exists.
In that case such a bug would no longer be addressed by a new point
release (emergency or otherwise) as it'd exist outside of both apt and
ca-certificates.

It is not outside of apt. The point is that earlier apt versions may overwrite a file after its hashsum is verified. GPG signatures for package lists are genuine, but attacker's code may be executed on the system.

Reply via email to