Lol Microsoft is not even close to a walled garden. This is just them removing the password manager feature that nobody used from their authenticator app.

It’s not exactly clear what the main goal is here and it sounds like a bad idea on first glance after reading your last paragraph. But it sounds like you might be looking for mandatory profiles.

https://learn.microsoft.com/en-us/windows/client-management/client-tools/mandatory-user-profile

But now that I read it again I think you might be conflating users/profiles with sessions. In which case no, this is neither possible nor a good idea. But it still might be okay with mandatory profiles if the device and app works the same from multiple sessions.

Anyway, you might get better answers if you state the full problem, including details on software and device, not just your proposed solution.

https://xyproblem.info/

The easiest way that doesn’t affect the main network would be to use a travel router. Its WAN IP would be the private IP it gets from the main network (over wireless since that’s your only option). And it would NAT your network onto that IP and then you can do whatever you want on your network.

I’m not sure if that Mikrotik router will do this but it might. You basically need something that can connect to an SSID and use that interface as its WAN interface. The wireless factor here is really limiting your choices. If you had a wired uplink to the main network you could use any router/gateway/firewall you wanted. You could also use an AP in bridge mode to connect to the main network’s SSID and wire it to the WAN port of any router of your choice.

You don’t really need to use VLANs to separate your network from the main network unless you want to share any of the same layer 2 segments (basically wired Ethernet) while keeping it isolated. But it doesn’t really sound like that applies in your scenario. Of course using VLANs within your network would still make sense if that applies (for example, to separate your server traffic from your IoT traffic).

Definitely malware, as everyone has already said.

https://www.malwarebytes.com/blog/news/2025/03/fake-captcha-websites-hijack-your-clipboard-to-install-information-stealers

They’re only killing the crappy store/UWP version that nobody used anyway and only caused confusion. The normal OneNote bundled in Office isn’t going anywhere as far as I know.

That said, I’ve moved a lot of my note taking to Obsidian. It’s not a perfect replacement but it’s a fantastic markdown editor and now I use both for different use cases.

Multiple monitor support would be nice but that’s probably a ways off.

I heavily use both and this is objectively untrue.

This is a good answer.

To add, for Linux kernels, the maintainer use a shim EFI package with the distro’s keys (e.g., Canonical’s keys for Ubuntu) which loads the maintainer-signed kernel. And Microsoft signs the shim to keep the chain intact.

I don’t deal with hardware much anymore, but I’d take Aruba over Cisco any day. But for everything else, yeah fuck HP.

I’m Ron Burgundy?

Yep that’s how I have Syncthing set up. All global and local discovery disabled, no firewall ports open on the clients, no broadcasting, no relay servers. Just syncing through a central server which maintains versioning and where the backups run. Works like a charm.

As another poster mentioned, QubesOS with anti evil maid will work, but that’s the defense against state actors too and is overkill for this threat model.

BitLocker or any FDE using SecureBoot and PCR 7 will be sufficient for this (with Linux you also need PCRs 8+9 to protect against grub and initramfs attacks). Even if they can replace something in the boot chain with something trusted, it’ll change PCR 7 and you’d be prompted to unlock with a recovery key (don’t blindly enter it without verifying the boot chain and knowing why you’re being prompted).

With Secure Boot alone, the malicious bootloader would still need to be trusted (something like BlackLotus).

Also make sure you have a strong BIOS password and disable boot from USB, PXE, and anything else that isn’t the specific EFI bootloader used by your OS(es).

And what about taking a nice drive down Jean Baptiste Pointe du Sable Lake Shore Drive?

Not that it’s my first recommendation for security reasons, and I would never do this in prod, but you can just add the self-signed cert to the local trusted root CA store and it should work fine. No reg changes needed.

If you do this, put it in the store of the user running the client, not LocalMachine. Then you just need to make sure you connect as something in the cert’s SAN list. An IP might work (don’t know since I never try to put IPs in the SAN list), but just use a hosts entry if you can’t modify local DNS.

Edit: after reading the full OP post (sorry), I don’t think it’s necessarily the self-signed cert. If the browser is connecting with https:// and presenting a basic auth prompt, then https is working. It almost sounds like there is a 301/302 redirect back to http after login. Check the Network tab of the browser’s dev pane (F12) to see what is going on.

Microsoft uses TPM PCRs 7+11 for BitLocker which is more secure than the Linux implementations mentioned in the article.

PCR 7 is the Secure Boot measurement which means it can’t be unlocked unless every signed boot component has not been tampered with up to the point of unlock by the EFI bootloader. PCR 11 is simply flipped from a 0 to a 1 by the bootloader to protect the keys from being extracted in user land from an already booted system.

The article is correct that most Linux implementations blindly following these kinds of “guides” are not secure. Without additional PCRs, specifically 8 and 9 measuring the grub commands (no single-user bypass) and initrd (which is usually on an unencrypted partition), it is trivial to bypass. But the downside of using these additional PCRs is that you need to manually unlock with a LUKS2 password and reseal the keys in TPM whenever the kernel and or initrd updates.

Of course to be really secure, you want to require a PIN in addition to TPM to unlock the disk under any OS. But Microsoft’s TPM-only implementation is fairly secure with only a few advanced vulnerabilities such as LogoFAIL and cold boot attacks.

I use it for providing a text summary of YouTube videos that I can parse quickly. Because everything has to be a gorram video these days.

Linux on enterprise user endpoints is an insane proposition for most organizations.

You clearly have no experience managing thousands of endpoints securely.

Looks like they found someone.

This is like the epitome of the XY Problem.

https://xyproblem.info/

I agree with the first part but vehemently disagree with the third paragraph.

I suspect it varies wildly based on where you live, but in Chicago there absolutely ARE places with waitstaff worth getting a burger from.