@gdog05 Automation tools built using particular llms tend to default to the model name as a UA. UA is largely meaningless nowadays for malicious bots (usually simulates a human UA)

So it’s an AI script kiddie?

I get frequent login attempts from Google servers too. I got the impression they offer a SaaS service some script kiddie uses to attack home assistant installations.

Make sure passwords are unique and activate incorrect login bans.

How are they casually finding these home assistant servers? I sure hope people dont have port 8123 exposed publicly like that? Are you all broadcasting it in to the ether?

Haha heck no, at least not in my case. It’s not impossible to just stumble on the domain. An infinite monkey sorta thing.

Nabu Casa offers a reverse proxy through their own services. I run mine this way. Haven’t really had any problems with it in the year I’ve been subscribed.

reverse proxy

Wait. Is this really all that nabu casa is doing? Why in the ever loving fuck would anybody agree to that, let alone pay them for it?

Looks like it’s true. They are even leaking all the hostnames in the certificate transparency logs for the bots to scrape. Just when you think you’ve seen it all…

Because there are a limited number of IPv4 addresses and bots scan them for open ports.

If you use a publicly trusted TLS certificate (either directly, or via Nabu Casa offering), domain will be recorded in Certificate Transparency logs.

You can slightly hide it if you use a wildcard domain in the TLS certificate (e.g. *.example.com), and then use a subdomain for the service. Ideally, something that’s not home/hass/ha.example.com, or otherwise easily guessable.

104.28.222.47? That one showed up in a log yesterday for me. Super glad the latest update added the option to see the IP for login attempts.

This one was 66.187.5.19. I tried to find a way to report their behavior on hostodo (the service they’re using) but they don’t seem to have a report option. I’m guessing because they don’t want to prevent this kind of thing.

Often you can just send them a email about the behavior, Using their generic contact info.

the whois info for the IP might have an abuse email address, try that

Can confirm it does:

OrgAbuseHandle: ABUSE9107-ARIN
OrgAbuseName:   ABUSE
OrgAbusePhone:  +1-313-516-8908
OrgAbuseEmail:  abuse@hostodo.com
OrgAbuseRef:    https://rdap.arin.net/registry/entity/ABUSE9107-ARIN

OrgAbuseHandle: ABUSE9107-ARIN

OrgAbuseName: ABUSE

OrgAbusePhone: +1-313-516-8908

OrgAbuseEmail: abuse@hostodo.com

OrgAbuseRef: rdap.arin.net/registry/entity/ABUSE9107-ARIN

They have options. No idea if they actually do anything for the reports, but it’s worth a shot.

I’ve not looked into Nabu Casa much, is it an outbound VPN from our HA instances to their servers? If so, maybe the Nabu Casa admins need a blocklist?

That is exactly it, yes. And I am thinking about reaching out to them about it and their blocklist. But after thinking about it, anyone with $16 in hosting can start an AI hacking instance. It’s just going to get worse.

This kind of thing has been going on since long before “AI.” Expect anything connected to the internet will have failed login requests. That’s why there is a login system.

I agree this is an old thing, my firewall has blocklists and allowlists to prevent known bad IPs and allow only the countries I travel to.

But if Nabu Casa is an outbound VPN, then my blocks won’t work. I’d need them to block

Either way it would make sense for HA to also use some crowdsourced blocklists as a 2nd level defense

What do you do if somebody adds your IP to the crowdsourced blocklist used by your server?

Ask for it to be removed

Not just that, companies are using “smart” devices as AI scraper botnets to utilize private, basically unbanable IPs. There are only very few companies who I might believe them not doing it (AllenAI and maybe Mistral - tell me if I’m proven wrong pls). But OpenAI, Anthropic, fucking Google and Meta, they all treat your network as their personal internet extension. It’s reasonable to assume any “Smart” device with wifi access that isn’t FOSS most likely being your enemy.

Our family Nextcloud already got taken down by OpenAI swarming it… overloaded and crashed php-fpm within a minute. At least one client blasting all endpoints still advertised themselves as OpenAI crawler.

Welcome to the Internet. If it’s exposed, people and bots are going to be banging on it.

Who’s getting banged?

My home assistant.

Good for him!

To shreds, you say?

I’ve got Nabu Casa and IPs from the “googleusercontent.com” domain have been hitting my instance about once per day for the last week or so.

Use strong passwords and let’s hope that the login page doesn’t have any vulnerabilities…

and let’s hope that the login page doesn’t have any vulnerabilities…

Home assistant also supports mTLS.
home assistant is something you’d likely ever want to allow from a handful of trusted devices, so deploying a client certificate on them can make sense.
This way a non-onboarded device doesn’t even get to the html part, it’s denied upstream by a reverse proxy before HA is involved.

I follow a guide for authelia and caddy, making it so that I can put authelia infront of anything without it the service behind supporting it. Still have to authenticate on the service behind but that is a small inconvenience.

You have anything online you will be hammered by bots trying to get through. That’s the sad reality of the world we live on. It has been like that forever. I get hundreds of bot malicious scans on my server each day.

All that’s left is to secure everything as hard as you can so they cannot break through.

It’s been that case since forever now. Nothing has really changed and the same rules still apply.

I still remember ‘hacking’ my ISP ADSL router 2+ decades ago and listening to the WAN interface. Hundreds of thousands of attempts per day in the early 2000s. Mostly US, Russia and China.

My personal #1 threat on the internet for the past 20 years has been DDOS and thanks to billions of IOT devices and privacy invading ‘smart’ appliances everywhere… The botnets have never been this plenty. DDOS as a Service for mere cents a day.

LLM’s joined the fray in recent years but it still can’t do anything it hasn’t learned from us humans.

Wireguard can solve secure remote access without* passwords. Android and iPhone both have clients that can be configured to bring the VPN up once you leave wi-fi. Bringing a third party into your home is just not wise.

My router (unifi dream machine) offers regional blocking, so I block Russia, China, and a few others that I’ve gotten hits from. So far so good.

Acronyms, initialisms, abbreviations, contractions, and other phrases which expand to something larger, that I’ve seen in this thread:

[Thread #102 for this comm, first seen 11th Sep 2026, 19:50] [FAQ] [Full list] [Contact] [Source code]

The only place that I use port 22, is my crowdsec pi. I use some other port than the traditional port for HA. I am slowly moving all my things out of the Apple ecosystem once I do I will setup some other blocking system for outgoing stuff, kinda like pihole but in reverse.

Would highly recommend using an alternative port other than the default one for HA and 22.

I had a reverse proxy and all services on the web for funsies, behind various security mechanisms. Authentik, CrowdSec…

But in the age of AI, it feels far too dangerous.

Now everything still has a DNS entry and is accessible in Lan with it through my pi hole and Tailscale resolves the local subnet too.

I’ve been seeing this on my HA instance too. I expose HA though cloudflared. So I did some investigation and google has several primary ASNs I blocked them all via Cloudflare. The GCP one is AS396982 so that may be enough but I blocked them all because fuck Google.