It's generally good advice, but I don't see that Safe Browsing did anything wrong in this case. First, it sounds like they actually were briefly hosting phishing sites:
> All sites on statichost.eu get a SITE-NAME.statichost.eu domain, and during the weekend there was an influx of phishing sites.
Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged. How else is Google supposed to know that subdomains belong to different users? That's what the PSL is for.
From my reading, Safe Browsing did its job correctly in this case, and they restored the site quickly once the threat was removed.
I'm not saying that Google or Safe Browsing in particular did anything wrong per se. My point is primarily that Google has too much power over the internet. I know that in this case what actually happened is because of me not putting enough effort into fending off bad guys.
The new separate domain is pending inclusion in the PSL, yes.
Edit: the "effort" I'm talking about above refers to more real time moderation of content.
> My point is primarily that Google has too much power over the internet.
That is probably true, but in this case I think most people would think that they used that power for good.
It was inconvenient for you and the legitimate parts of what was hosted on your domain, but it was blocking genuinely phishing content that was also hosted on your domain.
"Google does good thing, therefore Google has too much power over the internet" is not a convincing point to make.
This safety feature saves a nontrivial number of people from life-changing mistakes. Yes we publishers have to take extra care. Hard to see a negative here.
How does flagging a domain that was actively hosting phishing sites demonstrate that Google has too much power? They do, but this is a terrible example, undermining any point you are trying to make.
There are two aspects to the Internet: the technical and the social.
In the social, there is always someone with most of the power (distributed power is an unstable equilibrium), and it's incumbent upon us, the web developers, to know the current status quo.
Back in the day, if you weren't testing on IE6 you weren't serving a critical mass of your potential users. Nowadays, the nameplates have changed but the same principles hold.
> Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged.
NO, Google should be "mindful" (I know companies are not people but w/e) of the power it unfortunately has. Also, Cloudflare. All my homies hate Cloudflare.
Getting on the public suffix list is easier said than done [1]. They can simply say no if they feel like it and are making sure to be able to keep said rights as a "project" vs a "business," [2] which has its pros and cons.
> Getting on the public suffix list is easier said than done [1].
Can you elaborate on this? I didn't see anything in either link that would indicate unreasonable challenges. The PSL naturally has a a series of validation requirements, but I haven't heard of any undue shenanigans.
Is it great that such vital infrastructure is held together by a ragtag band of unpaid volunteers? No; but that's hardly unique in this space.
> Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged. How else is Google supposed to know that subdomains belong to different users? That's what the PSL is for.
A centralized list, where you have to apply to be included and it's up to someone else to decide whether you will be allowed in? How is this what they went for: "You want to specify some rules around how subdomains should be treated? Sure, name EVERY domain that this applies to."
Why not just something like https://example.com/.well-known/suffixes.dat at the main domain or whatever? Regardless of the particulars, this feels like it should have been an RFC and a standard that avoids such centralization.
There was an IETF working group that was working on a more distributed alternative based on a DNS record (so you could make statements in the DNS about common administrative control of subdomains, or lack of such common control, and other related issues). I believe the working group concluded its work without successfully creating a standard for this, though.
Yes, its generally good advice to keep user content on a separate domain.
That said, there are a number of IT professionals that aren't aware of the PSL as these are largely initiatives that didn't exist prior to 2023 and don't get a lot of advertisement, or even a requirement. They largely just started being used silently by big players which itself presents issues.
There are hundreds if not thousands of whitepapers on industry, and afaik there's only one or two places its mentioned in industry working groups, and those were in blog posts, not whitepapers (at M3AAWG). There's no real documentation of the organization, what its for, and how it should be used in any of the working group whitepapers. Just that it is being used and needs support; not something professional's would pay attention to imo.
> Second, they should be using the public suffix list
This is flawed reasoning as is. Its hard to claim this with a basis when professionals don't know about this, a small subset just arbitrarily started doing this, and seems more like false justification after-the-fact for throwing the baby out with the bath water.
Security is everyone's responsibility, and Google could have narrowly tailored the offending domain name accesses instead of blocking the top-level. They didn't do that, and worse that behavior could even be automated in a way that the process could be extended and there could be a noticing period to the toplevel provider before it started hitting everyone's devices. They also didn't do that apparently.
Regardless, no single entity should be able to dictate what other people perceive or see arbitrarily from their devices (without a choice; opt-in) but that is what they've designed these systems to do.
Enumerating badness doesn't work. Worse, say the domain names get reassigned to another unrelated customer.
Those people are different people, but they are still blocked as happens with small mail servers quite often. Who is responsible when someone who hasn't been engaged with phishing is being arbitrarily punished without due process. Who is to say that google isn't doing this purposefully to retain their monopolies for services they also provide.
Its a perilous torturous path where trust cannot be given because they've violated that trust in the past, and have little credibility with all net incentives towards their own profit at the expense of others. They are even willing to regularly break the law, and have never been held to account for it. (i.e. Google Maps WIFI wiretapping).
Hanlon's razor is a joke intended as a joke, but there are people that use it literally and inappropriately to deceitfully take advantage of others.
Gross negligence coupled with some form of loss is sufficient for general intent which makes the associated actions malicious/malice.
Throwing out the baby with the bath water without telling anyone or without warning, is gross negligence.
I'm not sure what to tell you. I'm a professional with nearly two decades of experience in this industry, and I don't read any white papers. I read web publications like Smashing Magazine or CSS Tricks, and more specifically authors like Paul Irish, Jake Archibald, Josh Comeau, and Roman Komarov. Developers who talk about the latest features and standards, and best practices to adopt.
The view that professionals in this industry exclusively participate in academic circles runs counter to my experience. Unless you're following the latest AI buzz, most people are not spending their time on arXiv.
The PSL is surely an imperfect solution, but it's solving a problem for the moment. Ideally a more permanent DNS-based solution would be implemented to replace it. Though some system akin to SSL certificates would be necessary to provide an element of third-party trust, as bad actors could otherwise abuse it to segment malicious activity on their own domains.
If you're opposed to Safe Browsing as a whole, both Chromium and Firefox allow you to disable that feature. However, making it an opt-in would essentially turn off an important security feature for billions of users. This would result in a far greater influx of phishing attacks and the spread of malware. I can understand being opposed to such a filter from an idealistic perspective, but practically speaking, it would do far more harm than good.
Putting user content on another domain and adding that domain to the public suffix list is good advice.
So good, in fact, that it should have been known to an infrastructure provider in the first place. There's a lot of vitriol here that is ultimately misplaced away from the author's own ignorance.
The PSL is something you find out about after it goes wrong.
It's a weird thing, to be honest, a Github repo mentioned nowhere in any standards that browsers use to treat some subdomains differently.
Information like this doesn't just manifest itself into your brain once you start hosting stuff, and if I hadn't known about its existence I wouldn't have thought to look for a project like this either. I certainly wouldn't have expected it to be both open for everyone and built into every modern internet-capable computer or anti malware service.
To be fair I’ve been in the space for close to 20 years now, worked on some of the largest sites and this is the first I’m hearing of the public suffix list.
I don't know about LiveJournal, but I don't believe you can host any interactive content on substack (without hacking substack at least). You can't sign up and host a phishing site, for instance.
User-uploaded content (which does pose a risk) is all hosted on substackcdn.com.
The PSL is more for "anyone can host anything in a subdomain of any domain on this list" rather than "this domain contains user-generated content". If you're allowing people to host raw HTML and JS then the PSL is the right place to go, but if you're just offering a user post/comment section feature, you're probably better off getting an early alert if someone has managed to breach your security and hacked your system into hosting phishing.
The public suffix list interferes with cookies. So on a service like livejournal, where you want users logged in across all subdomains, it's not an option
Exactly, this has been documented knowledge for many years now, even decades. Github and other large providers of user-generated content have public-facing documentation on the risks and ways to mitigate them. Any hosting provider that chooses to ignore those practices is putting themselves, and their customers, at risk.
> There's a lot of vitriol here that is ultimately misplaced away from the author's own ignorance.
For what it's worth, this makes it sound like you think the vitriol should be aimed at the author's ignorance rather than the circumstances which led to it, presuming you meant the latter.
I do think the author's ignorance was a bigger problem--both in the sense of he should have known better and also in the sense that the PSL needs to be more discoverable--than anything Google('s automated systems) did.
However, I'm now reflecting on what I said as "be careful what you wish for", because the comments on this HN post have done a complete 180 since I wrote it, to the point of turning into a pile-on in the opposite direction.
The good news is, once known, a lesson like this is hard to forget.
The PSL is one of those load-bearing pieces of web infrastructure that is esoteric and thanklessly maintained. Maybe there ought to be a better way, both in the sense of a direct alternative (like DNS), and in the sense of a better security model.
Since there's a lot of discussion about the Public Suffix list, let me point out that it's not just a webform where you can add any domain. There's a whole approval process where one very important criterion is that the domain to be added has a large enough user base. When you have a large enough user base, you generally have scammers as well. That's what happened here.
It basically goes: growing user base -> growing amount of malicious content -> ability to submit domain to PSL. In that order, more or less.
In terms of security, for me, there's no issue with being on the same domain as my users. My cookies are scoped to my own subdomain, and HTTPS only. For me, being blocked was the only problem, one that I can honestly admit was way bigger than I thought.
What sort of size would be needed to get on there?
My open source project has some daily users, but not thousands. Plenty to attract malicious content, I think a lot of people are sending it to themselves though (like onto a malware analysis VM that is firewalled off and so they look for a public website to do the transfer), but even then the content will be on the site for a few hours. After >10 years of hosting this, someone seems to have fed a page into a virus scanner and now I'm getting blocks left and right with no end in sight. I'd be happy to give every user a unique subdomain instead of short links on the main domain, and then put the root on the PSL, if that's what solves this
Based on what I've seen, there's no way to get that project into the PSL. I would recommend you to have the content available at projectcontent.com if the main site is project.com, though. :)
As a CISO I am happy with many of the protections that Google creates. They are in a unique position, and probably the only ones to be able to do it.
However, I think the issue is that with great power comes great responsibility.
They are better than most organisations, and working with many constraints that we cannot always imagine.
But several times a week we get a false "this mail is phishing" incident, where a mail from a customer or prospect is put in "Spam", with a red security banner saying it contains "dangerous links". Generally it is caused by domain reputation issues, that block all mail that uses an e-mail scanning product. These products wrap URLs so they can scan when the mail is read, and thus when they do not detect a virus, they become defacto purveyors of virii, and their entire domain is tagged as dangerous.
I have raised this to Google in May (!) and have been exchanging mail on a nearly daily basis. Pointing out a new security product that has been blacklisted, explaining the situation to a new agent, etc.
Not only does this mean that they are training our staff that security warnings are generally false, but it means we are missing important mail from prospects and customers. Our customers are generally huge corporations, missing a mail for us is not like missing one mail for a B2C outfit.
So far the issue is not resolved (we are in Oct now!) and recently they have stopped responding. I appreciate our organisation is not the US Government, but still, we pay upwards of 20K$ / year for "Google Workspace Enterprise" accounts. I guess I was expecting something more.
If someone within Google reads this: you need to fix this.
I'm old. I've been doing security for a very long time. Started back in the 1990s. Here's what I have learned over the last 30 years...
Half (or more) of security alerts/warnings are false positives. Whether it's the vulnerability scanner complaining about some non-existent issues (based on the version of Apache alone... which was back ported by the package maintaner), or an AI report generated by interns at Deloitte fresh out of college, or someone reporting www.example.com to Google Safe Browsing as malicious, etc. At least half of the things they report on are wrong.
You sort of have to have a clue (technically) and know what you are doing to weed through all the bullshit. Tools that block access, based on these things do more harm than good.
What this post might be missing is that it’s not just Google that can block your website. A whole variety of actors can, and any service that can host user-generated content, not just html (a single image is enough), is at risk, but really, any service is at risk. I’ve had to deal with many such cases: ISPs mistakenly blocking large IP prefixes, DPI software killing the traffic, random antivirus software blocking your JS chunk because of a hash collision, even small single-town ISPs sinkholing your domain because of auto-reports, and many more.
In the author’s case, he was at least able to reproduce the issues. In many cases, though, the problem is scoped to a small geographic region, but for large internet services, even small towns still mean thousands of people reaching out to support while the issue can’t be seen on the overall traffic graph.
The easiest set of steps you can do to be able to react to those issues are:
1. Set up NEL logging [1] that goes to completely separate infrastructure,
2. Use RIPE Atlas and similar services in the hope of reproducing the issue and grabbing a traceroute.
I’ve even attempted to create a hosted service for collecting NEL logs, but it seemed to be far too niche.
I don't see how a separate domain would solve the main issue here. If something on that separate domain was flagged, it would still affect all user content on that domain. If your business is about serving such user content, the main service of your business would be down, even though your main domain would still be up.
You are right, it would still affect all users. Until the pending PSL inclusion is complete, that is. But it now separates my own resources, such as the website and dashboard of statichost.eu from that.
A separate domain may not prevent users' conten from being blocked, but it may prevent blocking of the administrative interfaces. Which would help affected customers get their content and the service could more easily put a banner advising users of the situation, etc.
I second this for personal sites. Having run forums and chan sites without a CDN I found that not only is this true, it is 100% automated. The timing in emails to VPS/Registrars matches the times their scripts would crawl my sites and submit illicit content, screenshot it and automatically submit the screenshots to the VPS/server/registrar providers. That was incentive enough for me to take my sites private / semi-private. I would move them to .onion nodes but that's just too slow for me. I have my own theories as to what groups are running these scripts to push people to CDN's but no smoking gun.
Corporations are a little safer. They have mutually binding contracts with multiple internet service providers and dedicated circuits. They have binding contracts with DNS registrars. Having been on the receiving end of abuse@ they notify over phone and email giving plenty of time to figure out what is going on. I've never seen corporate circuits get nuked for such shenanigans.
Any services successfully offloading UGC to other moderated platforms? E.g. developer tools relying on GitHub instead of storing source/assets in the service itself, and Microsoft can take care of most moderation needs. But are there consumer apps that do things like this?
But something has definitely changed over the past few years. Back in the days, it felt completely normal for individuals to spin up and run their own forums. Small communities built and maintained by regular people. How many new truly independent, individual-run forums can you name today? Hardly any. Instead we keep hearing about long-time community sites shutting down because individuals can no longer handle the risks of user content. I've seen so many posts right here on HN announcing those kinds of shutdowns.
Your equally just one fake report to an automated system away having your account shut down. So, yes, your actions have consequences, but more worrying to me is the ability of someone with a grudge causing consequences for you as well.
Not sure who changed the HN headline, but I appreciate the change. Especially since the concept in the headline is buried at the bottom of the post.
Post author is throwing a lot of sand at Google for a process that has (a) been around for, what, over a decade now and (b) works. The fact of the matter is this hosting provider was too open, several users of the provider used it to put up content intended to attack users, and as far as Google (or anyone else on the web is concerned) the TLD is where the buck stops for that kind of behavior. This is one of the reasons why you host user-generated content off your TLD, and several providers have gotten the memo; it is unfortunate statichost.eu had not yet.
I'm sorry this domain admin had to learn an industry lesson the hard way, but at least they won't forget it.
Author here. I understand that my post and what I'm trying to say is unclear. And that there are too many different aspects to all this.
What I'm trying to say in the post specifically about Google is that I personally think that they have too much power. They can and will shut down a whole domain for four billion users. That is too much power no matter the intentions, in my opinion. I can agree that the intentions are good and that the net effect is positive on the whole, though.
On the "different aspects" side of things, I'm not sure I agree with the _works_ claim you make. I guess it depends on what your definition of works is, but having a blacklist as you tool to fight bad guys is not something that works very well in my opinion. Yes, specifically my own assets would not have been impacted, had I used a separate domain earlier. But the point still stands.
The fact that it took so long to move user content off the main domain is of course on me. I'm taking some heat here for saying this is more important than one (including me) might think. But nonetheless, let it be a lesson for those of you out there who think that moving that forum / upload functionality / wiki / CMS to its own domain (not subdomain) can be done tomorrow instead of today.
It can happen to anyone and cause a reputational risk. Once upon a time $workplace had a Zoho Form that would be blacklisted by Google Safe Browsing or Microsoft Edge for arbitrary periods of time, presumably because someone used Zoho to make a phishing site, leading to some very confused calls.
Sounds a very convenient mistake to do on your competitors? It does not sound believable that they would not know what zoho was or that it makes no sense to flag all the zoho domain.
In Github's case, I think it was also because a lot of security boundaries were using TLD which led x.github.com potentially grab cookies of y.github.com or worse, github.com itslef
Don't forget the `githubusercontent.com` domain, which is specifically used to host risky, user-generated content, and fully documented in https://docs.github.com/en/authentication/keeping-your-accou... (using an open source component that other companies could also use, if they were interested in similar levels of security)
I am a solo developer. I recently created a new web app for a client. Google has marked as phishing so they can't use it. Obviously I can't do anything about it except report error and wait. I'm worried if I move it to a new domain that one will get marked as well. Not sure what to do TBH.
No, however it does include a Microsoft entra/Azure AD/Microsoft 365 login for that clients tenant. It is also a newly registered domain so I can understand why it looks suspicious. The most frustrating thing is that this is all a machine I.e. no-one I can speak to, nothing I can do to fix it. My fate has been decided by an algorithm.
Google has some sort of internal flag for determining origin is different on some platforms. We don't get a complete takedown of Neocities every time there's a spam site reported. It is likely that they were not on that list but perhaps have been manually added to whatever that internal list is at this point.
The public suffix list (https://publicsuffix.org/) is good and if I were to start from scratch I would do it that way (with a different root domain) but it's not absolutely required, the search engines can and do make exceptions that don't just exclusively use the PSL, but you'll hit a few bumps in the road before that gets established.
Ultimately Google needs to have a search engine that isn't full of crap, so moving user content to a root domain on the PSL that is infested with phishing attacks isn't going to save you. You need to do prolific and active moderation to root out this activity or you'll just be right back on their shit list. Google could certainly improve this process by providing better tooling (a safe browsing report/response API would be extremely helpful) but ultimately the burdon is on platforms to weed out malicious activity and prevent it from happening, and it's a 24/7 job.
BTW the PSL is a great example of the XKCD "one critical person doing thankless unpaid work" comic, unless that has changed in recent years. I am a strong advocate of having the PSL management become an annual fee driven structure (https://groups.google.com/g/publicsuffix-discuss/c/xJZHBlyqq...), the maintainer deserves compensation for his work and requiring the fee will allow the many abandoned domains on the list to drop off of it.
If you're not using separate domains then I hope you don't have any kind of sensitive information stored in cookies. You can't rely on the path restrictions for cookies because it's easily bypassed.
Strict cookies crossing root to subdomains would be a major security bug in browsers. It's always been a (valid) theoretical concern but it's never happened on a large scale to the point I've had to address it. There is likely regression testing on all the major browsers that will catch a situation where this happens.
If youtube.com doesn't end up on the Safe Browsing blacklist because of phishing videos, but your own website can easily end up there, it's a pretty clear case of Google abusing their power.
YouTube doesn't allow you to put your credentials into text box and hit send. Google sites, on the other hand, does pose a disk, but they'll likely be treated the same as any other domain on the PSL.
In my experience, safe browsing does theoretically allow you to report scams and phishing in terms of user generated content, but it won't apply unless there's an actual interactive web page on the other end of the link.
There is the occasional false positive but many good sites that end up on that list are there because their WordPress plugin got hacked and somewhere on their site they are actually hosting malware.
I've contacted the owners of hacked websites hosting phishing and malware content several times, and most of the time I've been accused of being the actual hacker or I've been told that I'm lying. I've given up trying to be the good guy and report the websites to Google and Microsoft these days to protect the innocent.
Google's lack of transparency what exact URLs are hosting bad material does play a role there.
You are right, of course. I'm not sure if those of you who disagree with me think that Safe Browsing did its job (which it did!), that Safe Browsing is a good thing (which it maybe is, but which I slightly disagree with), or that it's ok that Google monitors everything everyone does.
The last point is actually the one I'm trying to make.
There should be a concept, sort of an inverse of tragedy of the commons, for the positive feedback loop of many users providing big data to a company that can use that data to benefit many users.
From spamblocking that builds heuristics fed by the spam people manually flag in GMail to Safe Browsing using attacks on users' Chrome as a signal to their voice recognition engine leapfrogging the industry standard a few years back because they trained it on the low-quality signal from GOOG411 calls, Google keeps building product by harvesting user data... And users keep signing up because the resulting product is good.
This puts a lot of power in their hands but I don't think it's default bad... If it becomes bad, users leave and Google starts to lose their quality signal, so they're heavily incentivized to provide features users want to retain them.
This does make it hard to compete with them. In the US at least, antitrust competition has generally been about user harm, not actually market harm. If a company has de-facto control but customers aren't getting screwed, that's fine because ultimately the customer matters (and nobody else is owed a shot at beeing a Google).
It's hard to get that point because you're conflating two different stories.
Folks around here are generally uneasy about tracking in general too, but remove big brother monitoring from Safe Browsing and this story could still be the same: whole domain blacklisted by Google, only due to manual reporting instead.
"Oh, but a human reviewer would've known `*.statichost.eu` isn't managed by us"—not in a lot of cases, not really.
Generally because Facebook polices Facebook (imperfectly, but the effort is demonstrated) and the damage radius is limited to Facebook users mostly. As long as the easiest way to avoid damage from the Facebook domain is "Don't use Facebook," the larger Internet doesn't need a mechanism to police it.
If Facebook became a trap that frequently hosted malware to strangers, the rest of the net would begin to interpret it as damage and route around it.
Many phishing attacks originate from Google's owns domains. Gmail users phishing others, scammy youtube videos, scammy comments with links to scams, scammy ads to fake banking pages, etc, etc. But Google would never be hypocritical, never!!
That is a great point. When I see these sites I'm always seeing a dozen red flags, and maybe the biggest one is that it's showing a "NatWest" banking site or something and is hosted on "portal-abc.statichost.eu". But the whole point is of course saving users from coming to harm, and if it did - great!
I’ve got a random subdomain hosting a little internal tool. About twice a year, Google Safe Browsing decides it’s phishing and flags it. Sometimes they flag the whole domain for good measure.
Search Console always points to my internal login page, which isn’t public and definitely isn’t phishing.
They clear it quickly when I appeal, and since it’s just for me, I’ve mostly stopped worrying about it.
I encountered something similar. I have `*.domain.tld` pointed to an internal IP address, and over the past few years it happened a few times where some subdomain would be flagged as dangerous by Google Safe Browsing.
Internal IP addresses in public DNS are sometimes used to do things like DNS rebind attacks. It's possible that's tripping up their detection mechanism.
My workaround is to use an IPv6 ULA for my publicly hosted private IP addresses, which is extremely unlikely to ever be reused by a bad actor.
As a result, some ISPs apparently block the domain. Why is it listed? I have no idea. There are no ads, there is no user content, and I've never sent any email from the domain. I've tried contacting spamhaus, but they instantly closed the ticket with a nonsensical response to "contact my IT department" and then blocked further communication. (Oddly enough, my personal blog does not have an IT department.)
Just like it's slowly become quasi-impossible for an individual to host their own email, I fear the same may happen with independent websites.
From reading that my guess would be that the IP of your host gotten from your hosting provider had some spammy history before you started hosting your blog on it.
Either that or your DNS provider hosts a lot of spam.
> As a result, some ISPs apparently block the domain
This is the infuriating part. I get that someone buying cheap hosting may end up with an IP address that used to send spam, but spam lists are not reliable indicators of website security.
Overzealous security products are a blight on the internet. I'd be less annoyed at them if they weren't so trivial to bypass as a hacker with access to a stolen credit card.
Anyone who can upload HTML pages to subdomain.domain.com can read and write cookies for *.domain.com, unless you declare yourself a public suffix and enough time has passed for all the major browsers to have updated themselves.
I've seen web hosts in the wild who could have their control panel sessions trivially stolen by any customer site. Reported the problem to two different companies. One responded fairly quickly, but the other one took several years to take any action. They eventually moved customers to a separate domain, so the control panel is now safe. But customers can still execute session fixation attacks against one another.
(Author here) This is all true. The main assumption from my part is that anything remotely important or even sensitive should be and is hosted on a domain that is _not_ companysubdomain.domain.com but instead www.company.com.
Many commenters are implying that there is a security issue here, and that I'm putting everyone in danger. That is quite frankly a pretty absurd claim to just casually make. I'm of course very curious to hear more details on what the security risk here actually would be?
Do you think I'm reading/writing sensitive data to/from subdomain-wide cookies?
Also, yes, the PSL is a great tool to mitigate (in practice eliminate) the problem of cross-domain cookies between mutually untrusting parties. But getting on that list is non-trivial and they (voluntary maintainers) even explicitly state that you can forget getting on there before your service is big enough.
I am not implying you’re putting “everyone” in danger. I’m merely implying that you’re putting your own service in danger by allowing clients to act like a trusted subdomain like controlpanel.statichost.eu, .secure, or Unicode similarities of www.
I don't like nor trust google, but "Use your own judgement and hard-earned Internet street smarts" doesn't work either, because the median internet user does not have anything resembling internet street smarts.
Still not sure why it's legal for Google to slander companies like this. They often have no proof or it's a false positive, meanwhile they're screaming about how malicious you are.
I'm not asking about this specific case. There are plenty of examples of Google wrongly accusing others of being malicious with massive business impact
Seems like a reasonable trade-off I mean six hours is not the worst thing in the world. What if you were hosting mission-critical such as such? Were you?
> To be fair, many or even most sites on the Google Safe Browsing blacklist are probably unworthy. But I’m pretty sure this was not the first false positive.
The bigger issue is that the internet needs governance. And, in the absence of regulation, someone has stepped in and done it in a way that the author didn't like.
Perhaps we could start by requiring that Google provide ways to contact a living, breathing human. (Not an AI bot that they claim is equivalent.)
why do you assume that the living, breathing human hired by theGoogs will be competent at handling all of the crazy that will be flung at them by the living, breathing human on the other end of the line. One single person cannot handle that. Naturally, you need a team of living, breathing humans. You might even have them in triage level groups like level 1 support, level 2 support and so on where each level is a more trained/experienced living, breathing human. Eventually, you'll have an entire department of people of varying degrees of skill. Oh, wait, I'm sorry, I thought it was the year 2000.
Hopefully, this helps you understand why your living, breathing human is such a farcical idea for theGoogs to consider.
Honestly, this is extremely basic stuff in hosting, not only due to safe browsing, but also—and more importantly—cookie safety, etc. If a hosting provider didn’t know (already bad enough) and turn to whining after being hit, then
> Static site hosting you can trust
is more like amateur hour static site hosting you can’t trust. Sorry.
The thing is, you cannot just add any domain to the PSL. You need a significant amount of users before they will include your domain. Before recently, there really was no point in even submitting, since the domain would have been rejected as too small. An increase in user base, increase in malicious content and the ability to add your domain to the PSL all happen sort of simultaneously.
I'm also trusting my users to not expose their cookies for the whole *.statichost.eu domain. And all "production" sites use a custom domain anyway, which avoids all of this anyway.
There are well-documented solutions to this that don't rely on the PSL. Choosing to ignore all of that advice while hosting user content is a very irresponsible choice, at best.
I have recently had the pleasure of speaking with Google senior leadership involved in the Safe Browsing product on the topic of getting my SaaS product placed on their, "naughty list." The platform was down for 6 or so hours due to a false positive hit for phishing.
I have read A LOT of blogs/rants/incidents on social media about startups, small businesses, and individuals getting screwed by large companies in similar capacities. I am VERY sympathetic to those cries into the sky, shaking fists at clouds, knowing very well we are all very small and how the large providers seem to not care. With that in mind, I am not blind to the privilege my organization has to rope in Google to discuss root causes for incidents.
I am writing about it here because I believe most people will never be able to pull a key Google stakeholder into a 40 minute video call to deeply discuss the RCAs. The details of the discussion are probably protected by NDA so I'll be speaking in general terms.
Google has a product called Web Risk (https://cloud.google.com/web-risk/docs/overview), I hear it's mostly used by Google Enterprise customers in regulated verticals and some large social media orgs. Web Risk protects the employees of these enterprise organizations by analyzing URLs for indicators of risk, such as phishing, brand impersonation, etc.
My SaaS platform is well established and caters mostly to large enterprise. I provide enterprise customers with optional branded SSO landing pages. Customers can either use sign-in from the branded site (SP-initiated) or redirect from their own internal identity provider to sign-in (IdP-initiated). The SSO site branding is directed by the customer, think along the lines of what Microsoft does for Entra ID branded sign-in pages. Company logo(s), name, visual styling, and other verbiage may be included. The branded/vanity FQDN is (company).productname.mydomain.com.
You may be able to see where I'm headed at this point... Why was my domain blocked? For suspected phishing.
A mutual enterprise customer was subscribed to Google's Web Risk. When their employees navigated to their SSO site, Google scanned it. Numerous heuristics flagged the branded SSO site as phishing and we were blocked by Safe Browsing across all major web browsers (Safari, Chrome, Firefox, Edge, and probably others). Google told us that had our customer put the SSO site on their Web Risk allow-list, we wouldn't have been blocked.
I'm no spring chicken, I cannot rely nor expect a customer to do that, so I pressed for more which led to a lengthy conversation on risk and the seemingly, from my perspective, arbitrary decisions made by a black box without any sort of feedback loop.
I was provided a tiny bit of insight into the heuristic secret sauce, which led to prescribed guidance on what could be done to greatly reduce the risk of getting false positive flag for phishing again. Those specifics I assume I cannot detail here, however the overall gist of it is domain reputation. Google was unable to positively ascertain my domain's reputation.
My recommendation is for those of you out there in the inter-tubes who have experienced false positive Safe Browsing blocks, think about what you can do to increase your domain's public reputation. Also, get a GCP account so if you do get blocked, you can open a ticket from that portal. I was told it would be escalated to the appropriate team and be actioned on within 10-15 minutes.
I have the same issue. Think of my site as WeTransfer, but instead of only files, you can also use it as a link shortener or pastebin. Abuse works the same as on every other site or service: I do spot checks and users can report content. This was fine until uBlock Origin decided the website was malicious, per one of the lists that is default-enabled for everyone
That list doesn't have a clear way to get off of it. I would be happy to give them the heads up that their users are complaining about a website being broken, but there is no such thing, neither for users nor for me. In looking around, there's many "sources" that allegedly independently decided around the same day that my site needs to not work anymore, so now there's a dozen parties I need to talk to and more popping up the further you look. Netcraft started sending complaints to the registrar (which got the whole domain put on hold), some other list said they sent abuse to the IP space owner (my ISP), public resolvers have started delisting the domain (pretending "there is no such domain" by returning NXDOMAIN), as well as the mentioned adblockers
There's only one person who hasn't been contacted: the owner. I could actually do something about the abusive content...
It's like the intended path is that users start to complaint "your site doesn't work" (works for me, wdym?) and you need to figure out what software is it they're using, what DNS resolver they use, what antivirus, what browser, if a DOH provider is enabled... to find out who it might be that's breaking the site. People don't know how many blocklists they're using, and the blocklists don't give a shit if you're not a brand name they recognize. That's the only difference between my site and a place like Github: if I report "github.com hosts malware", nobody thinks "oh, we need to nuke that malicious site asap!"
I'd broaden the submitted post to say that it's not only Google with too much power, but these blocklists have no notification mechanism or general recourse method. It's a whack-a-mole situation which, as an open source site with no profit model (intentionally so), I will never win. Big tech is what wins. Idk if these lists do a trademark registration check or how they decide who's okay and who's not, but I suspect it's simply a brand name thing and your reviewer needs to know you
> Luckily, Google provided me with a helpful list of the offending sites
Google is doing better than most others with that feature. Most "intelligence providers", which other blocklists like e.g. Quad9 uses, are secretive about why they're listing you, or never even respond at all
Another day, another IT company learning the hard way about the public suffix list, or well-known URIs, or some other well-documented-but-niche security technology.
I love that IT is a field where there's no required formal education track and you can succeed with any learning path, but we definitely need some better way to make sure new devs are learning about some of these gotchas.
The PSA is good, the article is meh. There is too much misdirected anger towards google here, IMO. I agree it sucks to be the false positive, but it'd also suck more to unknowingly be part of phishing campaigns and not know.
On top of that, it is also recommended to serve user content from another domain for security reasons. It's much easier to avoid entire classes of exploits this way. For the site admins: treat it as a learning experience instead of lashing out on goog. In the long run you'll be better off, having learned a good lesson.
I’m sure I don’t know ALL the "security best practices that have been around for 20+ years" and this is perfectly fine as long as I’m able to react quickly. See also https://xkcd.com/1053/.
PSA: Submitted title was 'PSA: Always use a separate domain for user content". We've changed it per https://news.ycombinator.com/newsguidelines.html. Might be worth knowing for context.
It's generally good advice, but I don't see that Safe Browsing did anything wrong in this case. First, it sounds like they actually were briefly hosting phishing sites:
> All sites on statichost.eu get a SITE-NAME.statichost.eu domain, and during the weekend there was an influx of phishing sites.
Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged. How else is Google supposed to know that subdomains belong to different users? That's what the PSL is for.
From my reading, Safe Browsing did its job correctly in this case, and they restored the site quickly once the threat was removed.
I'm not saying that Google or Safe Browsing in particular did anything wrong per se. My point is primarily that Google has too much power over the internet. I know that in this case what actually happened is because of me not putting enough effort into fending off bad guys.
The new separate domain is pending inclusion in the PSL, yes.
Edit: the "effort" I'm talking about above refers to more real time moderation of content.
> My point is primarily that Google has too much power over the internet.
That is probably true, but in this case I think most people would think that they used that power for good.
It was inconvenient for you and the legitimate parts of what was hosted on your domain, but it was blocking genuinely phishing content that was also hosted on your domain.
5 replies →
"Google does good thing, therefore Google has too much power over the internet" is not a convincing point to make.
This safety feature saves a nontrivial number of people from life-changing mistakes. Yes we publishers have to take extra care. Hard to see a negative here.
14 replies →
How does flagging a domain that was actively hosting phishing sites demonstrate that Google has too much power? They do, but this is a terrible example, undermining any point you are trying to make.
5 replies →
There are two aspects to the Internet: the technical and the social.
In the social, there is always someone with most of the power (distributed power is an unstable equilibrium), and it's incumbent upon us, the web developers, to know the current status quo.
Back in the day, if you weren't testing on IE6 you weren't serving a critical mass of your potential users. Nowadays, the nameplates have changed but the same principles hold.
2 replies →
> Google has too much power over the internet.
In this case they did use it for good cause. Yes, alternatively you could have prevented the whole thing from happening if you cared about customers.
Exactly.
> Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged.
NO, Google should be "mindful" (I know companies are not people but w/e) of the power it unfortunately has. Also, Cloudflare. All my homies hate Cloudflare.
1 reply →
Getting on the public suffix list is easier said than done [1]. They can simply say no if they feel like it and are making sure to be able to keep said rights as a "project" vs a "business," [2] which has its pros and cons.
[1] https://github.com/publicsuffix/list/blob/main/public_suffix...
[2] https://groups.google.com/g/publicsuffix-discuss/c/xJZHBlyqq...
> Getting on the public suffix list is easier said than done [1].
Can you elaborate on this? I didn't see anything in either link that would indicate unreasonable challenges. The PSL naturally has a a series of validation requirements, but I haven't heard of any undue shenanigans.
Is it great that such vital infrastructure is held together by a ragtag band of unpaid volunteers? No; but that's hardly unique in this space.
> Second, they should be using the public suffix list (https://publicsuffix.org/) to avoid having their entire domain tagged. How else is Google supposed to know that subdomains belong to different users? That's what the PSL is for.
How is this kinda not insane? https://publicsuffix.org/list/public_suffix_list.dat
A centralized list, where you have to apply to be included and it's up to someone else to decide whether you will be allowed in? How is this what they went for: "You want to specify some rules around how subdomains should be treated? Sure, name EVERY domain that this applies to."
Why not just something like https://example.com/.well-known/suffixes.dat at the main domain or whatever? Regardless of the particulars, this feels like it should have been an RFC and a standard that avoids such centralization.
There was an IETF working group that was working on a more distributed alternative based on a DNS record (so you could make statements in the DNS about common administrative control of subdomains, or lack of such common control, and other related issues). I believe the working group concluded its work without successfully creating a standard for this, though.
The problem is that you then have to trust the site's own statement about whether its subdomains are independent.
Yes, its generally good advice to keep user content on a separate domain.
That said, there are a number of IT professionals that aren't aware of the PSL as these are largely initiatives that didn't exist prior to 2023 and don't get a lot of advertisement, or even a requirement. They largely just started being used silently by big players which itself presents issues.
There are hundreds if not thousands of whitepapers on industry, and afaik there's only one or two places its mentioned in industry working groups, and those were in blog posts, not whitepapers (at M3AAWG). There's no real documentation of the organization, what its for, and how it should be used in any of the working group whitepapers. Just that it is being used and needs support; not something professional's would pay attention to imo.
> Second, they should be using the public suffix list
This is flawed reasoning as is. Its hard to claim this with a basis when professionals don't know about this, a small subset just arbitrarily started doing this, and seems more like false justification after-the-fact for throwing the baby out with the bath water.
Security is everyone's responsibility, and Google could have narrowly tailored the offending domain name accesses instead of blocking the top-level. They didn't do that, and worse that behavior could even be automated in a way that the process could be extended and there could be a noticing period to the toplevel provider before it started hitting everyone's devices. They also didn't do that apparently.
Regardless, no single entity should be able to dictate what other people perceive or see arbitrarily from their devices (without a choice; opt-in) but that is what they've designed these systems to do.
Enumerating badness doesn't work. Worse, say the domain names get reassigned to another unrelated customer.
Those people are different people, but they are still blocked as happens with small mail servers quite often. Who is responsible when someone who hasn't been engaged with phishing is being arbitrarily punished without due process. Who is to say that google isn't doing this purposefully to retain their monopolies for services they also provide.
Its a perilous torturous path where trust cannot be given because they've violated that trust in the past, and have little credibility with all net incentives towards their own profit at the expense of others. They are even willing to regularly break the law, and have never been held to account for it. (i.e. Google Maps WIFI wiretapping).
Hanlon's razor is a joke intended as a joke, but there are people that use it literally and inappropriately to deceitfully take advantage of others.
Gross negligence coupled with some form of loss is sufficient for general intent which makes the associated actions malicious/malice.
Throwing out the baby with the bath water without telling anyone or without warning, is gross negligence.
I'm not sure what to tell you. I'm a professional with nearly two decades of experience in this industry, and I don't read any white papers. I read web publications like Smashing Magazine or CSS Tricks, and more specifically authors like Paul Irish, Jake Archibald, Josh Comeau, and Roman Komarov. Developers who talk about the latest features and standards, and best practices to adopt.
The view that professionals in this industry exclusively participate in academic circles runs counter to my experience. Unless you're following the latest AI buzz, most people are not spending their time on arXiv.
The PSL is surely an imperfect solution, but it's solving a problem for the moment. Ideally a more permanent DNS-based solution would be implemented to replace it. Though some system akin to SSL certificates would be necessary to provide an element of third-party trust, as bad actors could otherwise abuse it to segment malicious activity on their own domains.
If you're opposed to Safe Browsing as a whole, both Chromium and Firefox allow you to disable that feature. However, making it an opt-in would essentially turn off an important security feature for billions of users. This would result in a far greater influx of phishing attacks and the spread of malware. I can understand being opposed to such a filter from an idealistic perspective, but practically speaking, it would do far more harm than good.
1 reply →
Putting user content on another domain and adding that domain to the public suffix list is good advice.
So good, in fact, that it should have been known to an infrastructure provider in the first place. There's a lot of vitriol here that is ultimately misplaced away from the author's own ignorance.
The PSL is something you find out about after it goes wrong.
It's a weird thing, to be honest, a Github repo mentioned nowhere in any standards that browsers use to treat some subdomains differently.
Information like this doesn't just manifest itself into your brain once you start hosting stuff, and if I hadn't known about its existence I wouldn't have thought to look for a project like this either. I certainly wouldn't have expected it to be both open for everyone and built into every modern internet-capable computer or anti malware service.
To be pedantic, the GitHub repo is not the source of truth, this is:
https://publicsuffix.org/list/public_suffix_list.dat
It even says so in the file itself. If Microsoft goes up in flames, they can switch to another repository provider without affecting the SoT.
If you don't know what you're doing and as a result bad things happen, that's on you.
I don't have a lot of sympathy for people who allow phishing sites suffering reputational consequences.
To be fair I’ve been in the space for close to 20 years now, worked on some of the largest sites and this is the first I’m hearing of the public suffix list.
Maybe it was effective from obscurity?
For something that you think is a de-facto standard, public suffix list seems kinda raw to me for now.
I checked it for two popular public suffixes that came to mind: 'livejournal.com' and 'substack.com'. Both weren't there.
Maybe I'm mistaken, it's not a bug and these suffixes shouldn't be included, but I can't think of the reason why.
I don't know about LiveJournal, but I don't believe you can host any interactive content on substack (without hacking substack at least). You can't sign up and host a phishing site, for instance.
User-uploaded content (which does pose a risk) is all hosted on substackcdn.com.
The PSL is more for "anyone can host anything in a subdomain of any domain on this list" rather than "this domain contains user-generated content". If you're allowing people to host raw HTML and JS then the PSL is the right place to go, but if you're just offering a user post/comment section feature, you're probably better off getting an early alert if someone has managed to breach your security and hacked your system into hosting phishing.
The public suffix list interferes with cookies. So on a service like livejournal, where you want users logged in across all subdomains, it's not an option
Exactly, this has been documented knowledge for many years now, even decades. Github and other large providers of user-generated content have public-facing documentation on the risks and ways to mitigate them. Any hosting provider that chooses to ignore those practices is putting themselves, and their customers, at risk.
> There's a lot of vitriol here that is ultimately misplaced away from the author's own ignorance.
For what it's worth, this makes it sound like you think the vitriol should be aimed at the author's ignorance rather than the circumstances which led to it, presuming you meant the latter.
I do think the author's ignorance was a bigger problem--both in the sense of he should have known better and also in the sense that the PSL needs to be more discoverable--than anything Google('s automated systems) did.
However, I'm now reflecting on what I said as "be careful what you wish for", because the comments on this HN post have done a complete 180 since I wrote it, to the point of turning into a pile-on in the opposite direction.
1 reply →
This is of course true! It just takes an incident like this to get ones head out of ones ass and actually do it. :)
The good news is, once known, a lesson like this is hard to forget.
The PSL is one of those load-bearing pieces of web infrastructure that is esoteric and thanklessly maintained. Maybe there ought to be a better way, both in the sense of a direct alternative (like DNS), and in the sense of a better security model.
1 reply →
This is the kind of thing that customers rely on you to do _before_ it causes an incident.
1 reply →
One can only imagine the other beginner mistakes made by this operator.
5 replies →
Since there's a lot of discussion about the Public Suffix list, let me point out that it's not just a webform where you can add any domain. There's a whole approval process where one very important criterion is that the domain to be added has a large enough user base. When you have a large enough user base, you generally have scammers as well. That's what happened here.
It basically goes: growing user base -> growing amount of malicious content -> ability to submit domain to PSL. In that order, more or less.
In terms of security, for me, there's no issue with being on the same domain as my users. My cookies are scoped to my own subdomain, and HTTPS only. For me, being blocked was the only problem, one that I can honestly admit was way bigger than I thought.
Hence, the PSA. :)
What sort of size would be needed to get on there?
My open source project has some daily users, but not thousands. Plenty to attract malicious content, I think a lot of people are sending it to themselves though (like onto a malware analysis VM that is firewalled off and so they look for a public website to do the transfer), but even then the content will be on the site for a few hours. After >10 years of hosting this, someone seems to have fed a page into a virus scanner and now I'm getting blocks left and right with no end in sight. I'd be happy to give every user a unique subdomain instead of short links on the main domain, and then put the root on the PSL, if that's what solves this
> [..] projects not serving more then (sic) thousands of users are quite likely to be declined.
from PSL's GitHub repo's wiki [0].
[0]: https://github.com/publicsuffix/list/wiki/Guidelines#validat...
Based on what I've seen, there's no way to get that project into the PSL. I would recommend you to have the content available at projectcontent.com if the main site is project.com, though. :)
> My cookies are scoped to my own subdomain
If you mean with the domain option, that's not really sufficient. You need to use the Host- prefix
As a CISO I am happy with many of the protections that Google creates. They are in a unique position, and probably the only ones to be able to do it.
However, I think the issue is that with great power comes great responsibility.
They are better than most organisations, and working with many constraints that we cannot always imagine.
But several times a week we get a false "this mail is phishing" incident, where a mail from a customer or prospect is put in "Spam", with a red security banner saying it contains "dangerous links". Generally it is caused by domain reputation issues, that block all mail that uses an e-mail scanning product. These products wrap URLs so they can scan when the mail is read, and thus when they do not detect a virus, they become defacto purveyors of virii, and their entire domain is tagged as dangerous.
I have raised this to Google in May (!) and have been exchanging mail on a nearly daily basis. Pointing out a new security product that has been blacklisted, explaining the situation to a new agent, etc.
Not only does this mean that they are training our staff that security warnings are generally false, but it means we are missing important mail from prospects and customers. Our customers are generally huge corporations, missing a mail for us is not like missing one mail for a B2C outfit.
So far the issue is not resolved (we are in Oct now!) and recently they have stopped responding. I appreciate our organisation is not the US Government, but still, we pay upwards of 20K$ / year for "Google Workspace Enterprise" accounts. I guess I was expecting something more.
If someone within Google reads this: you need to fix this.
I'm old. I've been doing security for a very long time. Started back in the 1990s. Here's what I have learned over the last 30 years...
Half (or more) of security alerts/warnings are false positives. Whether it's the vulnerability scanner complaining about some non-existent issues (based on the version of Apache alone... which was back ported by the package maintaner), or an AI report generated by interns at Deloitte fresh out of college, or someone reporting www.example.com to Google Safe Browsing as malicious, etc. At least half of the things they report on are wrong.
You sort of have to have a clue (technically) and know what you are doing to weed through all the bullshit. Tools that block access, based on these things do more harm than good.
What this post might be missing is that it’s not just Google that can block your website. A whole variety of actors can, and any service that can host user-generated content, not just html (a single image is enough), is at risk, but really, any service is at risk. I’ve had to deal with many such cases: ISPs mistakenly blocking large IP prefixes, DPI software killing the traffic, random antivirus software blocking your JS chunk because of a hash collision, even small single-town ISPs sinkholing your domain because of auto-reports, and many more.
In the author’s case, he was at least able to reproduce the issues. In many cases, though, the problem is scoped to a small geographic region, but for large internet services, even small towns still mean thousands of people reaching out to support while the issue can’t be seen on the overall traffic graph.
The easiest set of steps you can do to be able to react to those issues are: 1. Set up NEL logging [1] that goes to completely separate infrastructure, 2. Use RIPE Atlas and similar services in the hope of reproducing the issue and grabbing a traceroute.
I’ve even attempted to create a hosted service for collecting NEL logs, but it seemed to be far too niche.
[1]: https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/Net...
I don't see how a separate domain would solve the main issue here. If something on that separate domain was flagged, it would still affect all user content on that domain. If your business is about serving such user content, the main service of your business would be down, even though your main domain would still be up.
You are right, it would still affect all users. Until the pending PSL inclusion is complete, that is. But it now separates my own resources, such as the website and dashboard of statichost.eu from that.
A separate domain may not prevent users' conten from being blocked, but it may prevent blocking of the administrative interfaces. Which would help affected customers get their content and the service could more easily put a banner advising users of the situation, etc.
It feels like unless you're one of the big social media companies, accepting user content is slowly becoming a larger and larger risk.
It always was. You're one upload and a complaint to your ISP/Google/AWS/MS away from having your account terminated.
I second this for personal sites. Having run forums and chan sites without a CDN I found that not only is this true, it is 100% automated. The timing in emails to VPS/Registrars matches the times their scripts would crawl my sites and submit illicit content, screenshot it and automatically submit the screenshots to the VPS/server/registrar providers. That was incentive enough for me to take my sites private / semi-private. I would move them to .onion nodes but that's just too slow for me. I have my own theories as to what groups are running these scripts to push people to CDN's but no smoking gun.
Corporations are a little safer. They have mutually binding contracts with multiple internet service providers and dedicated circuits. They have binding contracts with DNS registrars. Having been on the receiving end of abuse@ they notify over phone and email giving plenty of time to figure out what is going on. I've never seen corporate circuits get nuked for such shenanigans.
Any services successfully offloading UGC to other moderated platforms? E.g. developer tools relying on GitHub instead of storing source/assets in the service itself, and Microsoft can take care of most moderation needs. But are there consumer apps that do things like this?
1 reply →
But something has definitely changed over the past few years. Back in the days, it felt completely normal for individuals to spin up and run their own forums. Small communities built and maintained by regular people. How many new truly independent, individual-run forums can you name today? Hardly any. Instead we keep hearing about long-time community sites shutting down because individuals can no longer handle the risks of user content. I've seen so many posts right here on HN announcing those kinds of shutdowns.
3 replies →
Your equally just one fake report to an automated system away having your account shut down. So, yes, your actions have consequences, but more worrying to me is the ability of someone with a grudge causing consequences for you as well.
3 replies →
Not sure who changed the HN headline, but I appreciate the change. Especially since the concept in the headline is buried at the bottom of the post.
Post author is throwing a lot of sand at Google for a process that has (a) been around for, what, over a decade now and (b) works. The fact of the matter is this hosting provider was too open, several users of the provider used it to put up content intended to attack users, and as far as Google (or anyone else on the web is concerned) the TLD is where the buck stops for that kind of behavior. This is one of the reasons why you host user-generated content off your TLD, and several providers have gotten the memo; it is unfortunate statichost.eu had not yet.
I'm sorry this domain admin had to learn an industry lesson the hard way, but at least they won't forget it.
Author here. I understand that my post and what I'm trying to say is unclear. And that there are too many different aspects to all this.
What I'm trying to say in the post specifically about Google is that I personally think that they have too much power. They can and will shut down a whole domain for four billion users. That is too much power no matter the intentions, in my opinion. I can agree that the intentions are good and that the net effect is positive on the whole, though.
On the "different aspects" side of things, I'm not sure I agree with the _works_ claim you make. I guess it depends on what your definition of works is, but having a blacklist as you tool to fight bad guys is not something that works very well in my opinion. Yes, specifically my own assets would not have been impacted, had I used a separate domain earlier. But the point still stands.
The fact that it took so long to move user content off the main domain is of course on me. I'm taking some heat here for saying this is more important than one (including me) might think. But nonetheless, let it be a lesson for those of you out there who think that moving that forum / upload functionality / wiki / CMS to its own domain (not subdomain) can be done tomorrow instead of today.
> In order to limit the impact of similar issues in the future, all sites on statichost.eu are now created with a statichost.page domain instead.
This read like a dark twist in a horror novel - the .page tld is controlled by Google!
https://get.page/
Thank you for this hacker-minded and sharp comment! <3 Seriously, not all comments on here are as fun to read for me as the author and fellow hacker.
And for what it's worth, it feels great to actually pay for something Google provides!
One way to build trust.
It can happen to anyone and cause a reputational risk. Once upon a time $workplace had a Zoho Form that would be blacklisted by Google Safe Browsing or Microsoft Edge for arbitrary periods of time, presumably because someone used Zoho to make a phishing site, leading to some very confused calls.
Sounds a very convenient mistake to do on your competitors? It does not sound believable that they would not know what zoho was or that it makes no sense to flag all the zoho domain.
Github discovered the same thing a long long time ago which is why you now have the github.io domain.
In Github's case, I think it was also because a lot of security boundaries were using TLD which led x.github.com potentially grab cookies of y.github.com or worse, github.com itslef
https://news.ycombinator.com/item?id=5500612
Don't forget the `githubusercontent.com` domain, which is specifically used to host risky, user-generated content, and fully documented in https://docs.github.com/en/authentication/keeping-your-accou... (using an open source component that other companies could also use, if they were interested in similar levels of security)
I am a solo developer. I recently created a new web app for a client. Google has marked as phishing so they can't use it. Obviously I can't do anything about it except report error and wait. I'm worried if I move it to a new domain that one will get marked as well. Not sure what to do TBH.
Is it phishing?
No, however it does include a Microsoft entra/Azure AD/Microsoft 365 login for that clients tenant. It is also a newly registered domain so I can understand why it looks suspicious. The most frustrating thing is that this is all a machine I.e. no-one I can speak to, nothing I can do to fix it. My fate has been decided by an algorithm.
Google has some sort of internal flag for determining origin is different on some platforms. We don't get a complete takedown of Neocities every time there's a spam site reported. It is likely that they were not on that list but perhaps have been manually added to whatever that internal list is at this point.
The public suffix list (https://publicsuffix.org/) is good and if I were to start from scratch I would do it that way (with a different root domain) but it's not absolutely required, the search engines can and do make exceptions that don't just exclusively use the PSL, but you'll hit a few bumps in the road before that gets established.
Ultimately Google needs to have a search engine that isn't full of crap, so moving user content to a root domain on the PSL that is infested with phishing attacks isn't going to save you. You need to do prolific and active moderation to root out this activity or you'll just be right back on their shit list. Google could certainly improve this process by providing better tooling (a safe browsing report/response API would be extremely helpful) but ultimately the burdon is on platforms to weed out malicious activity and prevent it from happening, and it's a 24/7 job.
BTW the PSL is a great example of the XKCD "one critical person doing thankless unpaid work" comic, unless that has changed in recent years. I am a strong advocate of having the PSL management become an annual fee driven structure (https://groups.google.com/g/publicsuffix-discuss/c/xJZHBlyqq...), the maintainer deserves compensation for his work and requiring the fee will allow the many abandoned domains on the list to drop off of it.
If you're not using separate domains then I hope you don't have any kind of sensitive information stored in cookies. You can't rely on the path restrictions for cookies because it's easily bypassed.
You can set cookies that strictly stay on the root domain and don't cross to subdomain origins, and vise versa (https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/Coo...). We've been doing this for 12 years without issue.
Strict cookies crossing root to subdomains would be a major security bug in browsers. It's always been a (valid) theoretical concern but it's never happened on a large scale to the point I've had to address it. There is likely regression testing on all the major browsers that will catch a situation where this happens.
Hosts phishing sites, gets blocked by anti phishing mechanism. Works as expected from my point of view.
Get yourself on public suffix list or get better moderation. But of course just moaning about bad google is easier.
If youtube.com doesn't end up on the Safe Browsing blacklist because of phishing videos, but your own website can easily end up there, it's a pretty clear case of Google abusing their power.
YouTube doesn't allow you to put your credentials into text box and hit send. Google sites, on the other hand, does pose a disk, but they'll likely be treated the same as any other domain on the PSL.
In my experience, safe browsing does theoretically allow you to report scams and phishing in terms of user generated content, but it won't apply unless there's an actual interactive web page on the other end of the link.
There is the occasional false positive but many good sites that end up on that list are there because their WordPress plugin got hacked and somewhere on their site they are actually hosting malware.
I've contacted the owners of hacked websites hosting phishing and malware content several times, and most of the time I've been accused of being the actual hacker or I've been told that I'm lying. I've given up trying to be the good guy and report the websites to Google and Microsoft these days to protect the innocent.
Google's lack of transparency what exact URLs are hosting bad material does play a role there.
What is a phishing video?
1 reply →
You are right, of course. I'm not sure if those of you who disagree with me think that Safe Browsing did its job (which it did!), that Safe Browsing is a good thing (which it maybe is, but which I slightly disagree with), or that it's ok that Google monitors everything everyone does.
The last point is actually the one I'm trying to make.
There should be a concept, sort of an inverse of tragedy of the commons, for the positive feedback loop of many users providing big data to a company that can use that data to benefit many users.
From spamblocking that builds heuristics fed by the spam people manually flag in GMail to Safe Browsing using attacks on users' Chrome as a signal to their voice recognition engine leapfrogging the industry standard a few years back because they trained it on the low-quality signal from GOOG411 calls, Google keeps building product by harvesting user data... And users keep signing up because the resulting product is good.
This puts a lot of power in their hands but I don't think it's default bad... If it becomes bad, users leave and Google starts to lose their quality signal, so they're heavily incentivized to provide features users want to retain them.
This does make it hard to compete with them. In the US at least, antitrust competition has generally been about user harm, not actually market harm. If a company has de-facto control but customers aren't getting screwed, that's fine because ultimately the customer matters (and nobody else is owed a shot at beeing a Google).
1 reply →
It's hard to get that point because you're conflating two different stories.
Folks around here are generally uneasy about tracking in general too, but remove big brother monitoring from Safe Browsing and this story could still be the same: whole domain blacklisted by Google, only due to manual reporting instead.
"Oh, but a human reviewer would've known `*.statichost.eu` isn't managed by us"—not in a lot of cases, not really.
2 replies →
You are right, but then again, nobody flags facebook because of the scamming taking place in some facebook pages.
Generally because Facebook polices Facebook (imperfectly, but the effort is demonstrated) and the damage radius is limited to Facebook users mostly. As long as the easiest way to avoid damage from the Facebook domain is "Don't use Facebook," the larger Internet doesn't need a mechanism to police it.
If Facebook became a trap that frequently hosted malware to strangers, the rest of the net would begin to interpret it as damage and route around it.
"Might makes right" as they say.
There is no real way a normal person even can flag facebook.
Google services simply behaved the way I would expect them to here. Who knows... they may even have saved some users from coming to harm.
Many phishing attacks originate from Google's owns domains. Gmail users phishing others, scammy youtube videos, scammy comments with links to scams, scammy ads to fake banking pages, etc, etc. But Google would never be hypocritical, never!!
That is a great point. When I see these sites I'm always seeing a dozen red flags, and maybe the biggest one is that it's showing a "NatWest" banking site or something and is hosted on "portal-abc.statichost.eu". But the whole point is of course saving users from coming to harm, and if it did - great!
I’ve got a random subdomain hosting a little internal tool. About twice a year, Google Safe Browsing decides it’s phishing and flags it. Sometimes they flag the whole domain for good measure.
Search Console always points to my internal login page, which isn’t public and definitely isn’t phishing.
They clear it quickly when I appeal, and since it’s just for me, I’ve mostly stopped worrying about it.
I encountered something similar. I have `*.domain.tld` pointed to an internal IP address, and over the past few years it happened a few times where some subdomain would be flagged as dangerous by Google Safe Browsing.
Internal IP addresses in public DNS are sometimes used to do things like DNS rebind attacks. It's possible that's tripping up their detection mechanism.
My workaround is to use an IPv6 ULA for my publicly hosted private IP addresses, which is extremely unlikely to ever be reused by a bad actor.
I was curious how other browsers handle this. Apparently Safari and Firefox delegate to Google.
https://www.apple.com/legal/privacy/data/en/safari/
https://support.mozilla.org/en-US/kb/how-does-phishing-and-m...
Microsoft seems to do its own thing for Edge, though.
https://learn.microsoft.com/en-us/deployedge/microsoft-edge-...
This is a bit of a tangent, the whole concept of "domain reputation" can be infuriating. For example, my blog has been marked as suspicious by spamhaus.org: https://check.spamhaus.org/results?query=dynomight.net
As a result, some ISPs apparently block the domain. Why is it listed? I have no idea. There are no ads, there is no user content, and I've never sent any email from the domain. I've tried contacting spamhaus, but they instantly closed the ticket with a nonsensical response to "contact my IT department" and then blocked further communication. (Oddly enough, my personal blog does not have an IT department.)
Just like it's slowly become quasi-impossible for an individual to host their own email, I fear the same may happen with independent websites.
From reading that my guess would be that the IP of your host gotten from your hosting provider had some spammy history before you started hosting your blog on it.
Either that or your DNS provider hosts a lot of spam.
Hmmm, I use https://njal.la/ for DNS. Could spamhaus really just auto-mark every njalla user as suspicious?
1 reply →
> As a result, some ISPs apparently block the domain
This is the infuriating part. I get that someone buying cheap hosting may end up with an IP address that used to send spam, but spam lists are not reliable indicators of website security.
Overzealous security products are a blight on the internet. I'd be less annoyed at them if they weren't so trivial to bypass as a hacker with access to a stolen credit card.
It's also good from a security perspective.
Anyone who can upload HTML pages to subdomain.domain.com can read and write cookies for *.domain.com, unless you declare yourself a public suffix and enough time has passed for all the major browsers to have updated themselves.
I've seen web hosts in the wild who could have their control panel sessions trivially stolen by any customer site. Reported the problem to two different companies. One responded fairly quickly, but the other one took several years to take any action. They eventually moved customers to a separate domain, so the control panel is now safe. But customers can still execute session fixation attacks against one another.
(Author here) This is all true. The main assumption from my part is that anything remotely important or even sensitive should be and is hosted on a domain that is _not_ companysubdomain.domain.com but instead www.company.com.
So… you were hosting user generated content on the same TLD as your website, without using the PSL, and you blamed G when things went south?
By putting UGC on the same TLD you also put your own security at risk, so they basically did you a favor…
Many commenters are implying that there is a security issue here, and that I'm putting everyone in danger. That is quite frankly a pretty absurd claim to just casually make. I'm of course very curious to hear more details on what the security risk here actually would be?
Do you think I'm reading/writing sensitive data to/from subdomain-wide cookies?
Also, yes, the PSL is a great tool to mitigate (in practice eliminate) the problem of cross-domain cookies between mutually untrusting parties. But getting on that list is non-trivial and they (voluntary maintainers) even explicitly state that you can forget getting on there before your service is big enough.
I am not implying you’re putting “everyone” in danger. I’m merely implying that you’re putting your own service in danger by allowing clients to act like a trusted subdomain like controlpanel.statichost.eu, .secure, or Unicode similarities of www.
1 reply →
Super interesting article! We were facing the exact same problem last Monday . Also write a couple of words about the incident: https://muetsch.io/how-google-accidentally-took-us-off-the-i....
Does anyone know if adding our domains to Public Suffix List will prevent incidents like this?
I don't like nor trust google, but "Use your own judgement and hard-earned Internet street smarts" doesn't work either, because the median internet user does not have anything resembling internet street smarts.
Still not sure why it's legal for Google to slander companies like this. They often have no proof or it's a false positive, meanwhile they're screaming about how malicious you are.
My site, a library for building services with the GOV.UK Design System, is currently blocked too.
Despite being a paying Google Workspace customer, I can't get in touch with anyone who can help.
https://github.com/x-govuk/govuk-components/
Notably this post did not examine whether any of the sites it was hosting on this domain was malicious/misleading.
I'm not asking about this specific case. There are plenty of examples of Google wrongly accusing others of being malicious with massive business impact
Because Google has absolutely nothing to lose and you do, besides that they can outlast anybody except for nation states in court.
How does this answer the question of legality?
3 replies →
Good question, it probably shouldn't be legal. Enforcing the law on these behemoths is another problem, though... :)
Seems like a reasonable trade-off I mean six hours is not the worst thing in the world. What if you were hosting mission-critical such as such? Were you?
> To be fair, many or even most sites on the Google Safe Browsing blacklist are probably unworthy. But I’m pretty sure this was not the first false positive.
The bigger issue is that the internet needs governance. And, in the absence of regulation, someone has stepped in and done it in a way that the author didn't like.
Perhaps we could start by requiring that Google provide ways to contact a living, breathing human. (Not an AI bot that they claim is equivalent.)
why do you assume that the living, breathing human hired by theGoogs will be competent at handling all of the crazy that will be flung at them by the living, breathing human on the other end of the line. One single person cannot handle that. Naturally, you need a team of living, breathing humans. You might even have them in triage level groups like level 1 support, level 2 support and so on where each level is a more trained/experienced living, breathing human. Eventually, you'll have an entire department of people of varying degrees of skill. Oh, wait, I'm sorry, I thought it was the year 2000.
Hopefully, this helps you understand why your living, breathing human is such a farcical idea for theGoogs to consider.
Well, Google did self-appoint itself the "internet police," and the general job of the police is to deal with screwballs.
So you can't take one part of the responsibility and abdicate the other part!
1 reply →
You really think talking to a human and a bot is the same?
2 replies →
Honestly, this is extremely basic stuff in hosting, not only due to safe browsing, but also—and more importantly—cookie safety, etc. If a hosting provider didn’t know (already bad enough) and turn to whining after being hit, then
> Static site hosting you can trust
is more like amateur hour static site hosting you can’t trust. Sorry.
The thing is, you cannot just add any domain to the PSL. You need a significant amount of users before they will include your domain. Before recently, there really was no point in even submitting, since the domain would have been rejected as too small. An increase in user base, increase in malicious content and the ability to add your domain to the PSL all happen sort of simultaneously.
I'm also trusting my users to not expose their cookies for the whole *.statichost.eu domain. And all "production" sites use a custom domain anyway, which avoids all of this anyway.
There are well-documented solutions to this that don't rely on the PSL. Choosing to ignore all of that advice while hosting user content is a very irresponsible choice, at best.
1 reply →
If user1.statichost.page gets blacklisted now will it affect user2.statichost.page as well?
Yes, unless they submit statichost.page to the public suffix list.
I have recently had the pleasure of speaking with Google senior leadership involved in the Safe Browsing product on the topic of getting my SaaS product placed on their, "naughty list." The platform was down for 6 or so hours due to a false positive hit for phishing.
I have read A LOT of blogs/rants/incidents on social media about startups, small businesses, and individuals getting screwed by large companies in similar capacities. I am VERY sympathetic to those cries into the sky, shaking fists at clouds, knowing very well we are all very small and how the large providers seem to not care. With that in mind, I am not blind to the privilege my organization has to rope in Google to discuss root causes for incidents.
I am writing about it here because I believe most people will never be able to pull a key Google stakeholder into a 40 minute video call to deeply discuss the RCAs. The details of the discussion are probably protected by NDA so I'll be speaking in general terms.
Google has a product called Web Risk (https://cloud.google.com/web-risk/docs/overview), I hear it's mostly used by Google Enterprise customers in regulated verticals and some large social media orgs. Web Risk protects the employees of these enterprise organizations by analyzing URLs for indicators of risk, such as phishing, brand impersonation, etc.
My SaaS platform is well established and caters mostly to large enterprise. I provide enterprise customers with optional branded SSO landing pages. Customers can either use sign-in from the branded site (SP-initiated) or redirect from their own internal identity provider to sign-in (IdP-initiated). The SSO site branding is directed by the customer, think along the lines of what Microsoft does for Entra ID branded sign-in pages. Company logo(s), name, visual styling, and other verbiage may be included. The branded/vanity FQDN is (company).productname.mydomain.com.
You may be able to see where I'm headed at this point... Why was my domain blocked? For suspected phishing.
A mutual enterprise customer was subscribed to Google's Web Risk. When their employees navigated to their SSO site, Google scanned it. Numerous heuristics flagged the branded SSO site as phishing and we were blocked by Safe Browsing across all major web browsers (Safari, Chrome, Firefox, Edge, and probably others). Google told us that had our customer put the SSO site on their Web Risk allow-list, we wouldn't have been blocked.
I'm no spring chicken, I cannot rely nor expect a customer to do that, so I pressed for more which led to a lengthy conversation on risk and the seemingly, from my perspective, arbitrary decisions made by a black box without any sort of feedback loop.
I was provided a tiny bit of insight into the heuristic secret sauce, which led to prescribed guidance on what could be done to greatly reduce the risk of getting false positive flag for phishing again. Those specifics I assume I cannot detail here, however the overall gist of it is domain reputation. Google was unable to positively ascertain my domain's reputation.
My recommendation is for those of you out there in the inter-tubes who have experienced false positive Safe Browsing blocks, think about what you can do to increase your domain's public reputation. Also, get a GCP account so if you do get blocked, you can open a ticket from that portal. I was told it would be escalated to the appropriate team and be actioned on within 10-15 minutes.
I have the same issue. Think of my site as WeTransfer, but instead of only files, you can also use it as a link shortener or pastebin. Abuse works the same as on every other site or service: I do spot checks and users can report content. This was fine until uBlock Origin decided the website was malicious, per one of the lists that is default-enabled for everyone
That list doesn't have a clear way to get off of it. I would be happy to give them the heads up that their users are complaining about a website being broken, but there is no such thing, neither for users nor for me. In looking around, there's many "sources" that allegedly independently decided around the same day that my site needs to not work anymore, so now there's a dozen parties I need to talk to and more popping up the further you look. Netcraft started sending complaints to the registrar (which got the whole domain put on hold), some other list said they sent abuse to the IP space owner (my ISP), public resolvers have started delisting the domain (pretending "there is no such domain" by returning NXDOMAIN), as well as the mentioned adblockers
There's only one person who hasn't been contacted: the owner. I could actually do something about the abusive content...
It's like the intended path is that users start to complaint "your site doesn't work" (works for me, wdym?) and you need to figure out what software is it they're using, what DNS resolver they use, what antivirus, what browser, if a DOH provider is enabled... to find out who it might be that's breaking the site. People don't know how many blocklists they're using, and the blocklists don't give a shit if you're not a brand name they recognize. That's the only difference between my site and a place like Github: if I report "github.com hosts malware", nobody thinks "oh, we need to nuke that malicious site asap!"
I'd broaden the submitted post to say that it's not only Google with too much power, but these blocklists have no notification mechanism or general recourse method. It's a whack-a-mole situation which, as an open source site with no profit model (intentionally so), I will never win. Big tech is what wins. Idk if these lists do a trademark registration check or how they decide who's okay and who's not, but I suspect it's simply a brand name thing and your reviewer needs to know you
> Luckily, Google provided me with a helpful list of the offending sites
Google is doing better than most others with that feature. Most "intelligence providers", which other blocklists like e.g. Quad9 uses, are secretive about why they're listing you, or never even respond at all
Another day, another IT company learning the hard way about the public suffix list, or well-known URIs, or some other well-documented-but-niche security technology.
I love that IT is a field where there's no required formal education track and you can succeed with any learning path, but we definitely need some better way to make sure new devs are learning about some of these gotchas.
Wow 19€ for web hosting. I pay like 8 for whole vps. Crazy
The PSA is good, the article is meh. There is too much misdirected anger towards google here, IMO. I agree it sucks to be the false positive, but it'd also suck more to unknowingly be part of phishing campaigns and not know.
On top of that, it is also recommended to serve user content from another domain for security reasons. It's much easier to avoid entire classes of exploits this way. For the site admins: treat it as a learning experience instead of lashing out on goog. In the long run you'll be better off, having learned a good lesson.
Exactly! For a web dev in 2025 to still not know security best practices that have been around for 20+ years is a failure on the part of the dev.
I’m sure I don’t know ALL the "security best practices that have been around for 20+ years" and this is perfectly fine as long as I’m able to react quickly. See also https://xkcd.com/1053/.
4 replies →
How dare google tag my website unsafe just because im hosting a bunch of phishing sites!
Like i get that google has a lot of power, but you think they would use a case where google was actually in the wrong.