> In my experience the kinds of people who run this exact combination of circumstances also tend to be the kind of people that have a wide variance in the level of kindness they display to the authors of open source programs that happen to be in their way.
Love this. There’s been past discussion on HN re how OSS maintainers are treated, and this is such a wry sentence. I really appreciate the tone / attitude to the problem.
Well a little translation would be that some are very kind, some are total assholes, and everything in between. But your reading is probably more accurate.
I think the literal interpretation is correct here. It's a way of saying "you, dear reader who runs this combination of things, might be very kind, but some of your compatriots will be unusually unkind and will turn up in my issue tracker and on my socials to complain".
> something I was doing with my "strict MVP" build of Anubis' WASM wasn't in fact sticking to just the MVP features of WebAssembly
Fun fact, this _used_ to be the case - wasm32-unknown-unknown had extra non-mvp features added later, which in my eyes has been a breaking change on stable. You're not the first person to have been bitten by this, there's [1] and a similar story in Ruffle [2].
Weirdly, the response from the rustc side (IIRC we also talked with one of the devs of wasm component on discord a bit) has contained a lot of (paraphrasing) "in retrospect we think wasm32-unknown-unknown target was designed badly*, so we'd prefer if you either dealt with it or switched to another wasm target <with its different set of downsides> than us fixing it". (*I acknowledge that the way std was implemented on that target was unusual and "not clean", but it works _really well_ for a lot of people.)
> I looked into the process involved for rebuilding the standard library twice: once with only MVP wasm features enabled and once with an "all yes config" like usual. Based on some research I did this seemed like a massive pain.
That's what Ruffle does [3], it's a bit unfortunate but wouldn't call it a massive pain.
I have every copy of Firefox here configured with webassembly disabled...because I don't tend to do what Webassembly was designed for, i.e. online games, video/audio editing, emulation, etc. [1] and because I dislike things running in the background without my knowledge.
So this is going to be interesting.
My plea (in a small voice) to the OP: just as many sites still do for JavaScript, please add a "This captcha required Webassembly to continue" message to your code when Webassembly is disabled. And, incidentally, for smaller platforms/browsers that don't offer a Webassembly engine at all.
The second half of this post chronicles the author's Herculean efforts to make everything work well in the cases you describe.
> One of the big things that blocked this shipping for so long was not having an escape hatch of some kind to allow clients that disable WebAssembly by policy to get through the gate. [...] This combination of factors means that there would need to be another implementation of the proof of work code in JavaScript that would actually execute the number crunching.
Thanks. You're right; it was a very thorough article and I trailed off about halfway through. I still say there are many improper uses of Webassembly out there, but I credit the developer for a highly thoughtful approach to the project. I really wonder if Cloudflare et al. does as much due diligence?
I really don't see why you'd disable WebAssembly and allow JavaScript instead. Expecting developers to make a fallback for a baseline-supported feature for any reason but backwards compatibility with runtimes that do not so support WASM is extremely silly.
Hats off to Xe for spending so much time on backwards compatibility, especially the tidbit about targeting Chrome 66.
I have a Mac from 2014 running Yosemite that I occasionally use to test for backwards compatibility in my own frontend code (for fun!). But IMO the best way to ensure compatibility is to use period-correct toolchains or toolchains where the pace of change is slower, like ClojureScript.
Is there a place where I can try out if my browser is compatible? Easier to find out now than when I'm trying to get work done and a million websites now have it deployed
On https://wasm-feature-detect.surma.technology it shows that I don't have 3 of all these features but I'm not sure if Anubis needs any of them to not kick me back to the pure JS solution
Which would apparently be bad because
> The WebAssembly that's shipped with this flow is ridiculously performant. This may mean you need to adjust the difficulty [upward, to avoid that bots solve it trivially when they support wasm, I assume this means]
combined with
> The wasm2js flow doesn't currently have a way to update the progress bar [so you have no clue about remaining time]
(wouldn't be the first time that I gave up on a page because it was stuck on 0 hashes per second)
It would be cool if the POW could be done ahead of time. That way I don't get stuck waiting while I'm working. Some type of credit/tokens that my browser could then spend.
That just creates another problem: if you're taking any measures to reduce tracking (ie. clearing cookies on shutdown or using temporary containers), this won't work. If anubis was being deployed on a site that a user visits often (eg. HN), the user might be convinced to whitelist it, but most anubis deployments are on random blogs or fediverse instances that I might not visit again in months. I'm certainly am not going to whitelist those sites, nor am I going to enable cookies wholesale just to avoid solving challenges.
Look, if you're going out of your way to break expected behaviour on websites you shouldn't be surprised when people code to the most common denominator and then you have weird subtle breakage as a result.
Gotta agree. Turning tracking cookies off? Sure! Turning session cookies off and then complaining that the server doesn’t remember you? Uhhh…
And while I’m sympathetic to the idea of not wanting to run JS, to a first approximation modern browsers are JS engines that have graphical displays. How things should be vs how they are is a classic is/ought problem. The world took a vote on what a browser’s meant for and we lost. Fighting it today is rough; tomorrow, futile.
There's no mechanistic difference between the two types of cookies, let's not pretend this is an actual config choice that people are making.
Whenever I hit Anubis, I simply go "keep your secrets then" and take my leave. The vast majority of the sites posted to HN (that catch my fancy) work fine or fine enough, and are better for it.
If I really, really desperately want to check something out, I can always just turn stuff back on. Turns out, I rarely do.
Shoutout to the particularly dogshit few that don't just require cookies and JS, but even third party JS. For reading a blogpost or a message thread!
Then we could implement an exchange, so that if you generate too much Anubis POW, you could exchange it with others.
We could link it to a site, you generate for HN, I for Reddit, but it so happens that you visited Reddit more and I HN, so we depleted our Anubis POW, so we could exchange some Reddit Anubis with some HN Anubis.
The interesting part is that bots mitigation is actually an economic problem. You don't need to make scraping impossible, you only need to just make it expensive enough so that abusive traffic stops being worth it.
The git host example is probably the one which Anubis is the dumbest defense for: the main reason to have those links is for easy machine interaction. So that's dead when you implement it.
So why even host the links? It's a git repo: send them a local got client and let them clone the repo in browser or something.
Not to mention it is the least efficient representation of git commits. It seems a lot of people are unaware that git generates the diff representation you see with "git show" on the fly -- internally, commits are Merkle trees representing a snapshot of the repo file contents.
As you said, if you actually want to interact with a remote git repo you'd do so with packfiles not diffs.
> It also means that the "hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead.
I assume that the previous challenges will still be available for circumstances where WASM is not available (unless we are just working on the basis that it is available if JS is?) or perhaps as a fallback. TFA explicitly mentions “smart” TVs, and I'm sure there are plenty of those out there that run browsers old enough to not know what WASM is. TBH these would not be a priority for me, but they obviously are for this project given they (well, TVs generally, not specifically elderly ones) get specific mention in this write-up.
If so, what is to stop scrapers from just forcing the fallback POW option instead of trying to run the new memory-hard algorithm at all?
Ah I must have skipped by that bit. And presumably the memory-hard algorithm is CPU-easy enough that this won't be a problem. That leaves me wondering "why not just use a JS memory-hard routine in the first place, but maybe I've shouted over that too. I'll give TFA a reread later.
I think it is still compute-heavy if not compute-hard, and stuff like WASM SIMD makes it less compute heavy (not really available in JS). And also, for memory bound problems, you probably don't want the JS GC needing to clean up afterwards when WASM gives you that control.
I assume that SIMD acceleration is usable with WebCrypto, but I have no evidence either way. It's also not part of WebAssembly so there's that to consider too lol
Others have answered your question. A lot of the sites listed here https://anubis.techaro.lol/docs/user/known-instances/ have dynamic content; Git web interfaces in particular (Codeberg, the Linux kernel, FFMPEG, and more are on the list) are vulnerable to poorly or maliciously configured scrapers.
I'm not a fan of Anubis for various reasons but the idea that bot traffic in only harmful with dynamic pages must die. CPU (yes, even to serve static pages) is not free, bandwidth is definitely not free. There's an idea that serving a static page to a bot has a marginal cost tending to zero, but it's never really zero and serving them by millions definitely has a cost.
Also, while some pages may look like static HTML pages, they may be generated on the fly by an expensive and/or slow backend, which adds to the cost. I happen to maintain servers for academics and some content management systems are slow and have an expensive CPU cost. While it's OK for the low number of humans interested in the subjects they deal with, it's definitely not fit for massive bot scrapping. And before you ask, no, it's not always practical to have cache upfront or to pre-generate all pages.
There are numerous services that will let you host static pages for free or nearly free. There are also numerous services that sit in front of your website that can block bots and reduce load on your origin server, many of which are also free, or very low cost relative to the service they provide.
The situation you are in is far less dire sounding when you consider that you have these options available to you.
Shitty scrapers hitting images over and over again is also a problem. Not just /pic/1.webp - /pic/2.webp, but in practice I've seen three AI scrapers that would request things like /pic/1.webp?version=webp&foo=bar and other random query parameters that none of my domains even use. It's just stuck brute-forcing files or something, sucking up loads of bandwidth downloading the same image for no reason.
It also depends on if you want your content to be scraped. Any bot ignoring robots.txt should be considered potentially malicious and worth considering such mechanisms for.
The anubis docs website is used as an example of how anubis works.
Also, if you are not using a big CDN (e.g. Cloudflare), you still need to pay for the traffic (even if you don't pay as much for the CPU). This is doubly true for smaller CDNs like Bunny.
How much data does it download off the server to run the challenge? I've made some tiny webassembly binaries before, but I've never seen Rust generate something tiny.
Anubis is fine if you want to block bots that are bothering you, but I never understood why they don't have robots.txt at the root level too. I happily obey that for my private crawler. I suppose you don't want to duplicate implementations. Fair enough. I guess I should add an Anubis detector so I can just blacklist because that person clearly doesn't want bots.
and the latter (at least) is Anubis protected. It could be just an omission. But if it's intentional, I am just curious why so that as a bot author I can be well-behaved.
Great article, WebAssembly is such a good technology when you have use case for it.
I've just finished creating a WASM 1.0 implementation (simple JIT on x86 and x86_64) for plugin usage and I'm looking on how to support more languages beyond C (for which I have "micro-libc" to create really small binaries). In particular C++ and Rust so the article contains useful info for that. C++ that would be more problematic I think, would have to resort to full Emscripten or something... I've tried to compile standalone C++ stdlib but wasn't much sucessful yet.
BTW, I see in the screenshots you also test Firefox, I think it would make sense to test the latest two ESR versions (these are often used by Linux distributions).
It's still very unclear to me how Anubis is supposed to work in the long term or even the medium term. The assumption is that scrappers, a.k.a. AI companies, a.k.a. those who hog all the RAM, are short of memory...?
The assumption is that requiring 1 second of compute before allowing a page load will have more impact on the people requesting 1000 pages/sec than it will on consumers requesting 1 page every minute.
Personally I don’t understand where this confusion comes from - it’s a simple economic tactic. Yes the large scale scrapers won’t run out of compute but putting anubis in front of a website demonstrably reduces the amount of crawling and that is the entire goal.
> requiring 1 second of compute before allowing a page load
After 1 solve, you get a cookie and can load tons of pages with it. Scrapers and user agents alike will need to spend the compute only once to get a valid session for themselves
Seems to be working so far though. Fwiw, lots of silly things stop bots until someone puts their mind (or tokens) to bypassing it, and then the blocking mechanism has to adapt. We'll see how it plays out, for now it's a lot better than begging big brother (cloudflare, google, or amazon captcha) for access, so I'm quite appreciative of what Techaro is doing
The issue is that scrapers are rotating their IP and essentially performing a DDoS attack. Anubis is part of a defense-in-depth solution. If scrapers reuse a cookie, traditional anti-abuse mechanisms will work.
>If scrapers reuse a cookie, traditional anti-abuse mechanisms will work.
Scrapers can't change their strategy? It's not hard to think of mitigations: keep on using a session until you're banned, and failing that rotate your IP. Nowadays you don't even need to think. You can probably tell a LLM "write me a scraper that has adaptive anti-anti-bot mechanism".
Yes, but you normally don't try to fight all strategies up front, you add stuff on a need-to basis. Currently, lots of people are in the stage of "need to get this thing using bunch of IPs to stop hammering my servers", apparently Anubis helps them with that. If that stops working, other tactics are tried, and so it goes on for the life of your service. Hasn't stopped yet, doesn't seem to be stopped soon either.
That argument also works against anubis. Why bother forcing clients to solve a PoW challenge, when you're fighting a bunch of unsophisticated attackers that will be deterred by a javascript challenge that requires them to solve `result = x + y`?
Point is, you get problems because of something like this, you try either approach. And when that stops working (because eventually it will), you add something else. Some work better against some actors, but also feels like it never ends, so slowly you add more things. I don't think Anubis is uniquely able to stop that particular type of annoyance, it's just one of the tools you can see if it helps. If you want to try something else before, makes a lot of sense to do so.
You're not fighting unsophisticated attackers. 50% of them are, maybe. The rest? They're actively hostile, detecting blocking attempts, moving on to botnets, DDoSing you and trying everything they can to bypass you.
People have been sent to jail for participating in LOIC attacks which were less destructive that what these bots do on the daily.
Caution: Techaro is very hostile to this line of thinking. He believes the PoW is essential. And that's fine - different projects use different approaches, like go-away and dogwaf.
Plenty of these abusive scrapers are utilizing retail residential proxies, which will be applying forced rotations to avoid "burning" their compromised and or otherwise surreptitiously utilized IP address.
I would disagree with the Anubis effectiveness falling off, but the llm bros are throwing life itself into training to defeat any “perceived” challenge.
Yep, on my aging phone it takes quite w noticable amount of time to go through. But on my more recent laptop wnd workstations, it's a quick flash of a manga lolita.
You could just spend 30 seconds googling instead of confidently asserting nonsense. There are plenty of people demonstrating that Anubis reduces bot traffic.
That’s because its weird and mostly irrelevant, it’s nothing to do with the proof of work being any sort of economic impediment. It takes Claude a few minutes to write a near instant solver, that’s the maximum proof of work cost this adds.
This strain of nihilism about anti-bot efforts baffles me. Cloudflare built a gigantic business on stopping DDoS attacks. There will never be good faith self-restraint by AI bros, there is too much money at stake and no regulation, doing nothing means your only alternative is disconnect from the Internet. Anubis introduces enough friction that some percentage of the lunacy is throttled and possibly give your server a fighting chance.
>There are plenty of people demonstrating that Anubis reduces bot traffic.
"Anubis reduces bot traffic" and "[the proof of work] is not stopping anyone" isn't contradictory, especially if anubis isn't being benchmarked against non-POW alternatives like IP/ASN bans, or tls fingerprinting.
>The assumption is that requiring 1 second of compute before allowing a page load will have more impact on the people requesting 1000 pages/sec than it will on consumers requesting 1 page every minute.
Which doesn't hold. A 1s delay, despite being inperceptible is still a cost to human users. After all, 3600 x 1s delay adds up to an hour. Worse yet, the typical human visitor will always have to solve a challenge, because chances are it's the first time they're visiting your blog or whatever. Someone scraping reuse a valid session for at least a few more pages.
There's also the problem of time. It might be tempting to think the cost of a challenge is 1M cpu cycles or whatever, but the cost is better modeled as two parts, a time cost (for the human) and a cpu cost (for the computer). A scraper might have to pay the cpu cost, but not the time cost, because they can have their scraper solve challenges in their sleep. Unless the user is utterly destitute, their time is worth far more than whatever the compute cost is.
The assumption is that there will always be some population distribution in terms of effort expended as well as in terms of problems caused. At present it's the lowest effort actors that are causing the highest number of problems. So an 80% (or even just a 20%) solution largely solves the practical problem faced by site operators.
If against all odds it turns out that skiddies (or agents or AI billionaires or what have you) are willing to burn inordinate amounts of RAM and CPU just to scrape autogenerated cgit pages that are of approximately zero value to them to begin with then I guess we'll just have to cross that bridge when we come to it.
Regular users store the cookie and do the challenge once. Bots do it over and over again throwing out the cookies so they're not rate-limited.
Extreme configurations with high difficulty do make for a frustrating experience on very old devices, but in practice it rarely gets in the way.
Only the tail end of what one could consider normal users (those running custom browsers or having JS disabled or nonfunctional) run into issues here, but they'll have these issues with any other bot-blocking mechanism that still offers some protection these days.
Apple's "ask your vendor for tokens to exchange for access" web standard gets around this, assuming those bespoke and Javascript-less browsers implement PATs, but that standard seems to be completely dead in the water outside of Safari.
> And then regular users: who now can't really browse your site.
That's entirely up to how the site operator configures the PoW challenge.
> anyone with even a moderate amount of funding in the AI space, who aren't even slightly slowed down
Do you have any evidence that it is the large well funded AI companies DoSing sites by (for example) scraping absurd numbers of permutations of cgit url parameters?
Even if it is (which I doubt), if all this accomplishes is getting them to spend 10 minutes optimizing their bots to not do that it will be a major win.
I think the Rust feature you’re looking for regarding recompiling the standard library is called “build-std”, that should be enough for you to search for it. (For similar reasons you also need that flag if you are trying to use Rust to build multithreaded wasm binaries, so it might come up for you!)
It's wild to me that PoW systems are how we sort the bot problem... the bummer is that all this "work" is just wasted cycles, at least in crypto there is a token you can sell.
Scraping would be costly in this world: scrapers would have to spend tokens in order to get the webpage.
But, in this world, the website owner would receive tokens that they can then use to do whatever they want, including paying for servers and bandwidth. This is the sense in which the cycles aren’t wasted: the website owner now has cash to spend.
Effectively, both scrapers and ordinary users would be paying for the privilege of getting website bytes.
This also solves the problem of having to wait for your phone to solve the challenge while you’re browsing: you can buy or mine some tokens ahead of time and pay them as soon as challenged. So can the scrapers, but because they’re accessing enormous numbers of pages it’s hopefully prohibitively expensive for them.
Yup was thinking the same: make honest people pay $0.00001 when they visit the site (in electricity/compute), have the challenge made so that only the website wins a tiny something. Bleed the bots dry.
> This makes Anubis challenges use a memory-hard proof of work function (argon2id) instead of just a CPU hard one. It also means that the "hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead.
I want to do this eventually, but it's hard to split things into the micro-tasks that would be required to make this work on Anubis. One of the ideas I'm throwing around is a world where Anubis helps fuzz old games to find timesaves in tool-assisted speedruns. It's harder than you think.
The tools that exist for fuzzing speedruns (TASers call it "botting") are fairly primitive, excluding a few game-specific ones. Because of that, there really aren't that many TASes where a distributed randomised search could make improvements which don't get immediately overshadowed within a day of human attention. Improving botting tools and increasing their adoption would be more effective IMO. (Incidentally, TASVideos.org recently banned one of the people pioneering bot development.)
Meanwhile BOINC is well-established as the platform for distributed computation. If you can figure out how to squeeze its work units into Wasm challenges, I'm sure a lot of researchers would thank you.
Yes, I also strongly suspect that this is only going to move more parts of scrapers onto consumer devices. The egress proxies are already there, why not use a little bit of the compute as well?
The goal isn't to eliminate scrapers, its to prevent a distributed scraping network from requesting 10000 pages a second each from 10000 different websites.
> In my experience the kinds of people who run this exact combination of circumstances also tend to be the kind of people that have a wide variance in the level of kindness they display to the authors of open source programs that happen to be in their way.
Love this. There’s been past discussion on HN re how OSS maintainers are treated, and this is such a wry sentence. I really appreciate the tone / attitude to the problem.
Can you translate the American for me? “Wide variance”, does that they’re being assholes?
Well a little translation would be that some are very kind, some are total assholes, and everything in between. But your reading is probably more accurate.
I think the literal interpretation is correct here. It's a way of saying "you, dear reader who runs this combination of things, might be very kind, but some of your compatriots will be unusually unkind and will turn up in my issue tracker and on my socials to complain".
Thanks!
> something I was doing with my "strict MVP" build of Anubis' WASM wasn't in fact sticking to just the MVP features of WebAssembly
Fun fact, this _used_ to be the case - wasm32-unknown-unknown had extra non-mvp features added later, which in my eyes has been a breaking change on stable. You're not the first person to have been bitten by this, there's [1] and a similar story in Ruffle [2].
Weirdly, the response from the rustc side (IIRC we also talked with one of the devs of wasm component on discord a bit) has contained a lot of (paraphrasing) "in retrospect we think wasm32-unknown-unknown target was designed badly*, so we'd prefer if you either dealt with it or switched to another wasm target <with its different set of downsides> than us fixing it". (*I acknowledge that the way std was implemented on that target was unusual and "not clean", but it works _really well_ for a lot of people.)
> I looked into the process involved for rebuilding the standard library twice: once with only MVP wasm features enabled and once with an "all yes config" like usual. Based on some research I did this seemed like a massive pain.
That's what Ruffle does [3], it's a bit unfortunate but wouldn't call it a massive pain.
[1] https://internals.rust-lang.org/t/can-we-have-either-a-new-t...
[2] https://github.com/ruffle-rs/ruffle/pull/18397
[3] See lines 55 and 117-127 in https://github.com/ruffle-rs/ruffle/blob/6890c618daab3153343...
I have every copy of Firefox here configured with webassembly disabled...because I don't tend to do what Webassembly was designed for, i.e. online games, video/audio editing, emulation, etc. [1] and because I dislike things running in the background without my knowledge.
So this is going to be interesting.
My plea (in a small voice) to the OP: just as many sites still do for JavaScript, please add a "This captcha required Webassembly to continue" message to your code when Webassembly is disabled. And, incidentally, for smaller platforms/browsers that don't offer a Webassembly engine at all.
[1] https://webassembly.org/docs/use-cases/
The second half of this post chronicles the author's Herculean efforts to make everything work well in the cases you describe.
> One of the big things that blocked this shipping for so long was not having an escape hatch of some kind to allow clients that disable WebAssembly by policy to get through the gate. [...] This combination of factors means that there would need to be another implementation of the proof of work code in JavaScript that would actually execute the number crunching.
Thanks. You're right; it was a very thorough article and I trailed off about halfway through. I still say there are many improper uses of Webassembly out there, but I credit the developer for a highly thoughtful approach to the project. I really wonder if Cloudflare et al. does as much due diligence?
> because I dislike things running in the background without my knowledge
Point of order: WASM doesn't let things run in the background. To do that you need WebWorkers (a javascript feature, not WASM-specific).
I don't mean to be rude, but please search "wasm2js" in the article.
I really don't see why you'd disable WebAssembly and allow JavaScript instead. Expecting developers to make a fallback for a baseline-supported feature for any reason but backwards compatibility with runtimes that do not so support WASM is extremely silly.
"I disable things and expect others to do things for me, in their own time, without being paid, to bring solutions to my paranoia"
You do see how selfish it is, right? While the author did introduce such a solution, it's such a massive waste of time.
Hats off to Xe for spending so much time on backwards compatibility, especially the tidbit about targeting Chrome 66.
I have a Mac from 2014 running Yosemite that I occasionally use to test for backwards compatibility in my own frontend code (for fun!). But IMO the best way to ensure compatibility is to use period-correct toolchains or toolchains where the pace of change is slower, like ClojureScript.
In the non-webdev realm, it's gobsmacking that you can put "FROM debian/eol:buzz" in a Dockerfile, and it will do what it says...
Is there a place where I can try out if my browser is compatible? Easier to find out now than when I'm trying to get work done and a million websites now have it deployed
On https://wasm-feature-detect.surma.technology it shows that I don't have 3 of all these features but I'm not sure if Anubis needs any of them to not kick me back to the pure JS solution
Which would apparently be bad because
> The WebAssembly that's shipped with this flow is ridiculously performant. This may mean you need to adjust the difficulty [upward, to avoid that bots solve it trivially when they support wasm, I assume this means]
combined with
> The wasm2js flow doesn't currently have a way to update the progress bar [so you have no clue about remaining time]
(wouldn't be the first time that I gave up on a page because it was stuck on 0 hashes per second)
You can use Rust's `wasm32v1-none` target to get baseline WASM with no extra target features. (This restricts you to `#[no_std]`, however.) https://doc.rust-lang.org/rustc/platform-support/wasm32v1-no...
It would be cool if the POW could be done ahead of time. That way I don't get stuck waiting while I'm working. Some type of credit/tokens that my browser could then spend.
I'm working on this with a private prototype. I'm probably going to lean towards using a Service Worker (https://developer.mozilla.org/en-US/docs/Web/API/Service_Wor...) to renew challenges at a reduced difficulty. Stay tuned!
That just creates another problem: if you're taking any measures to reduce tracking (ie. clearing cookies on shutdown or using temporary containers), this won't work. If anubis was being deployed on a site that a user visits often (eg. HN), the user might be convinced to whitelist it, but most anubis deployments are on random blogs or fediverse instances that I might not visit again in months. I'm certainly am not going to whitelist those sites, nor am I going to enable cookies wholesale just to avoid solving challenges.
Look, if you're going out of your way to break expected behaviour on websites you shouldn't be surprised when people code to the most common denominator and then you have weird subtle breakage as a result.
Clearing cookies when all you want to do is read static content is not "breaking expected behaviour on websites".
Gotta agree. Turning tracking cookies off? Sure! Turning session cookies off and then complaining that the server doesn’t remember you? Uhhh…
And while I’m sympathetic to the idea of not wanting to run JS, to a first approximation modern browsers are JS engines that have graphical displays. How things should be vs how they are is a classic is/ought problem. The world took a vote on what a browser’s meant for and we lost. Fighting it today is rough; tomorrow, futile.
>Gotta agree. Turning tracking cookies off? Sure! Turning session cookies off and then complaining that the server doesn’t remember you? Uhhh…
No, because there are technologies that don't have this issue, eg. privacy pass.
There's no mechanistic difference between the two types of cookies, let's not pretend this is an actual config choice that people are making.
Whenever I hit Anubis, I simply go "keep your secrets then" and take my leave. The vast majority of the sites posted to HN (that catch my fancy) work fine or fine enough, and are better for it.
If I really, really desperately want to check something out, I can always just turn stuff back on. Turns out, I rarely do.
Shoutout to the particularly dogshit few that don't just require cookies and JS, but even third party JS. For reading a blogpost or a message thread!
Or something that shadier websites could calculate ahead of time in your browser, then spend on your behalf.
I feel like Anubis is ironically speedrunning a lot of discoveries the crypto folks have already made several years ago...
What if the work output was bound to a particular source IP? That'd stop most "theft" at the expense of discards when your device switches networks.
Anubis does this. Happy Eyeballs (https://en.wikipedia.org/wiki/Happy_Eyeballs), IPv6 privacy extensions (https://oneuptime.com/blog/post/2026-01-15-ipv6-privacy-exte...), and devices migrating between home and mobile IP addresses as the person moves have all caused issues with this. Fixing this is hard, but I have been trying.
Privacy Pass does something similar to what you describe: https://developers.cloudflare.com/waf/tools/privacy-pass/
I wonder if you could convince the Anubis author to implement x402 payments to bypass the PoW
This could be done at the web server level, not requiring changes to Anubis
Then we could implement an exchange, so that if you generate too much Anubis POW, you could exchange it with others.
We could link it to a site, you generate for HN, I for Reddit, but it so happens that you visited Reddit more and I HN, so we depleted our Anubis POW, so we could exchange some Reddit Anubis with some HN Anubis.
And then, we can inextricably link it to a new cryptocurrency -- let's call it BotCoin -- and make Sam Altman pay for it!
Basically reinventing crypto one step at the time.
> This means that adding one (1) to the difficulty of a challenge makes it 1024 (one thousand twenty-four) times as hard to solve in the worst case.
I don't understand the units here, wouldn't a bit versus nibble difference make the multiplier 16, instead of 1024?
Oops, there's your proof I didn't write this with AI! Fixing, sorry.
The interesting part is that bots mitigation is actually an economic problem. You don't need to make scraping impossible, you only need to just make it expensive enough so that abusive traffic stops being worth it.
Scrapers scrape all sorts of nonsense, notably every possible git diff from git hosts. They clearly aren't concerned with rationality.
Scrapers scrape every link they can find.
The git host example is probably the one which Anubis is the dumbest defense for: the main reason to have those links is for easy machine interaction. So that's dead when you implement it.
So why even host the links? It's a git repo: send them a local got client and let them clone the repo in browser or something.
The main reason to have those links is for people to click on them. No sensibly programmed machine is scraping diffs from a git web front end.
Not to mention it is the least efficient representation of git commits. It seems a lot of people are unaware that git generates the diff representation you see with "git show" on the fly -- internally, commits are Merkle trees representing a snapshot of the repo file contents.
As you said, if you actually want to interact with a remote git repo you'd do so with packfiles not diffs.
> why even host the links? It's a git repo: send them a local got client and let them clone the repo in browser or something.
I document my code, in issue and elsewhere on the Web, by referring to specific commits so having a URL for each is useful to me at least.
> It also means that the "hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead.
I assume that the previous challenges will still be available for circumstances where WASM is not available (unless we are just working on the basis that it is available if JS is?) or perhaps as a fallback. TFA explicitly mentions “smart” TVs, and I'm sure there are plenty of those out there that run browsers old enough to not know what WASM is. TBH these would not be a priority for me, but they obviously are for this project given they (well, TVs generally, not specifically elderly ones) get specific mention in this write-up.
If so, what is to stop scrapers from just forcing the fallback POW option instead of trying to run the new memory-hard algorithm at all?
the answer to this is in the post: where wasm is not available, he transpiles the wasm binary back to Javascript
Ah I must have skipped by that bit. And presumably the memory-hard algorithm is CPU-easy enough that this won't be a problem. That leaves me wondering "why not just use a JS memory-hard routine in the first place, but maybe I've shouted over that too. I'll give TFA a reread later.
I think it is still compute-heavy if not compute-hard, and stuff like WASM SIMD makes it less compute heavy (not really available in JS). And also, for memory bound problems, you probably don't want the JS GC needing to clean up afterwards when WASM gives you that control.
I assume that SIMD acceleration is usable with WebCrypto, but I have no evidence either way. It's also not part of WebAssembly so there's that to consider too lol
tangential fyi: the author is not a he
Tangential but all the pages I see protected by Anubis are just static HTML. But bot traffic is only harmful on dynamic pages right?
Others have answered your question. A lot of the sites listed here https://anubis.techaro.lol/docs/user/known-instances/ have dynamic content; Git web interfaces in particular (Codeberg, the Linux kernel, FFMPEG, and more are on the list) are vulnerable to poorly or maliciously configured scrapers.
I'm not a fan of Anubis for various reasons but the idea that bot traffic in only harmful with dynamic pages must die. CPU (yes, even to serve static pages) is not free, bandwidth is definitely not free. There's an idea that serving a static page to a bot has a marginal cost tending to zero, but it's never really zero and serving them by millions definitely has a cost.
Also, while some pages may look like static HTML pages, they may be generated on the fly by an expensive and/or slow backend, which adds to the cost. I happen to maintain servers for academics and some content management systems are slow and have an expensive CPU cost. While it's OK for the low number of humans interested in the subjects they deal with, it's definitely not fit for massive bot scrapping. And before you ask, no, it's not always practical to have cache upfront or to pre-generate all pages.
There are numerous services that will let you host static pages for free or nearly free. There are also numerous services that sit in front of your website that can block bots and reduce load on your origin server, many of which are also free, or very low cost relative to the service they provide.
The situation you are in is far less dire sounding when you consider that you have these options available to you.
Except that I don't have these options per employer policies.
Shitty scrapers hitting images over and over again is also a problem. Not just /pic/1.webp - /pic/2.webp, but in practice I've seen three AI scrapers that would request things like /pic/1.webp?version=webp&foo=bar and other random query parameters that none of my domains even use. It's just stuck brute-forcing files or something, sucking up loads of bandwidth downloading the same image for no reason.
It also depends on if you want your content to be scraped. Any bot ignoring robots.txt should be considered potentially malicious and worth considering such mechanisms for.
The anubis docs website is used as an example of how anubis works.
Also, if you are not using a big CDN (e.g. Cloudflare), you still need to pay for the traffic (even if you don't pay as much for the CPU). This is doubly true for smaller CDNs like Bunny.
Not only is it an example, it specifically is one of the first things I use to test experimental features so I get data back instantly.
some people don't want to be scrapped at all, and most bots ignore `robots.txt`
no, absolutely not, any malicious or dishonest bot traffic is harmful, generating unnecessary transfer and literally stealing data
How much data does it download off the server to run the challenge? I've made some tiny webassembly binaries before, but I've never seen Rust generate something tiny.
It's in the range of 32-40Ki before compression. After compression it gets to be completely negligible.
Anubis is fine if you want to block bots that are bothering you, but I never understood why they don't have robots.txt at the root level too. I happily obey that for my private crawler. I suppose you don't want to duplicate implementations. Fair enough. I guess I should add an Anubis detector so I can just blacklist because that person clearly doesn't want bots.
In this day and age, only good bots, the ones you actually want, obey robots.txt. You obtain no benefit by having a restrictive one.
That is true, but if you have Anubis you presumably do intend to `Disallow` all bots.
One of the standard library rules enables robots.txt passthrough by default. I can't control people not choosing to use it.
That makes sense, Xe, but you don't enable it at:
- https://techaro.lol/robots.txt
- https://anubis.techaro.lol/robots.txt
and the latter (at least) is Anubis protected. It could be just an omission. But if it's intentional, I am just curious why so that as a bot author I can be well-behaved.
thanks for trying to make the web suck a little less these days xena!
Great article, WebAssembly is such a good technology when you have use case for it.
I've just finished creating a WASM 1.0 implementation (simple JIT on x86 and x86_64) for plugin usage and I'm looking on how to support more languages beyond C (for which I have "micro-libc" to create really small binaries). In particular C++ and Rust so the article contains useful info for that. C++ that would be more problematic I think, would have to resort to full Emscripten or something... I've tried to compile standalone C++ stdlib but wasn't much sucessful yet.
BTW, I see in the screenshots you also test Firefox, I think it would make sense to test the latest two ESR versions (these are often used by Linux distributions).
It's still very unclear to me how Anubis is supposed to work in the long term or even the medium term. The assumption is that scrappers, a.k.a. AI companies, a.k.a. those who hog all the RAM, are short of memory...?
The assumption is that requiring 1 second of compute before allowing a page load will have more impact on the people requesting 1000 pages/sec than it will on consumers requesting 1 page every minute.
Personally I don’t understand where this confusion comes from - it’s a simple economic tactic. Yes the large scale scrapers won’t run out of compute but putting anubis in front of a website demonstrably reduces the amount of crawling and that is the entire goal.
> requiring 1 second of compute before allowing a page load
After 1 solve, you get a cookie and can load tons of pages with it. Scrapers and user agents alike will need to spend the compute only once to get a valid session for themselves
Seems to be working so far though. Fwiw, lots of silly things stop bots until someone puts their mind (or tokens) to bypassing it, and then the blocking mechanism has to adapt. We'll see how it plays out, for now it's a lot better than begging big brother (cloudflare, google, or amazon captcha) for access, so I'm quite appreciative of what Techaro is doing
The issue is that scrapers are rotating their IP and essentially performing a DDoS attack. Anubis is part of a defense-in-depth solution. If scrapers reuse a cookie, traditional anti-abuse mechanisms will work.
>If scrapers reuse a cookie, traditional anti-abuse mechanisms will work.
Scrapers can't change their strategy? It's not hard to think of mitigations: keep on using a session until you're banned, and failing that rotate your IP. Nowadays you don't even need to think. You can probably tell a LLM "write me a scraper that has adaptive anti-anti-bot mechanism".
> Scrapers can't change their strategy?
Yes, but you normally don't try to fight all strategies up front, you add stuff on a need-to basis. Currently, lots of people are in the stage of "need to get this thing using bunch of IPs to stop hammering my servers", apparently Anubis helps them with that. If that stops working, other tactics are tried, and so it goes on for the life of your service. Hasn't stopped yet, doesn't seem to be stopped soon either.
That argument also works against anubis. Why bother forcing clients to solve a PoW challenge, when you're fighting a bunch of unsophisticated attackers that will be deterred by a javascript challenge that requires them to solve `result = x + y`?
This is why there's the meta refresh challenge as an option: https://anubis.techaro.lol/docs/admin/configuration/challeng...
Turns out that in practice there's some bots that implement support for this but not JavaScript.
Point is, you get problems because of something like this, you try either approach. And when that stops working (because eventually it will), you add something else. Some work better against some actors, but also feels like it never ends, so slowly you add more things. I don't think Anubis is uniquely able to stop that particular type of annoyance, it's just one of the tools you can see if it helps. If you want to try something else before, makes a lot of sense to do so.
You're not fighting unsophisticated attackers. 50% of them are, maybe. The rest? They're actively hostile, detecting blocking attempts, moving on to botnets, DDoSing you and trying everything they can to bypass you.
People have been sent to jail for participating in LOIC attacks which were less destructive that what these bots do on the daily.
Caution: Techaro is very hostile to this line of thinking. He believes the PoW is essential. And that's fine - different projects use different approaches, like go-away and dogwaf.
Plenty of these abusive scrapers are utilizing retail residential proxies, which will be applying forced rotations to avoid "burning" their compromised and or otherwise surreptitiously utilized IP address.
https://people.kernel.org/monsieuricon/creepy-crawlies reports anubis effectiveness falling off
I would disagree with the Anubis effectiveness falling off, but the llm bros are throwing life itself into training to defeat any “perceived” challenge.
If you change your IP address, it invalidates the cookie. If you don't change your IP address, you can be blocked by IP address.
It’s not seconds though, it’s in the order of tens of milliseconds of work, it’s faster to complete the challenge than most http requests will take.
last difficulty 5 nonce 645376 in 9 ms (101.6MH/s, avx512-x16)
This takes many minutes to solve on a cellphone, 9 milliseconds on a cheap VPS. It’s not stopping anyone.
Yep, on my aging phone it takes quite w noticable amount of time to go through. But on my more recent laptop wnd workstations, it's a quick flash of a manga lolita.
>It’s not stopping anyone.
You could just spend 30 seconds googling instead of confidently asserting nonsense. There are plenty of people demonstrating that Anubis reduces bot traffic.
That’s because its weird and mostly irrelevant, it’s nothing to do with the proof of work being any sort of economic impediment. It takes Claude a few minutes to write a near instant solver, that’s the maximum proof of work cost this adds.
This strain of nihilism about anti-bot efforts baffles me. Cloudflare built a gigantic business on stopping DDoS attacks. There will never be good faith self-restraint by AI bros, there is too much money at stake and no regulation, doing nothing means your only alternative is disconnect from the Internet. Anubis introduces enough friction that some percentage of the lunacy is throttled and possibly give your server a fighting chance.
>There are plenty of people demonstrating that Anubis reduces bot traffic.
"Anubis reduces bot traffic" and "[the proof of work] is not stopping anyone" isn't contradictory, especially if anubis isn't being benchmarked against non-POW alternatives like IP/ASN bans, or tls fingerprinting.
"it's not stopping anyone" and "it is stopping some people" are mutually exclusive statements.
You're selectively quoting the comment. It's clearly talking about the PoW aspect not being a serious impediment.
Who cares if its PoW or something else? Does it matter in any meaningful way?
>The assumption is that requiring 1 second of compute before allowing a page load will have more impact on the people requesting 1000 pages/sec than it will on consumers requesting 1 page every minute.
Which doesn't hold. A 1s delay, despite being inperceptible is still a cost to human users. After all, 3600 x 1s delay adds up to an hour. Worse yet, the typical human visitor will always have to solve a challenge, because chances are it's the first time they're visiting your blog or whatever. Someone scraping reuse a valid session for at least a few more pages.
There's also the problem of time. It might be tempting to think the cost of a challenge is 1M cpu cycles or whatever, but the cost is better modeled as two parts, a time cost (for the human) and a cpu cost (for the computer). A scraper might have to pay the cpu cost, but not the time cost, because they can have their scraper solve challenges in their sleep. Unless the user is utterly destitute, their time is worth far more than whatever the compute cost is.
The assumption is that there will always be some population distribution in terms of effort expended as well as in terms of problems caused. At present it's the lowest effort actors that are causing the highest number of problems. So an 80% (or even just a 20%) solution largely solves the practical problem faced by site operators.
If against all odds it turns out that skiddies (or agents or AI billionaires or what have you) are willing to burn inordinate amounts of RAM and CPU just to scrape autogenerated cgit pages that are of approximately zero value to them to begin with then I guess we'll just have to cross that bridge when we come to it.
Okay but it's like no RAM at all. Thats the problem: you've got stupid bots causing a bunch of issues - these can be stopped.
And then regular users: who now can't really browse your site.
And then smart bots - anyone with even a moderate amount of funding in the AI space, who aren't even slightly slowed down.
The solution basically makes the favored customer well funded AI scrapers, not users.
Regular users store the cookie and do the challenge once. Bots do it over and over again throwing out the cookies so they're not rate-limited.
Extreme configurations with high difficulty do make for a frustrating experience on very old devices, but in practice it rarely gets in the way.
Only the tail end of what one could consider normal users (those running custom browsers or having JS disabled or nonfunctional) run into issues here, but they'll have these issues with any other bot-blocking mechanism that still offers some protection these days.
Apple's "ask your vendor for tokens to exchange for access" web standard gets around this, assuming those bespoke and Javascript-less browsers implement PATs, but that standard seems to be completely dead in the water outside of Safari.
> And then regular users: who now can't really browse your site.
That's entirely up to how the site operator configures the PoW challenge.
> anyone with even a moderate amount of funding in the AI space, who aren't even slightly slowed down
Do you have any evidence that it is the large well funded AI companies DoSing sites by (for example) scraping absurd numbers of permutations of cgit url parameters?
Even if it is (which I doubt), if all this accomplishes is getting them to spend 10 minutes optimizing their bots to not do that it will be a major win.
See also:
https://github.com/eternal-flame-AD/pow-buster - Browser extension that already used WASM to accelerate the anubis solver, among others.
https://github.com/kasper93/anubis_webgpu - Browser extension that uses webgpu, for another order-of-magnitude speedup (depending on your GPU of course).
afaik argon2 should make the GPU less helpful
Argon2 is already GPU-accelerated.
I think the idea is to just raise marginal costs for scraping high enough to make it expensive and non-profitable.
I think the Rust feature you’re looking for regarding recompiling the standard library is called “build-std”, that should be enough for you to search for it. (For similar reasons you also need that flag if you are trying to use Rust to build multithreaded wasm binaries, so it might come up for you!)
It's wild to me that PoW systems are how we sort the bot problem... the bummer is that all this "work" is just wasted cycles, at least in crypto there is a token you can sell.
> at least in crypto there is a token you can sell
That would defeat the purpose. The goal is to make scraping costly, not profitable.
Scraping would be costly in this world: scrapers would have to spend tokens in order to get the webpage.
But, in this world, the website owner would receive tokens that they can then use to do whatever they want, including paying for servers and bandwidth. This is the sense in which the cycles aren’t wasted: the website owner now has cash to spend.
Effectively, both scrapers and ordinary users would be paying for the privilege of getting website bytes.
This also solves the problem of having to wait for your phone to solve the challenge while you’re browsing: you can buy or mine some tokens ahead of time and pay them as soon as challenged. So can the scrapers, but because they’re accessing enormous numbers of pages it’s hopefully prohibitively expensive for them.
I guess it could in principle be profitable for the website, not the client?
unfortunately at that point it would be indistinguishable from running cryptojacking on your website
No. The client is doing the work.
The client is doing the work in both cases
Yup was thinking the same: make honest people pay $0.00001 when they visit the site (in electricity/compute), have the challenge made so that only the website wins a tiny something. Bleed the bots dry.
> This makes Anubis challenges use a memory-hard proof of work function (argon2id) instead of just a CPU hard one. It also means that the "hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead.
Nice.
Sure, then just replace it with something that is useful to society but not immediately profitable to a scraper, like science research
I want to do this eventually, but it's hard to split things into the micro-tasks that would be required to make this work on Anubis. One of the ideas I'm throwing around is a world where Anubis helps fuzz old games to find timesaves in tool-assisted speedruns. It's harder than you think.
The tools that exist for fuzzing speedruns (TASers call it "botting") are fairly primitive, excluding a few game-specific ones. Because of that, there really aren't that many TASes where a distributed randomised search could make improvements which don't get immediately overshadowed within a day of human attention. Improving botting tools and increasing their adoption would be more effective IMO. (Incidentally, TASVideos.org recently banned one of the people pioneering bot development.)
Meanwhile BOINC is well-established as the platform for distributed computation. If you can figure out how to squeeze its work units into Wasm challenges, I'm sure a lot of researchers would thank you.
This would be rad.
Well, there was CoinHive which did this exact thing 6-7 years ago, but that system got abused a bit much
PoW was originally anti-spam technology.
Comparatively, this is what it's like getting all of this working across browser versions, platforms, and so many other things:
…<img not found/>…
I don’t know what this picture was supposed to be, but a 404 demonstrates your point perfectly well.
Refresh, the Anubis docs site regenerates its cookie secret on every deploy as part of the "holy shit did I break everything?" verification step.
the enemy of my phone's temperature is my friend.
> hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead.
Lmao yeah no. I don't think a little argon2 is going to change shit all.
I mean the thesis of Anubis itself is "scrappers are compute limited (in ways that consumer devices are not)" which has its own massive flaws.
Yes, I also strongly suspect that this is only going to move more parts of scrapers onto consumer devices. The egress proxies are already there, why not use a little bit of the compute as well?
When you force a low end device to burn CPU or fill ram constantly, the owner throws it out and buys a new one.
Haha, no. The average tech illiterate person tolerates a LOT when it comes to bad performance on low end devices.
The goal isn't to eliminate scrapers, its to prevent a distributed scraping network from requesting 10000 pages a second each from 10000 different websites.
Somehow WebAssembly does not want to become as popular as HTML, CSS or JavaScript. Very disappointing to see.