[jchw]

With anti-cheat the obvious (lazy, stupid) future is remote attestation. It's another way to kick the can down the road of actually going to "real" approaches to anti-cheat like less client trust, behavioral analysis, and statistics that you would have to do if you had no choice, like for online Chess games or something like that. Of course even for fast-paced games like FPS games, you can now cheat using a capture card, ML models and a fake HID keyboard/mouse device so I'm sure the arms race will evolve to include forced HDCP and signed, encrypted HID devices and other dumb bullshit before there is finally some realization that there is no longer any possible, reasonable way to shortcut anti-cheat anymore. The shortcuts are just too much cheaper and easier. (I'm sure we'll keep remote attestation anyways afterwards, because it's impossible to have nice things.)

[reginald78]

I think a lot of these problems are exacerbated by the developers themselves. Dedicated servers were removed (so that games could be sunset-ed to avoid competing with sequels) erasing community and human moderation leaving players entirely dependent upon algorithms and spotty reporting. F2P games are pushed to get people on the micro-transaction treadmill which makes the cost of being cause nothing but also can create a financial incentive to cheat for items. Now everyone has to jump through a bunch of extra hoops because the control they demanded isn't enough to solve the problem they made worse with it. The answer is always to add more hoops.

[SirMaster]

Dedicated servers were mainly removed to balance matches with matchmaking IMO.

How do you do proper matchmaking with user-hosted dedicated servers?

Like what would an end-user dedicated server system look like for call of duty warzone, or Fortnite, battle royale games. A round lasts about 20 minutes and you want 150 players filled in right at the start and you don't want to wait more than a couple minutes to fill the server for each round.

The other popular types of FPS games are mostly like 5v5 modes, and you want really balanced teams based on an MMR system. I don't really see how you do that with dedicated servers.

[steelbrain]

> forced HDCP

Fly-by, but HDCP is already cracked. There's no shortage of HDCP strippers from AliExpress; although they use clever marketing terms to avoid spelling out the fact (presumably to avoid legal troubles)

[jchw]

Oh, I know. It's perfect for anti-cheat though: it doesn't fix the problem, it slightly increases the cost/entry barrier, and it's a pain in the ass for consumers. It's just like all of the other current anti-cheat technology!

[perching_aix]

Why do you think cheating is a fixable problem? Or am I misinterpreting your words?

[jchw]

Oh no, it's definitely not "fixable", it's hard enough to define cheating precisely less fix it.

Like there's obvious cheating. Playing online chess games using an engine to tell you every move is cheating, and thankfully if you really don't know what you are doing, you are likely to get caught.

Then there's less obvious cheating. It's pretty easy to cheat in speedrunning by passing off a carefully constructed "TAS" run as a regular speedrun; though just like with chess it's going to be obvious if you don't know what you are doing, but it's pretty hard to inspect in detail so you might get away with it as long as nobody pays too much attention.

Then there's things that may or may not be cheating in the first place. A good example is key binds in Source Engine games, like null-cancelling movement configurations. The game allows it and keybinds are not traditionally treated as cheating, so is it? (I think they may have changed this in newer games/updates but I'm behind on the times.) Or custom controllers for Super Smash Bros Melee. Merely emulating degraded analog sticks in a way that is advantageous is generally not considered to be cheating, but e.g. warping the stick movement depending on an internal state machine to make certain things easier to do is usually considered too far. I can think of many examples. "Calcing" and charting in PangYa is a nice obscure one.

But I do say this: If you took one aspect of cheating, like aimbots, you could attack it surgically. You could detect the current state-of-the-art, do your best to make it hard for aimbot developers to figure out how you're detecting them, ban players and make it hard for them to get new accounts. The problem is that this is not easy or cheap. People just want to throw some money at the problem and have it go away, they want something like Easy Anti-Cheat.

I think though that eventually, it might all just be futile. It depends. If machine learning continues to progress, eventually it will be pretty hard to distinguish human and machine inputs no matter how hard you try to. It's already not very easy looking at "state of the art" technology from 2022. I linked this elsewhere but it has a nice chart of people mostly failing to detect their ML-based aimbot:

https://arxiv.org/abs/2205.07060

I guess I'm just disappointed in these lazy approaches because they wind up being so vapidly anti-consumer. I know that they accept the temporary nature of these fixes since they don't really intend on these games running forever (which is also a sort of problem, though not related since obviously they really can't worry about and invest in anti-cheat forever) so they just need something that works for now. But it seems like no bridge is too far, and now we're nearly all the way to "you need to install this rootkit and enable secure boot and use Windows 11 and have TPM2" or whatever else.

edit: I really wish Hacker News pushed downvoters to leave a reply... If whoever did that comes across this post again I'm actually genuinely interested in hearing about whatever you disagreed with strongly enough in this post. Unfortunately I said so much that I can only guess what and that's a bummer because I bet it would be interesting.

[perching_aix]

For the record, I did not downvote (and I do think there are valid reasons to downvote without saying anything, even if I appreciate that it is frustrating sometimes from the other side).

I think there's merit to recognizing how loose the concept of cheating can be, it's much akin to the concept of vulnerabilities.

That said, a very very very large part of cheating in videogames revolves around these points:

- ensuring the inputs are not altered: not shaped, not injected into, not filtered, not replaced

- ensuring the inputs are issued by a human

- ensuring that the inputs issued by said human are made in response to the output of the game being processed by said human

- ensuring the that these happen at roughly the same time, and in real time

- ensuring that the identity of the human in question is unique among the playerbase (to prevent veterans sockpuppeting as newbies, as well as multiboxing, and many other forms of abuse)

- [several more I probably forget]

And I'd go as far as to say that even individually these are impossible to actually solve, let alone combined. The equivalent problems for machine-to-machine communication are already hard enough, but making one of the participants a human really twists the knife. It's also essentially the analog hole problem.

[jchw]

Ultimately, I'd say it's all about trying to ensure a level playing field and trying to make sure nobody has an unfair advantage, and "unfair advantage" is always somewhat subjective. In real world sports, it often manifests in hard-to-interpret rules like F1's new-ish rules around moving under breaking. In "eSports" it manifests in many ways, like debate over what hardware and software is cheating.

> and I do think there are valid reasons to downvote without saying anything

I think people should be generally encouraged to reply. There's some comments where it's obvious why someone would disagree or take offense, but lately I've noticed that even comments that seem incredibly benign often get grayed out at least briefly. This is confusing and hard for observers to make sense of, less the poster themselves. With everything left implied, it's not even clear if people even agree on why they don't agree with the post.

Stack Overflow encourages people to explain themselves probably for the same basic reason. HN on the other hand asks commenters to not discuss votes. I do generally agree (in fact I wouldn't even really want people to start comments with "I'm down voting because...") but damnit, sometimes you just want to know why.

And of course, sometimes the reason why is someone mistapped on their mobile phone, or something like that, so you'll never know.

[johnfonesca]

> HDCP is already cracked

It's being 15 years already

[dontlaugh]

Less client trust means some game designs are impossible. The latency caused by the speed of light leads to unacceptably bad player experiences if you don't trust the client, at least for shots from their p.o.v.

Personally I don't mind if fast-paced adversarial multiplayer FPS games stop existing, but that's a minority opinion.

[armarr]

Anti-cheat does not need to run real-time to be effective

[Demiurge]

So you put in the effort to share some nuanced opinion on a subject on a forum, and then you say you'd prefer if the whole subject did not exist.

[dontlaugh]

I’ve worked in the games industry for long enough to understand why anti-cheat exists in the form that it does.

I also understand that there are many players that like adversarial FPS games and many companies that make money from that.

Just because I personally don’t think the trade-off is worth it doesn’t mean I expect the world to comply to my preference.

[jchw]

To clarify I don't really mean "no client trust", I just mean less. There's a lot of ways to skin a cat when it comes to multiplayer architecture. (I realize you probably know this, but it's always worth repeating. I'm routinely surprised by what people manage to come up with.)

For something like a 2D fighting game it seems like they've landed on determinism + rollback most of the time. That basically only trusts inputs from any given client and nothing else is trusted. Totally possible for the server to "verify" a client's claims by running the replay if needed, though obviously that can be rather expensive. Either way, that eliminates a huge avenue of cheating; as long as the developers did it right, you're basically just left with cheating by botting and using macros.

For something like an FPS game life is hard. The client has to react "as if" there is no latency even though there is. If an arena "twitch" shooter (does... anybody still make these?) ever had a situation where the cross-hair lined up but it didn't count, that would basically break the game. Still, there are many more things that are typically client-controlled, and there should be ways to not give the client as much control without adding too much latency. Any situation where the client and the server or peers can both track something and the server or peers can verify the correctness of the state of the client after the fact, it's probably possible to make some improvements. Basically, make an effort to ensure as little of the information received from any given client is inherently "trusted".

I won't lie, the approach of using heuristics in an otherwise client-trusted environment is definitely "fraught with peril" to some degree; bad heuristics can be pretty damaging. Blizzard ran into a problem where they were accidentally banning Wine users at one point, and it took them a really long time to actually admit the bans were incorrect, and as far as I know they never bothered reversing any of them. Don't get me wrong, they can say "Wine is not supported" and call it a day, but even ignoring that you'll always have some false positives, due to bugs or even weirder issues like clients with bad DRAM.

FPS games are definitely in for a world of hurt though. It's one thing to combat rudimentary aim bots that read internal game data and move the mouse with exact precision, totally different ball game when you have vision models controlling HID mice and keyboards. A Raspberry Pi Pico and a second computer with a capture card isn't a non-zero expense... but, a lot of people already have these things lying around. No amount of not trusting the client can do anything for you with this. You would definitely have to get deeper into behavioral analysis.

[dontlaugh]

There are a few things you can do, for sure. The Valorant team has a good blog post on it (https://technology.riotgames.com/news/peeking-valorants-netc...). The Overwatch team did an excellent GDC talk on it a few years ago (https://www.gdcvault.com/play/1024001/-Overwatch-Gameplay-Ar...).

Behavioural analysis is indeed useful anyway, but I don't see anyone that has it giving up memory integrity anti-cheat.

Ultimately, I do think the game design makes it impossible to prevent cheating. It's unfortunate that it's precisely this design that many enjoy.

[hypercube33]

Game clients need to have zero trust. there has never been a truly working anticheat that doesn't punish the user in some way.

Be it you can't game on Linux (steam deck) or if you have some specific software installed you're flagged as a cheater (autohotkey) or other dumb things.

That said even in some AAA games it feels like the game cheats anyway either intentionally or unintentionally (read about how modern warfare basically decides the outcome of an encounter before it happens to keep your k/d ratio close to 1)

[lukan]

"Game clients need to have zero trust"

Good luck implementing anything fast this way.

Have you thought about a bit, what this means?

Only ever give the client what it can see?

With common real world ping times around 60 ms?

It would result in very slow shooters. Where you have to wait a bit to see what is around the corner after going there. Not that practical.

In my opinion most anti cheat is BS anyway and even if some gets banned, the main problem is, game companies seem happy with the solution of just selling new accounts to cheaters.

[Ukv]

> It would result in very slow shooters. Where you have to wait a bit to see what is around the corner after going there. Not that practical.

I think ideally you'd send players only what they could possibly see before next the next packet comes in. If their ping is 60ms and the game's tickrate is 128Hz, each tick they'd be receiving everything visible from within 68ms walking/running/jumping/etc. of their position 60ms ago.

Could be combined with server-side behavior analysis to detect players seemingly using information their client was sent but they shouldn't have seen. Packets from the client of what they actually did in that 60ms should be half received by the server and half in flight to the server by that point, so the client can't retroactively lie.

Wouldn't eliminate the advantage of this kind of cheating entirely, but might make it small enough that for many current cheaters it no longer justifies the risk.

[lukan]

"I think ideally you'd send players only what they could possibly see before next the next packet comes in."

That still makes wallhacks possible if standing close to a corner.

And otherwise what you describe makes netcode way more complicated, than it already is. Resulting in a worse game experience.

I am annoyed by any lag or stuttering and find it immersion breaking.

[Ukv]

> That still makes wallhacks possible if standing close to a corner.

That's what I'm thinking with the server-side behavior analysis to detect players seemingly using information their client was sent (in case they made a certain movement) but they shouldn't have actually seen (because they didn't make that movement). Though even without that, restricting wallhacks to corners that you could already almost immediately see around would already be a significant improvement.

> And otherwise what you describe makes netcode way more complicated, than it already is. Resulting in a worse game experience.

It's what I think is the ideal and would definitely take some work to implement in practice, but if implemented well I'd argue the measures would be invisible to players in almost all scenarios, and game experience would be superior to solutions that have rampant wallhackers and/or don't work on many machines due to invasive client-side anti-cheat.

[anton-c]

Your last paragraph is spot on. From a players perspective, more than one MMO I played clearly had bot/script problems that the devs always addressed publicly but never took the easiest and obvious solutions.

Hard to not see it - esp in smaller games - as the bots and cheaters providing too much revenue to cut out completely.

[Razengan]

I guess the ultimate "anticheat" would be to have a camera pointed at the user and their screen, and have some kind of super AI watch it.

And the ultimate "cheat" would be an android hooked up to the computer and indistinguishable from a meat-based person :)

[zihotki]

Even that can be cheated by mimiking the device and feeding it a generated video stream.

[immibis]

That's why the device is locked down like a payment card reader. Comes with a key already loaded; if the device is ever opened or the battery runs out, the key is erased.

[reginald78]

Meanwhile my retirement account is with a financial services company that pushes voice match authentication, an authentication mechanism that was a joke in 1980s.

[Razengan]

How about sending a company droid to the user's premises? To monitor them in person

[anton-c]

Wait how does using a capture card let you cheat? Not disputing, very curious. Like a bot sits inbetween and makes the correct moves?

I don't play fps anymore not trying to cheat lol

[jchw]

In the future, you no longer need to spy on internal game data to make an aimbot.

https://arxiv.org/abs/2205.07060

[anton-c]

Thanks, that's wild. I chuckled when I read that it was using deep neural network "YOLO".