5 ms·

this approach is used by the game, Counter-Strike: Global Offensive. the system is called Overwatch (not to be confused with the game of the same name..). how

this approach is used by the game, Counter-Strike: Global Offensive. the system is called Overwatch (not to be confused with the game of the same name..). how it works: once a player has received enough reports for suspicious behaviour (i.e., cheating), or bad behaviour (griefing, throwing, etc.), a replay of the match is submitted to tens of other players. these players review the "demo", decide on the likelihood of the "suspect" being guilty of those offenses. if all reviewers come to a unanimous decision, then the suspect receives a ban. the system is thoroughly abused[1] by hackers/cheaters who are able to spam reports for players they don't like, and by controlling hundreds, if not thousands, of bot accounts, can reach unanimous verdicts that no normal person would decide, inflicting bans on innocent players. [1] - https://www.youtube.com/watch?v=N0OiOiqmi-c https://www.youtube.com/watch?v=N0OiOiqmi-c

esyir 5y ago
So, I think the base concept is fine, but requires oversight. It's an augmentation of anticheat, not a replacement. Hidden reliability scores should aid the classification of "judges". Misclassify often? Get downprioritized. As CS is a paid game, the typical issue of mass account creation to game this should be at least slightly mitigated. Anticheat has to be defense in depth.
- foxhill 5y ago
  well, therein lies the issue: this approach, one of scoring jurors (i think that’s a more appropriate term :), already happens. almost exactly as you describe. defense in depth, indeed: Overwatch is one of three systems, the other two being VAC, and Trust. the former being more traditional cheat detection (which, empirically, is not effective at all), and the latter being an observation of a user’s interaction on the entire steam platform, amongst other things, to determine a “trust factor”. anecdotally, myself and many others can tell you that this is also ineffective. for what it’s worth, 2020 and by the looks of it, 2021, will go down in CSGO’s history as the years that the cheaters won.
  - esyir 5y ago
    So, looking at Riot's anticheat for Valorant, they're generally considered to do a good job. Similarly for CSGO, FaceIT exists, and from what I've seen, is considered a good, low-cheater environment. The issue with CSGO was not that the methods were ineffective, but that it's a constant war. The moment the devs stop fighting, the cheaters can and will most definitely win. Valve gave up, so CSGO is cheater infested. FaceIt didn't, and thus is a better environment. For paid games, there's a fun addon there. If cheaters have to buy a new account every time a ban hits, in theory, effective anticheat can subsidize it's own cost. The catch being that they'll then start using hacked accounts to cheat on. Then 2FA comes in... Such is the rabbit hole of anticheat, there's always a next step. Somehow, cheating at games is a prime objective to far too many people.
    - foxhill 5y ago
      Valorant’s anti-cheat is considered to be good, but hacks are available. it’s also incredibly invasive - running in the kernel, meaning it must be enabled at boot time - something i don’t think is either necessary or desirable. faceit/esea isn’t much better. and i suspect it’s only better at all because of the higher barrier to entry. as for MFA - Trust takes this into account in determining its score. to stay on topic here, i think what i was originally trying to say was, yes, having humans in the mix for self-moderation is a good idea in principle. in practise, it’s easy to manipulate because it’s very hard, potentially even impossible, to determine if you actually have a human using the machine. well, except in china, i suppose. but i don’t think we want to live in that world :)
- PoignardAzur 5y ago
  > Anticheat has to be defense in depth. I'm pretty sure every layer of defense you can think of, they've tried. - Machine learning to spot obvious cases? Check. - Federated moderation? Check. - Various trust schemes to weigh accounts based on how much they payed, whether they cheated in the past, etc? Check. The only thing I don't think they've tried too much is sending false/stale game data in cases where they know the client has no way to know otherwise (eg lying about an enemy's position when the player has no way to see the enemy). It's an arms race and it's hard to stay ahead at scale. (though I'm not actually sure I trust accounts of how prevalent cheating is; I haven't had that many cheaters in my own games, and a lot of complaints I've seen come from either free users or people who admitted to cheating themselves, which is what I'd expect to see if Trust Factor worked; though it's a crap deal for free users)

esyir 5y ago

So, I think the base concept is fine, but requires oversight. It's an augmentation of anticheat, not a replacement. Hidden reliability scores should aid the classification of "judges". Misclassify often? Get downprioritized. As CS is a paid game, the typical issue of mass account creation to game this should be at least slightly mitigated. Anticheat has to be defense in depth.

foxhill 5y ago
well, therein lies the issue: this approach, one of scoring jurors (i think that’s a more appropriate term :), already happens. almost exactly as you describe. defense in depth, indeed: Overwatch is one of three systems, the other two being VAC, and Trust. the former being more traditional cheat detection (which, empirically, is not effective at all), and the latter being an observation of a user’s interaction on the entire steam platform, amongst other things, to determine a “trust factor”. anecdotally, myself and many others can tell you that this is also ineffective. for what it’s worth, 2020 and by the looks of it, 2021, will go down in CSGO’s history as the years that the cheaters won.
- esyir 5y ago
  So, looking at Riot's anticheat for Valorant, they're generally considered to do a good job. Similarly for CSGO, FaceIT exists, and from what I've seen, is considered a good, low-cheater environment. The issue with CSGO was not that the methods were ineffective, but that it's a constant war. The moment the devs stop fighting, the cheaters can and will most definitely win. Valve gave up, so CSGO is cheater infested. FaceIt didn't, and thus is a better environment. For paid games, there's a fun addon there. If cheaters have to buy a new account every time a ban hits, in theory, effective anticheat can subsidize it's own cost. The catch being that they'll then start using hacked accounts to cheat on. Then 2FA comes in... Such is the rabbit hole of anticheat, there's always a next step. Somehow, cheating at games is a prime objective to far too many people.
  - foxhill 5y ago
    Valorant’s anti-cheat is considered to be good, but hacks are available. it’s also incredibly invasive - running in the kernel, meaning it must be enabled at boot time - something i don’t think is either necessary or desirable. faceit/esea isn’t much better. and i suspect it’s only better at all because of the higher barrier to entry. as for MFA - Trust takes this into account in determining its score. to stay on topic here, i think what i was originally trying to say was, yes, having humans in the mix for self-moderation is a good idea in principle. in practise, it’s easy to manipulate because it’s very hard, potentially even impossible, to determine if you actually have a human using the machine. well, except in china, i suppose. but i don’t think we want to live in that world :)
PoignardAzur 5y ago
> Anticheat has to be defense in depth. I'm pretty sure every layer of defense you can think of, they've tried. - Machine learning to spot obvious cases? Check. - Federated moderation? Check. - Various trust schemes to weigh accounts based on how much they payed, whether they cheated in the past, etc? Check. The only thing I don't think they've tried too much is sending false/stale game data in cases where they know the client has no way to know otherwise (eg lying about an enemy's position when the player has no way to see the enemy). It's an arms race and it's hard to stay ahead at scale. (though I'm not actually sure I trust accounts of how prevalent cheating is; I haven't had that many cheaters in my own games, and a lot of complaints I've seen come from either free users or people who admitted to cheating themselves, which is what I'd expect to see if Trust Factor worked; though it's a crap deal for free users)