I have gone ahead and e-mailed the security team with some details. Thanks for the great suggestions everyone! This thread has served its purpose : )
CCP mentioned theyâre already working on something like this, aka using AI to detect anomalous behaviour. Iâm not convinced that it is going to help in the long run, but Good Luck nonetheless.
2 Likes
Agreed Solstice. In the long run I suspect such systems will simply become a form of adversarial feedback that will train bots to be harder to detect.
That said, itâs the best approach I can think of and ya have to try something⌠It should at least make bots more computationally expensive.
2 Likes
I have to say, I agree. Maybe you really do have some value you can add, but coming at it via the forums is ratherâŚodd.
1 Like
Careful. I made that point once that one could up in an âevolutionary arms raceâ and that we are dealing with complex adaptive systems and holy crap did one person on the forums lose their â â â â . 
But yes, this kind of thing is a danger. We see it with spam and spam filters. Why do you see spam with obvious spelling errors? The spam filter which often use a Bayesian filter. To get around the filter the spammer will use grossly misspelled wordsâŚwhich to some extent become self-defeating.
And rising computational expense can dissuade some botters from continuing. The whole idea is raising the cost of botting drives out the lower cost botters.
3 Likes
Exactly best way to kill botâs is to make it not worth the cpu / electricity costâs make them compute harder make them jump through more loopâs work them so hard that the profitâs they get at the end literally arenât there anymore and they will stop.
1 Like
Uhm, that wouldnât even work with me, and I used straight C, PERL, and LISP.
Today, you guys are using objective-c, ruby, and etcetera that I donât even comprehend. The power available to you is not a hundred times what I was used to, but more like a million times.
This game is not going to present a challenge to bots, computationally.
1 Like
Just put in those capcha things like when you log into stuff everytime you undock. click all the tiles with cars in them! Itâs but a small price to pay everytime you undock.
The key flaw iâm seeing with your whole plan here, is that you are assuming CCP doesnât already have systems in place to identify bots, and that, without having actually seen the systems they already have in place. assuming that you can improve upon them. and, even if you could, that ccp would be willing to give you access to their database and code on the chance that you could develop something better than they already have. (not to mention how long it would take to test your new system to see if its any more effective than what they already use)
Maybe you can come up with a better way to identify bots, maybe not, but since it seems you have overlooked something so glaringly obvious, I am somewhat skeptical of your ability to design anything that could properly identify bots as opposed to players simply doing highly repetitive tasks without missing something similarly obvious, resulting in either a horrible success rate, or a massive number of false positives.
This has nothing to do with a specific programming language or how challenging the gameâs activities are to write bots forâŚ
The field of anomaly detection is huge, with many specialized approaches for various fields (like detecting bots in games, for example). I would guess the OP is using neural networks and/or clustering techniques, but there are many other possibilities as well. Beyond specific existing libraries restrictions, none of these techniques are tied to a specific programming language.
The idea is to classify characters as players or bots based on their behavior, represented by feature vectors. A good system of this type adds a requirement for any successful bot beyond simply functioning in the role it is designed for in game⌠it must also appear to be a player based on a feature vector that it doesnât even know. It can only really learn what behaviors to avoid and which to pattern through trial and error, with the errors likely being detection. It doesnât make it impossible to go undetected as a bot, but it becomes much more challenging. And even when such a bot is built that can avoid detection, it probably has to compute more than just the direct challenge of its in game task (hence the potential for increased computational expense).
LISP was used as the language of choice in my era because the code, itself, was a list. It could modify itself. Somewhat frowned upon, really, in any other language, and difficult to implement, to boot.
Not really. Whatever you code it in, youâre gonna have to hit Bayesâ Theorem, and itâs gonna be your tool. What is really important to building any sort of âinteligent system,â is the input data quality, that it is complete.
A neural network fed garbage makes mistakes. You feed it all the data, and complete data, as in over a long enough time period of complete data, and it will churn out relationships you never knew existed.
You can do more, now, Iâm sure.
The way to beat bots shouldnât make the game tedious for actual players.
That type of approach will just piss people off no end.
2 Likes
Yes, it will.
Greater minds than mine have tried this crap. And failed to get rid of bots. What one man can do, another can do, or undo. Sorry. Itâs just the reality. Security is an ever present battle, where youâre always a step behind.
LISP is nice, but itâs been a long time since Iâve needed it. Clustering and neural nets are more about parallel computation, often involving simple matrix operations. The core of many neural network libraries is written in C.
Most neural network and clustering implementations do not use Bayesâ Theorem. Exceptions exist. Bayesâ Theorem is very useful, but itâs not always going to âbe your tool.â Markov chains, for example, are very powerful and rely rather exclusively on Bayesâ Theorem.
This, I agree with. Which is why I said in the long run Iâm skeptical this approach will have much of an impact beyond making successful botting a bit harder than it is now.
Removing alpha would probably be a more effective way to limit bots.
No. I get that you have advanced beyond me, technologically.
Morally? No. Iâm not voting to harm innocents en mass to salvage peopleâs pride of accomplishment.
I was definitely kidding lol oh man Iâd quit 
I know another game where botting is out of hand, perhaps you can contact them?
Itâs called Black Desert Online and itâs owned by Pearl AbyssâŚhey, wait a minute!!
No they did not, because bigger inerests are around âŚ
I hope CCP gives OP a listen, at least. Sounds like partnering could be mutually beneficial, and at no cost to CCP in terms of money or staff time.
I guess I need to, again, point out that theyâve announced something like this last year already.