Gaming Anti-Cheat: Real-Time ML with Carly Taylor & Joe Reis

0:04Maybe get some footage of everyone walking around later because it is amazing. I think people are really excited. People are excited about data. San Francisco is lava red, so that's pretty cool. >> I'm Joe Reese. >> Carly Taylor. >> I'm Carly Taylor. I'm a field CTO in gaming here at Datab Bricks. And

0:26>> I'm Joe Ree. I am a data engineer, author, speaker, educator. Um, here at Databricks Data and AI Summit. >> You're the chief nerd. Also, >> I'm chief nerd. Yes. >> Yes. >> So, I'd say that data in gaming is becoming more and more important. It used to be treated almost like a secondass citizen. Game companies have

0:50always collected a lot of data, but they used it more for the development process and didn't really leverage it for anything like analytics, marketing, recommendation engines, machine learning, anti-toxicity, anti-fraud until I saw an explosion maybe in the last 5 years. So now all of the big players are realizing that this is super important. And I'm even talking to a bunch of indie studios and they're

1:13planning their data strategy alongside doing green lights for their studio. So, it's incredible to see it being pushed as more of a first class citizen and actually a blocker to ship, which it never used to be. You know, you can ship a game without data and analytics, but do you want to? >> Those were the good old days of back when I was just running, you know, wild on

1:35video games and just, yeah, cheating, not getting caught. And now it's Now I have to show up to conferences. I have nothing else to do with my time. >> Would you have you may have been a professional video game cheater hacker if you didn't become a data engineer. >> Yeah. >> Yeah. >> Yeah. >> Well, there's still time. >> Don't let your dreams be dreams. >> Oh, thank you. It's very inspiring.

2:02I >> would say obviously learn the fundamentals. Those don't change. But couple that with I would say going full in on AI. There's a lot of really cool stuff happening it seems on almost an hourly basis. And so you know I I I would just say master master master the current state of AI you know and become part of the

2:24future. The cool thing is in a lot of ways a reset button has been hit on the entire industry. And what I tell especially aspiring uh not just data engineers but software engineers, data scientists like you have a chance to not just get into an industry but reinvent it and come up with new roles, new ways of doing stuff that even you know old people like us couldn't figure out cuz we're you know stuck in our ways or

2:46whatever. But like there's a such a a huge once in a generation opportunity to do something completely new, completely different. Um, and I think really set the course for not just, you know, the next couple years, but for your career, for the next decade. So, I think it's an exciting time to be at the field. >> So, >> I like that. Reinvent things. >> Now is the time, right? >> It is. It really is, though. Yeah.

3:09>> Like, I'm going to become a video game cheater. Just kidding. >> Oh my god. Yeah. And I'm going to get you. >> Yes. Probably will, actually. >> I'm like staying. Every step you take, I'll be watching you. >> Oh, okay. Well, Well, it's actually interesting that you asked that. So, I have used Databricks long before I started working here. And I think the biggest way that it helped

3:31me, especially in my role, was the ability to do real-time machine learning and analytics, right? Like having that data almost as soon as it left the game server, being able to make recommendations with it, being able to find cheating and toxicity, right? Like those are things that you can't really delay. And then being able to take in all of our data. So, we have obviously a massive amount of structured data, but

3:53we had a ton of unstructured data, too. Logs, crash files from the servers, um logs from, you know, things like the kernel drivers. They put out these massive logs. And being able to put those somewhere, parse them later, deal with them, uh was like so so powerful. And back in the day, to do any of that, you would have had to probably use like 10 different products stitched together.

4:15And for us, it was like I had a team of machine learning engineers. I had data scientists, I had business analysts. They were all in the same tool together, which was really powerful because they could share code and best practices. We built our gold layer in our database right in Databricks and had everyone knowing like how to use it, what KPIs uh meant, what they were, how we defined things as a business, and it centralized

4:38and governed everything we had to do. >> Isn't that crazy? >> That's crazy cuz that's all the data I generated. So, >> wow. Yeah. That was just you running around and losing every gunfight. >> Yeah, I'm not dumb. So, I mean, just it's a lot of total noob. >> Yeah. >> Yeah, it's uh it happens. >> Yeah, we saw you. You actually had a

5:00negative KD. >> It's crazy. Negative kill death ratio. >> Yes. >> Well, it's fun. I play with my kids a lot, though. And so they they always do come across uh game cheating um you know games like Fortnite for example and uh yeah it it irritates them to no end for sure. So >> well in a competitive environment you have to believe that you're on a level

5:22playing field or else it's not fun >> you know and so >> yep that's why I don't play video games. >> That's why I'm kidding. >> Anything and everything you can think of you need to bring to your arsenal to catch cheaters. So, we're talking real-time machine learning, uh, analytics and heruristics on the game server for like ultra real time, and

5:45then you're using things like a kernel driver to try to inspect what people are actually doing on their PC and if they're injecting nefarious code into the game or exfiltrating data that they shouldn't have. Um, for things like pixel bots, which are computer vision hacks that look at the screen, you're trying to see superhuman aim if they're too fast. There are scripts people run

6:09that make that reduce recoil and make it easier for them to fire their weapon. This is all for first-person shooters, by the way. Um, you can see those in the data because the data is quantized. It doesn't look like normal movement. It actually has these patterns that you can see with like discrete nonhuman movement in the joysticks. So, like honestly, any touch point for a game that you can

6:32think of, cheaters are trying to exploit and anti-che teams are trying to protect. So the entirety of the game executable. >> Let me ask you this. So with uh AI, maybe agentic AI, what do you foresee as sort of the future of the anti- sheet field? I'm guessing this stuff is going to keep evolving at a

6:54very exponentially fast rate. >> It's an arms race like everything else. So, we actually have a video that I'll send to you that you can link to of someone who built a fully autonomous aimbot and it's a robot and it has hands and it runs a computer vision algorithm on its screen and it can play the game, right? >> That's pretty cool.

7:16>> And so then how do you find that? Like you basically need like proof of life now for human gamers to differentiate them, right? So that is going to be a future where like it's 100% dependent on heruristics. So hopefully you have a deep roster of professionals and people you understand who are like the best ever and then you benchmark against that, right? I think it's a similar way that you would probably find people

7:39likely to be cheating or doping in sports is what is the best capable of assuming we continue to push the envelope and people get better. What is a normal rate of progression in terms of skill and where is someone an outlier and then where are they continuously an outlier and if there's anything you can find that might flag it to make it look not great. The good thing about cheaters

8:01though is that a lot of them are trolls. So they want you to know that they're cheating which means they're not trying to blend in all the time. Some of them do. They're not trying to blend in and just seem like a really good player. Some of them want you to be like, "Oh no, this game is infected with cheaters." cuz they like to grief. Those people are easier to find than the people who try to blend in and seem maybe like they're just a professional.

8:24>> Okay. So, it's not like Lance Armstrong would be like, >> "Do you think he also cheats in video games?" >> Does he play video games? >> Who knows? >> We can find out. He might be one of the best players in the world, actually. >> Oh my gosh. What if he was? >> Yeah. >> And then no one ever believed that he was like really good, >> but he's like >> he ruined it. His name is Gamertag is Lance Armstrong.

8:47>> It's I'm not cheating. >> Yeah. Auntie Roids. Um >> Yeah. I don't know. But I mean it be like him like having like like a doping company shirt on his logo on his jersey. >> There's that uh meet that's going to be like where they're allowing people to dope. >> You seen this? >> No. That's pretty cool. >> Yeah. They're like letting there's like a whole competition where they're

9:10letting swimmers do performance-enhancing drugs to see what they can do. Like a bunch of athletes are going to this competition, but they're allowed to be on peeds. >> That's cool. >> See some of the guys there like yolked. >> Yeah, they look like me. I'm kidding. So, but that's actually an interesting point with the with the game. stuff. So maybe is there is there a situation

9:32where you you kind of just have a tournament of all the agents, all the bots, just let them do their thing and that's kind of your benchmark. Like that's what that's what that would look like. >> Them just mess with each other. >> So that is a known thing in the gaming industry. Players call it cheater island. So it's where you put them all together so they can just ruin each other's day and not ruin normal people's day. Um, the thing about cheaters, too,

9:54is that they like to know they're ruining someone else's day and they don't care to ruin each other's day. So, they want to be in Genpop, right? Like, they want to break out of Cheater Island because it's not fun for them. It's only fun for them to hack against like a normal person. >> We could do a whole psychological study. >> Basically, be a psychic here. >> In fact, we should do that right now. Here, >> let's unpack this. Why are you the way

10:16you are? >> Yes. >> I don't know. >> Don't know. >> Sorry. We I mean it's it's the Joe and Carly show. Um >> uh >> yeah. >> Yeah. I don't know man. >> I don't know. Chess though. Chess is interesting. >> Chess is very interesting. But see there's like a very well understood

10:38skill algorithm for chess right. Like ELO is the gold standard. So like if someone is consistently outperforming. >> Yeah. >> Then you just have a documentary made about you. Yeah. >> Yeah. Exactly. >> Yeah. People say you're a cheater and then you have to play live and I'm totally not doing that. >> Then you go to crazy links to cheat live games. >> You remember that guy? >> Yeah. >> Yeah. Crazy. That was wild. >> Yeah.

11:02>> Don't cheat at games. >> And come to Databricks AI Summit. >> And come to Databricks AI Summit. Yeah. >> Thank you.