SARD Update

Hi everyone,

Since the launch of New Horizon in August 2024, we've been working hard to roll out SARD Anti-Cheat. It's our next step toward a fairer, more enjoyable APB experience, and SARD has been a committed partner in helping us build something uniquely suited to the challenges of our game.

Today, I want to give an update on all-things-SARD.

The Goal of SARD Anti-Cheat and Our Unique Approach

I think its important to start with a recap of what we're trying to achieve.

My vision for anti-cheat in APB Reloaded is straightforward: we want to create a system where we never have to ban anyone outright. The typical ban approach is black and white - either someone's in or they're permanently out - and it demands an extremely high level of confidence in the data we use to justify those bans. Worse, each ban we do gives cheat makers valuable information about whether they are being detected or not. This allows them to improve their cheat and make it harder for us to catch them in the future.

I'm done fighting that battle. We're now taking a different approach by implementing cheat points.

The goal is to bring together detections from multiple systems, each with its own strengths and weaknesses, and assign points based on that combined data. Some systems can spot cheaters with pinpoint accuracy, but only under very specific circumstances. Other systems require extended amounts of time to detect cheats, but the detection might only rank 80% confidence. By assigning and tracking cheat points across all these inputs, I feel we will be able to manage cheaters in a way that's dynamic and layered, without requiring an irreversible "ban" decision to course correct the player.

This new system has (2) major sanction actions at its disposal:

• Escalating time-limited bans for specific features or game wide
• Matchmaking rules that take cheat points into account

Here's how we've designed this approach to be fair and flexible for players:

1. Minimizing Impact from False Positives: Because this system relies on multiple detections and a flood of data, a single hit won't necessarily trigger major consequences for a player. We've designed the system around thresholds that need to be hit before the system steps in to segment players away from the general population. This way, if one detection was incorrect, it's unlikely to affect that player's experience. When players do accumulate enough points, they're matched with others at similar cheat-point levels, keeping things balanced.

2. Cheat Points Expire: In our system, cheat points aren't permanent; they fade out over time. If a player stops doing things that trigger detections, they'll gradually reintegrate with the regular player pool. This design lets us handle minor or accidental flags without long-term impact and gives players a path back if they choose to play fairly. This feature also allows the system to self-correct over time, making it fairer for everyone.

3. High confidence detections trigger a timed ban: If a player is misbehaving in a specific gameplay mode, we can ban them from that mode. If we get a high confidence detection, then we can kick them out of the game for escalating windows of time while we wait for cheat points to accumulate and segment them from the non-cheating players. Importantly, this is a great new disciplinary feature that we can apply outside of cheaters. The game has long needed more granular features to punish griefing and toxicity. If you spew racist garbage over chat, then we will soon be able to block you from chatting.

4. Leveraging All Our Detection Data: The cheat points approach means we can use every bit of data available-client-side, server-side, and AI detections - without needing each detection to be 100% certain before taking action. In traditional banning models, server-side and AI detections alone wouldn't give us the confidence needed for a permanent ban, as they can sometimes yield unreliable results. But with cheat points or time limited bans, we're able to draw on data from every system to create a well-rounded, adaptive approach to handling cheating.

Feedback from the Community

Next, I want to acknowledge that things haven't gone as planned. Many of our players have posted on social media, on the forums, notified GMS, or sent information to customer support. Here are some of the concerns I found that players have raised:

1. Perceived Ineffectiveness in Detecting Cheaters: Many players feel that SARD has not significantly reduced cheating within the game, and in fact that cheating has actually risen. Reports indicate that known cheaters continue to play without repercussions, leading to frustration among the community.

2. Technical and Performance Issues: Some users have encountered technical problems attributed to SARD, such as initialization failures preventing the game from launching. Additionally, there are concerns about the system's impact on game performance, with players reporting increased resource usage and decreased frame rates.

3. Privacy and Intrusiveness Concerns: The implementation of SARD has raised privacy issues among players, particularly regarding its kernel-level access and the extent of system scanning. Some community members are apprehensive about the potential for false positives and the overall intrusiveness of the anti-cheat measures.

Before I move on to some of the explanations, I want to apologize to you, the players. This rollout has hit a bunch of different problems, and the game suffered. Period. I am sorry. I can't address of these concerns today, but know that I am tracking them.

Issues We've Encountered

I also want to restate that there isn't a clear roadmap for what we're trying to achieve. To my knowledge, no other game is attempting this. I knew going in that we might have to "break some eggs" before we got the "omelet right". Over the past 90 days, boy oh boy, have we broken some eggs.

Without mitigating my apology, I want to provide at least a surface level explanation for some of the problems.

1. SARD + Steam Compatibility: Despite several months of testing, a critical issue surfaced right after launch: the SARD Kernel Driver wasn't playing well with Steam, effectively blocking Steam players from logging in. After quickly identifying the problem, we had to make the tough call to disable the kernel driver to get our Steam players back in the game.

2. Sanction/Rules Engine Delays: Our custom sanction and rules engine is the heart of our approach, built to take in data from SARD and apply cheat points based on a clear set of criteria. Unfortunately, it wasn't fully functional in August, which held us back from making the most of the incoming data.

3. Certification Delays with Microsoft: As soon as we identified the Steam compatibility issue, the SARD team moved fast to develop an update for the kernel driver. Unfortunately, getting this update certified by Microsoft has been a slower process than we'd hoped. Without that certification, we can't sign and deploy the updated driver, so for now, we're stuck waiting for Microsoft's green light.

4. Training Server-Side AI Systems: The server-side AI detection system in SARD is powerful and multi-layered but it needs time to learn and adjust to patterns. You've heard that before, but the important piece is that we took a very careful approach to only train one piece at a time. Originally, we believed that the client-side kernel protection would be "good enough" to stand alone for a short period while the AI trained. Once the driver needed to be disabled, we were left with something that barely functions.

Moving Forward: Solutions

So that's where we've been. But this blog would be meaningless without an actual plan to resolve these issues.

In fact, a month ago, we had an internal meeting to decide if things were so bad that we needed to abandon this effort and just go back to a different traditional anti-cheat. However, as bad as things are, I decided against that. We've put a lot of effort into the new approach, and we're going to stay the course.

Here is how we are going to move forward:

1. Fix Kernel Driver + Other Solutions for Client-side Detection: SARD has been hard at work fixing things. We have a new driver that works with Steam, and we have several other solutions to prevent this from happening in the future.

2. Certification Process: This week, we made the decision to put Little Orbit in charge of the certification process with Microsoft. We have had a successful relationship there for over 12 years, and our business is much more established than SARD. We'll keep everyone posted as soon as there's any movement.

3. Sanction/Rules Engine: The rules system is now up and running as intended, actively analyzing data from SARD. In some cases it is applying cheat points, but we're monitoring this very closely as we expand the types of data we are acting on. This system is the lynchpin piece of our cheat-management strategy, and I'm excited to be finally rolling it out.

4. Server-side Training: A month ago, we got our initial look at the server-side training data. And while I think it was a good start, I didn't feel confident enough to start acting on it. But last week, we reviewed the latest detections, and frankly, it impressed the hell out of me. The server-side properly identified many different cheats with near 100% accuracy - even cheats we didn't know were out there. I've never seen anything like it. APB has never had an active client-side + server-side anti-cheat.

In the coming weeks, once we get the kernel driver operational, I believe this will be the best anti-cheat APB Reloaded has seen to date, and we can get back to having fun.

As always, this game wouldn't exist without you guys. I appreciate the feedback you've given. I hear you. We're working on it.

I look forward to sharing an exciting update on the next major content release which will include new features as well.

Thanks, Matt

Got any questions? Want to chat with other players? Then discuss this article in our Forums! You can also follow us on Twitter and Facebook.