PaimonAnalyzer | Source Included v1.0

[~] Feature Introduction
PaimonAnalyzer is an unofficial implementation of the server-side AIAC3, designed to handle billing and detect whether AIAC3 format mouse data is generated by hakced clients. In short, it can detect several cheats that are considered extremely difficult to identify.

This product includes the PaimonAnalyzer program and source code, featuring multiple functions and advantages:

Multi-Tenant & Flexible Billing
PaimonAnalyzer allows you to create licenses with a complete billing system:
- Expiry Date
A license has an expiration date, starting from the first time the license is used with AIAC, and it will automatically terminate when expired.
- IP Binding
Licenses are automatically bound to the access IP, and you can limit the number of times a license can change IP to prevent sharing.
- Tokens
Like mainstream AI applications, PaimonAnalyzer bills based on Tokens.

Model Market & Sharing
PaimonAnalyzer includes a revolutionary model marketplace.
- Allows Uploading Datasets
Users can upload their own created datasets.
- Paid Fine-Tuning
Users can pay in Tokens to fine-tune models on their datasets, optimizing the model to suit their needs.
- Sharing Profits
Users can share their fine-tuned models for others to use and earn Tokens. This greatly encourages users to share your products.

Security & Extensibility
- Includes Extras
I provide a range of extras, so even if you don’t intend to use AIAC3, you can still integrate PaimonAnalyzer into your network and anti-cheat solutions.
- No Dependencies
PaimonAnalyzer’s dependencies are all open-source under very loose licenses. It’s purely Java (no Kotlin or Scala), compatible with Java8, and has no native components. This makes extending and calling PaimonAnalyzer extremely easy.
- No Statistical Algorithms
PaimonAnalyzer contains no game-specific algorithms tailored to MC gameplay. This makes it adaptable for other anti-cheat or AI applications without being restricted to Minecraft.
- Retained interfaces
A number of enhanceable interfaces have been retained. For example, the password for AIAC3's encryption protocol is set as a variable. This can be very helpful if you wish to use it in your own non-open source anti-cheat solution.

You also don't need any statistical knowledge to maintain it. There's no need to update algorithms, even if cheats are updated. You simply collect data and retrain the model, which is a lightweight and automated process.
(However, basic linear algebra knowledge can be beneficial if you wish to optimize PaimonAnalyzer's internal AI.)
[~] Contact
Discord

[~] Testing
You can test PaimonAnalyzer’s detection effects at Loyisa's Test Server (choose AIAC3). A web-based testing server is not available at the moment.

[~] EULA Exception
In addition to the rights granted by the standard EULA, after purchasing and rebranding, you can host, modify, relicense, and resell PaimonAnalyzer or use it as your anti-cheat backend without needing any permission from me.

Since the product includes open-format model weights and source code, DRM measures are impossible — I cannot ensure you aren’t purchasing the product just to obtain the source code. If you have any doubts about the product’s content, please consult me via Discord before purchasing. Only buy after you are sure you fully understand its functionality.

[~] Extras
To help you fully utilize the product, I’ve included the following extras:
1. Minimum Implementation of AIAC3-Killaura Protocol
This is a dependency-free plugin (yes, it does not rely on AIAC3), designed to work on Spigot 1.8.x. It’s the minimum implementation of the AIAC3 Killaura cloud-check protocol.
It collects and pre-processes player Yaw and Pitch information, sends it to PaimonAnalyzer, and based on the returned results, punishes players using Killaura. You can use this to write your own anti-cheat plugin with no security or licensing concerns. It can also be used to generate datasets for training.
2. Default AIAC3 Weight Files
Pre-trained weight files used by the official AIAC3 cloud service, obtained from huzpsb and converted into PaimonAnalyzer format.
3. Some Datasets
huzpsb refused to share his datasets over concerns of biometric information leakage, so I’ve created a few (small, but decent quality) datasets for training purposes.

I provide Best-effort support for these extras. With good conscience, I promise they will work in most cases, but I cannot guarantee they are as flawless as PaimonAnalyzer itself. I’m willing to offer free support for them during my spare time, but it may not be as timely or thorough as the support for PaimonAnalyzer.

[~] Model Card
Details of the models mentioned in the extras:
Parameter Count: 14,760
Training Set: [Undisclosed]
Test Set Accuracy (disclosed): [82%]
My Own Test Set Accuracy: [77%]

• Accuracy refers to single-prediction accuracy. With multiple predictions (via the vl system), accuracy can stack significantly.
• PaimonAnalyzer has a built-in 3-step voting system, which is not accounted for in the data above. Therefore, if you directly test with PaimonAnalyzer, the accuracy might appear unusually high.

[!] Warning
Unless you have experience developing pure Java Spigot plugins or are developing your own anti-cheat plugin, or if you have already purchased AIAC3 (unfortunately, it’s currently only sold on Tencent), you should not buy this product.