Lantern takes a different approach to detecting X-ray users. Instead of relying solely on simple heuristics or ore-to-stone ratios, this plugin utilizes a CatBoost machine learning model to analyze player behavior in real-time.
This project started as an experiment on my own SMP server and yielded impressive results, but please note that this is currently an Alpha release.
How it works The detection model is trained on a large dataset of legitimate player behavior. Because gathering high-quality data on cheaters is difficult, the plugin is currently configured with a philosophy of "innocent until proven guilty."
- The model is heavily biased to avoid false positives. It requires a high confidence level (default is 90%) to flag a player.
- While this makes it very safe for legitimate players, it may miss very subtle cheaters. However, if a player is blatantly tunneling for ores, the plugin will detect it with high confidence.
- I strongly recommend using Lantern for Staff Alerts (via chat or Discord) rather than auto-bans. On my server, it proved to be an excellent tool for guiding moderators to suspicious players they might have otherwise missed.
Machine learning relies on data. If you run a server and are willing to help collect gameplay data to refine the detection of cheaters, please join our Discord. In exchange for contributing datasets, I can offer early access to the improved Premium model.
Key Features
- Smart Analysis: Uses 19 different behavioral factors to judge mining patterns.
- Discord Integration: Native support for Webhooks to send alerts directly to your staff channels.
- Multilingual: Supports English, Russian, German, Spanish, Chinese, and Ukrainian.
