Identify As A Human Does: A Pathfinder of Next-Generation Anti-Cheat Framework for First-Person Shooter Games
Offical Inspection Escapees Demo
Dataset
Cheat Sophistication
Playback Speed
Blatant aimbot escapee at slow motion. Sniper rifles generally require an open scope to hit accurately, and in this demo we can see that the cheater hits the enemy quickly and accurately and without opening the scope. Before the kill, the aimer quickly moves to the enemy's position, and after the shot is fired, the aimbot used artificially imitates the jerky movement of the mouse after a human presses the mouse. But as you can see, even in slow motion, the jiggle is still done within a frame, making it look ripped and unnatural.
Note
1. HUDs (in-game UI overlays, e.g., health bar, navigation map, avatar's ID, kill info, etc.) that may disclose any information are hidden to fulfill anonymous submission requirements;
2. X-Ray (highlighted character borders can be seen through obstacles) is activated for better illustration purpose. In real-world matches, the normal players will not obtain this function.
Selected Feature Average Growth Rate
Selected Feature Average Growth Rate
We chose representative features to further illustrate the differences between the cheater and normal players at the data level. For comparison purpose, we introduce average growth rate, which represents the percental difference between the average and the selected sample.
$$Average Growth Rate=\frac{{\text{Value}_{\text{cheater}} - \text{AVG}_{\text{normal}}}}{\text{AVG}_{\text{normal}}} \times 100\%$$
Where \(\text{Value}_{\text{cheater}}\) denotes the feature's value of the cheater; \(\text{AVG}_{\text{normal}}\) denotes the average value among all participants of the same feature in the match.
▼ represents the decrease compared to the feature's average value in the match; ▲ indicates the increase compared to the feature's average value in the match.