通用於第一人稱射擊遊戲外掛檢測機制之研究

Abstract

隨著科技的飛速發展，玩家可以在一台個人電腦上遊玩各種類型的遊戲，在各類型遊戲中，網路遊戲是大多數玩家最喜愛的遊戲類型，玩家為了在網路遊戲中獲得更好的成就，開始使用外掛程式達成個人無法實現的目標，基於上訴原因，作弊偵測成為了遊戲廠商的重大課題。本研究提出了一種基於影像辨識並以數據檢測輔助的作弊檢測系統，並分別使用VGG16、ResNet50、MobileNet V2、Xception和Inception v3 對誠實玩家和作弊玩家的瞄準軌跡進行檢測，研究結果表明，Inception V3 能最準確的分辨誠實玩家與作弊玩家。

With the rapid development of technology, players can use a personal computer to play a variety of games. Of all kinds of games, online games are the most popular game type for most players. To obtain better achievements in online games, players begin to use game cheat to achieve goals that cannot be achieved by individuals. Due to the above, cheat detection becomes the most important issue for game manufacturers.This research proposes a cheat detection system based on image recognition and supplemented by data detection and compared VGG16, ResNet50, MobileNet V2, Xception, and Inception V3 in an attempt to classify honest players and cheater aiming trajectories. The results of the research show that Inception V3 is the most accurate detector of honest player aiming trajectories.