Ferramenta antitrapaça para jogos digitais multijogador desenvolvidos na Unity com integração do ML-Agent

Abstract

Cheating in online multiplayer digital games is one of the biggest problems and provides an unpleasant experience for honest players. This paper presents the categorization of these cheats according to the taxonomy of Yan and Rendell (YAN; RANDELL, 2005). In addition, a bibliometric study was carried out in the database Web of Science to determine how academia solves these problems. The works were sought in two moments. In the first, the key terms were used game* and (cheat* or security) with a period of five years and in the second moment with search terms online game* and cheat*. The paper also evaluated studies that used the Unity tool in the game development. With this analysis, it was observed the use of reinforced learning to identify possible malicious players. It is proposed a prototype using the Unity Engine is proposed to create an environment in which one can use the ML-Agents library and validate through artificial intelligence whether a player is human or a bot. The prototype implementation was developed in Unity with the C# language. The algorithm created with most of the created classes. Afterwards, the integration steps of the ML-Agents library with Engine Unity are detailed and it is executed the analysis about the results. [resumo fornecido pelo autor]