A teenager has developed an AI system that can detect over 98% of cheaters in competitive online game CounterStrike: Global Offensive.
The global eSports industry has exploded in the past several years, with tournament prize pools now reaching into the tens of millions of pounds and top players around the world becoming millionaires. With such high pressure to succeed, some competitive online games have become breeding grounds for people using cheating software such as auto-aim and wall-visibility hacks to get an edge on the competition.
One teenage player of the game CounterStrike: Global Offensive has had enough of cheaters ruining the game and has developed an AI that can successfully detect cheating with a reported 98.36% success rate. The player, who goes by the name 2eggs, was also recently awarded over $11,000 US in bug bounties by CounterStrike's developer Valve for helping to identify security risks, and has helped develop databases that keep track of banned players.
The AI, named HestiaNet, was trained on a respository of video footage of suspected cheaters who are reported by suspicious players online. These videos are normally voted on by expert players who study them for visual clues of obvious cheating, so the AI used a player ban database to classify which videos turned out to be actual cheaters and which were genuine players.
The resulting AI can do the job of hundreds of experts studying the footage and has achieved a 98.36% success rate, with 15,104 out of 15,356 players it classified as cheaters going on to be banned. The AI periodically checks the ban status of users it has reported as cheating and updates its records, so it gets more accurate over time as it collects more data.
Source: Massively Overpowered, Rock Paper Shotgun
