How does crowdmark detect cheating?

Could you elaborate on the mechanisms employed by Crowdmark to identify potential cheating among its users? Do they utilize advanced algorithms to analyze patterns in submissions, or do they rely on manual review by experts? Is there a particular focus on flagging similarities between answers, or do they also consider other factors such as submission timing and IP addresses? Additionally, how does Crowdmark ensure that its cheating detection methods are fair and do not unfairly accuse innocent users?