GaitAuth™ Feature Scoring

Once trained, UnifyID GaitAuth™ models are used to produce scores for collected gait features. Those scores can then be utilized in your mobile application’s authentication flow.

Producing Gait Feature Scores

Android

model.score(gaitFeatures, new OnScoreListener() {
    @Override
    public void onComplete(List<GaitScore> newScores) {
        // Utilize scores for provided features
    }

    @Override
    public void onFailure(GaitModelException e) {
        // Error computing scores for provided features
    }
});

iOS

gaitModel.score(features) { result in
    switch result {
    case .success(let scores):
        // Utilize scores for provided features
    case .failure(let error):
        // Error computing scores for provided features
    }
}

Interpreting Gait Feature Scores

A GaitScore contains a timestamp date and a value, which is a number in the range (-1.0, 1.0):

A value above 0.0 means that it is more likely the user than an attacker. A value below 0.0 means that it is more likely an attacker than the user. The threshold that you choose to base authentication results on depends on your use case. Each business may have different considerations as it relates to false positives and negatives, but in general:

Scores between 0.8 and 1.0 reflect a very high level of confidence that this is the user. Scores between 0.6 and 0.8 continue to show high confidence that this is the user, but have a slightly increased likelihood that a potential attacker could have produced this score. Scores below 0.6 (and above 0.0) are still more likely to be the user than an attacker, but continue to increase the possibility of letting an attacker through.

Finding the exact threshold that works best for your business may require continued evaluation. To begin your integration, a score of 0.6 or higher is recommended. If you need any help or have any questions, we’re happy to help!

Updated: