Digital keys and systems for preventing relay attacks

A computer system for training a machine learning model for preventing relay attacks, the system comprising:

• a plurality of first portable electronic devices, each intended to be borne by a user and each comprising:
• a first wireless transceiver for transmitting a beacon signal, and
• a first acquisition module for acquiring, during at least one acquisition session, signals relating to the movements of the first portable electronic device;
• at least a first road vehicle comprising:
• a second wireless transceiver for receiving the beacon signal, and
• a first processor for opening or closing, contactlessly, at least one access to the first road vehicle according to the beacon signal;
• and
• a machine learning server for training the machine learning model on the basis of the signals from the acquisition modules, the machine learning server being provided to:
• obtain the signals generated by each first acquisition module, in response to the associated first portable electronic device being moved by the user, when the first processor opens or closes the access to the first road vehicle,
• calculate, for each acquisition session, a direction vector of the first portable electronic device, on the basis of the acquired signals,
• associate, for each acquisition session, the direction vector with a first predefined item of information which is either indicative of the movement of the first portable electronic device in the direction of the first road vehicle, or indicative of the movement of the first portable electronic device in the opposite direction to the first road vehicle,
• create, for each acquisition session, a characteristic vector on the basis of the acquired signals and the first predefined item of information, and
• apply a supervised classification algorithm to the machine learning model according to all or some of the characteristic vectors.