Learning method and learning device for training obfuscation network capable of obfuscating original data for privacy and testing method and testing device using the same

A testing method for testing an obfuscation network capable of obfuscating original data for privacy, comprising steps of:

• (a) on condition that a learning device has trained an obfuscation network to acquire a trained obfuscation network by performing processes of (i) inputting first training data into the obfuscation network, to thereby instruct the obfuscation network to filter frequency information of the first training data and thus generate first obfuscated data, and (ii) (ii-1) inputting the first obfuscated data into a first learning network designed to perform a first task, to thereby instruct the first learning network to generate first characteristic information by performing learning operation on the first obfuscated data, (ii-2) generating a first task loss by referring to the first characteristic information and its corresponding ground truth or a first task-specific output, generated by using the first characteristic information, and its corresponding ground truth, and (ii-3) training at least one of the first learning network and the obfuscation network through a backpropagation of the first task loss, a testing device inputting second training data into the trained obfuscation network, to thereby instruct the trained obfuscation network to filter frequency information of the second training data and thus generate second obfuscated data;
• and
• (b) the testing device (i) inputting the second obfuscated data into a second learning network designed to perform a second specific task, to thereby instruct the second learning network to generate second characteristic information by performing the learning operation on the second obfuscated data, (ii) generating a second task loss by referring to (ii-1) the second characteristic information and its corresponding ground truth or (ii-2) a second task-specific output, generated by using the second characteristic information, and its corresponding ground truth, and (iii) training the second learning network through a backpropagation of the second task loss.

A learning device for training an obfuscation network capable of obfuscating original data for privacy, comprising:

• at least one memory that stores instructions;
• and
• at least one processor configured to execute the instructions to perform:
• processes of (I) inputting training data into the obfuscation network, to thereby instruct the obfuscation network to filter frequency information of the training data and thus generate obfuscated data;
• and (II) (i) inputting the obfuscated data into a learning network designed to perform a specific task, to thereby instruct the learning network to generate characteristic information by performing learning operation on the obfuscated data, (ii) generating at least one task loss by referring to (ii-1) the characteristic information and its corresponding ground truth or (ii-2) a task-specific output, generated by using the characteristic information, and its corresponding ground truth, and (iii) training at least one of the learning network and the obfuscation network through a backpropagation of the task loss.

A testing device for testing an obfuscation network capable of obfuscating original data for privacy, comprising:

• at least one memory that stores instructions;
• and
• at least one processor configured to execute the instructions to perform:
• processes of (I) on condition that a learning device has trained an obfuscation network to acquire a trained obfuscation network by performing processes of (i) inputting first training data into the obfuscation network, to thereby instruct the obfuscation network to filter frequency information of the first training data and thus generate first obfuscated data, and (ii) (ii-1) inputting the first obfuscated data into a first learning network designed to perform a first task, to thereby instruct the first learning network to generate first characteristic information by performing learning operation on the first obfuscated data, (ii-2) generating a first task loss by referring to the first characteristic information and its corresponding ground truth or a first task-specific output, generated by using the first characteristic information, and its corresponding ground truth, and (ii-3) training at least one of the first learning network and the obfuscation network through a backpropagation of the first task loss, inputting second training data into the trained obfuscation network, to thereby instruct the trained obfuscation network to filter frequency information of the second training data and thus generate second obfuscated data;
• and (II) (i) inputting the second obfuscated data into a second learning network designed to perform a second specific task, to thereby instruct the second learning network to generate second characteristic information by performing the learning operation on the second obfuscated data, (ii) generating a second task loss by referring to (ii-1) the second characteristic information and its corresponding ground truth or (ii-2) a second task-specific output, generated by using the second characteristic information, and its corresponding ground truth, and (iii) training the second learning network through a backpropagation of the second task loss.