Method and apparatus for improving the robustness of a machine learning system
AssigneeRobert Bosch GmbH
InventorJan Hendrik Metzen
Life of this patent
Estimating the term…
Abstract
A method for operating a detector that is set up to check whether a data signal that is supplied to a machine learning system has been manipulated. The machine learning system is first trained in adversarial fashion using a manipulated data signal, the manipulated data signal having been ascertained by manipulation of a training data signal, and the machine learning system being trained to provide in each case the same output signal when the training data signal or the manipulated data signal is supplied to it. The detector is trained using another manipulated data signal that is produced as a function of the trained machine learning system.
Claims8
A method for operating a detector that is set up to check whether a data signal that is supplied to a machine learning system has been manipulated, the method comprising the following steps:
- adversarially training the machine learning system using a manipulated data signal, the manipulated data signal having been ascertained by manipulation of a training data signal, the machine learning system being trained to provide in each case the same output signal when the training data signal or the manipulated data signal is supplied to the machine learning system;
- and
- training the detector, wherein the detector is separate from the machine learning system and is configured to output a signal indicating a probability that the data signal supplied to the machine learning system has been manipulated, the training the detector using an additional manipulated data signal that is produced as a function of the trained machine learning system.
A non-transitory machine-readable storage medium on which is stored a computer program for operating a detector that is set up to check whether a data signal that is supplied to a machine learning system has been manipulated, the computer program, when executed by a computer, causing the computer to perform the following steps:
- adversarially training the machine learning system using a manipulated data signal, the manipulated data signal having been ascertained by manipulation of a training data signal, the machine learning system being trained to provide in each case the same output signal when the training data signal or the manipulated data signal is supplied to it;
- and
- training the detector, wherein the detector is separate from the machine learning system and is configured to output a signal indicating a probability that the data signal supplied to the machine learning system has been manipulated, the training the detector using an additional manipulated data signal that is produced as a function of the trained machine learning system.
A computer configured to operate a detector that is set up to check whether a data signal that is supplied to a machine learning system has been manipulated, the computer configured to:
- adversarially train the machine learning system using a manipulated data signal, the manipulated data signal having been ascertained by manipulation of a training data signal, the machine learning system being trained to provide in each case the same output signal when the training data signal or the manipulated data signal is supplied to the machine learning system;
- and
- train the detector, wherein the detector is separate from the machine learning system and is configured to output a signal indicating a probability that the data signal supplied to the machine learning system has been manipulated, the training the detector using an additional manipulated data signal that is produced as a function of the trained machine learning system.
Dependencies are inferred from the claim text.