Risk evaluation method, computer-readable recording medium, and information processing apparatus
AssigneeFujitsu Limited
InventorToshio Endoh
Life of this patent
Estimating the term…
Abstract
A risk evaluation method is disclosed. A machine learning is conducted by using a neural network by inputting training data. A data distance corresponding to a permission level is calculated based on restoration data and the training data, the permission level being among a plurality of layers of the neural network, the restoration data being generated by using at least one weight of a plurality of permission level weights at the permission level, the plurality of permission level weights being among a plurality of weights of synapses at the plurality of layers, the plurality of weights being generated by the machine learning.
Claims15
A risk evaluation method performed by a computer, the method comprising:
- conducting machine learning of a neural network by inputting training data into the neural network, so that the neural network has a plurality of trained weights in a plurality of layers after the machine learning;
- calculating restoration data by using only a part but not all of the trained weights, the restoration data being a restored version of the training data, the part of the trained weights being situated in part but not all of the plurality of layers;
- and calculating, based on a data distance between the restoration data and the training data, a risk that the training data are restored from the part of the trained weights in transfer learning.
A non-transitory computer-readable storage medium having stored therein a risk evaluation program that causes a computer to execute a process comprising:
- conducting machine learning of a neural network by inputting training data into the neural network, so that the neural network has a plurality of trained weights in a plurality of layers after the machine learning;
- calculating restoration data by using only a part but not all of the trained weights, the restoration data being a restored version of the training data, the part of the trained weights being situated in part but not all of the plurality of layers;
- and calculating based on a data distance between the restoration data and the training data, a risk that the training data are restored from the part of the trained weights in transfer learning.
An information processing apparatus comprising:
- a memory;
- and a processor coupled to the memory and the processor configured to:
- conduct machine learning of a neural network by inputting training data into the neural network, so that the neural network has a plurality of trained weights in a plurality of layers after the machine learning;
- calculate restoration data by using only a part but not all of the trained weights, the restoration data being a restored version of the training data, the part of the trained weights being situated in part but not all of the plurality of layers;
- and calculate, based on a data distance between the restoration data and the training data, a risk that the training data are restored from the part of the trained weights in transfer learning.
Dependencies are inferred from the claim text.