Adding a fingerprint to a machine learning model
Various embodiments relate to a method of producing a machine learning model with a fingerprint that maps an input value to an output label, including: selecting a set of extra input values, wherein the set of extra input values does not intersect with a set of training labeled input values for the machine learning model; selecting a first set of artificially encoded output label values corresponding to each of the extra input values in the set of extra input values, wherein the first set of artificially encoded output label values are selected to indicate the fingerprint of a first machine learning model; and training the machine learning model using a combination of the extra input values with associated first set of artificially encoded output values and the set of training labeled input values to produce the first learning model with the fingerprint.
1. A method of producing a first machine learning model and a second machine learning model that maps an input value to an output label, comprising:
selecting a set of common training labeled input values, a first set of extra input values and a second set of extra input values from a set of valid input values for the first machine learning model and the second machine learning model, wherein the set of common training labeled input values, first set of extra input values, and second set of extra input values do not intersect with one another such each valid input value is in at most one of the set of common training labeled input values, first set of extra input values, and second set of extra input values;
selecting a first set of artificially encoded output label values corresponding to each of the extra input values in the first set of extra input values, wherein the first set of artificially encoded output label values are selected to indicate a first fingerprint of the first machine learning model;
training the first machine learning model using a combination of the first extra input values with associated first set of artificially encoded output values and the set of common labeled training input values to produce the first learning model with the first fingerprint;
selecting a second set of artificially encoded output label values corresponding to each of the extra input values in the second set of extra input values, wherein the second set of artificially encoded output label values are selected to indicate a second fingerprint of a second machine learning model; and
training the second machine learning model using a combination of the second extra input values with associated second set of artificially encoded output values and the set of common labeled training input values to produce the second learning model with the second fingerprint,
wherein the first machine learning model and the second machine learning produce the same output for the set of common labeled training inputs and the first machine learning model and the second machine learning model have different fingerprints.
2. The method of claim 1 , wherein the fingerprint of the first machine learning model is based upon specific bits from each of the artificially encoded output label values in the first set of artificially encoded output label values.
3. The method of claim 2 , wherein the specific bits are one of a specific number of least significant bits, a specific number of most significant bits, or a specific number of contiguous bits.
4. The method of claim 2 , wherein the fingerprint includes error correction bits.
5. The method of claim 1 , wherein each member of the first set of extra input values to the first machine learning model corresponds to the first fingerprint and the first machine learning model produces a predetermined output indicating the first fingerprint only when members of the first set of extra inputs are input into the first machine learning model.
6. A non-transitory machine-readable storage medium encoded with instructions for producing a first machine learning model and a second machine learning model that maps an input value to an output label:
instructions for selecting a set of common training labeled input values, a first set of extra input values, and a second set of extra input values from a set of valid input values for the first machine learning model and the second machine learning model, wherein the set of common training labeled input values, first set of extra input values, and second set of extra input values do not intersect with one another such each valid input value is in at most one of the set of common training labeled input values, first set of extra input values, and second set of extra input values;
instructions for selecting a first set of artificially encoded output label values corresponding to each of the extra input values in the first set of extra input values, wherein the first set of artificially encoded output label values are selected to indicate a first fingerprint of the first machine learning model;
instructions for training the first machine learning model using a combination of the first extra input values with associated first set of artificially encoded output values and the set of common labeled training input values to produce the first learning model with the first fingerprint;
instructions for selecting a second set of artificially encoded output label values corresponding to each of the extra input values in the second set of extra input values, wherein the second set of artificially encoded output label values are selected to indicate a first second fingerprint of a second machine learning model; and
instructions for training the second machine learning model using a combination of the second extra input values with associated second set of artificially encoded output values and the set of common labeled training input values to produce the second learning model with the second fingerprint,
wherein the first machine learning model and the second machine learning produce the same output for the set of common labeled training inputs and the first machine learning model and the second machine learning model have different fingerprints.
7. The non-transitory machine-readable storage medium of claim 6 , wherein the fingerprint of the first machine learning model is based upon specific bits from each of the artificially encoded output label values in the first set of artificially encoded output label values.
8. The non-transitory machine-readable storage medium of claim 7 , wherein the specific bits are one of a specific number of least significant bits, a specific number of most significant bits, or a specific number of contiguous bits.
9. The non-transitory machine-readable storage medium of claim 7 , wherein the fingerprint includes error correction bits.
10. The non-transitory machine-readable storage medium of claim 6 , wherein each member of the first set of extra input values to the first machine learning model corresponds to the first fingerprint and the first machine learning model produces a predetermined output indicating the first fingerprint only when members of the first set of extra inputs are input into the first machine learning model.