Verification and identification of a neural network
The application relates to a method for verifying characteristic features of a neural network, comprising obtaining the neural network as well as an identifier assigned to the neural network, determining the characteristic features of the neural network, calculating a first hash code using a predetermined hash function from the characteristic features of the neural network, obtaining a second hash code assigned to the identifier from a secure database, as well as verifying the neural network by comparing the first hash code to the second hash code. The application furthermore comprises a computer software product which can be downloaded to the internal memory of a digital computer and which comprises software code sections with which the steps according to the method described here are carried out when the software is executed on a computer.
1. A method for verifying characteristic features of a trained neural network, the method comprising:
obtaining the trained neural network as well as an identifier assigned to the trained neural network;
determining one or more characteristic features of the trained neural network;
calculating a first hash code using a predetermined hash function from the one or more characteristic features of the trained neural network;
obtaining a second hash code assigned to the identifier from a secure database; and
verifying or rejecting the trained neural network by comparing the first hash code to the second hash code, wherein the one or more characteristic features of the trained neural network comprise one or more details on a training method used for training the trained neural network.
2. The method according to claim 1 , wherein the trained neural network is obtained by downloading from the Internet, from an Internet-based distribution platform, or as an attachment to an email.
3. The method according to claim 1 , wherein the secure database is part of a blockchain network.
4. The method according to claim 1 , wherein the one or more characteristic features of the trained neural network further include:
an architecture and a weight of the trained neural network.
5. The method according to claim 4 , wherein the one or more characteristic features further include:
an identifier of a provider of the trained neural network.
6. The method according to claim 1 , wherein the predetermined hash function includes a collision resistant hash function.
7. The method according to claim 1 , wherein the method, before obtaining the trained neural network and the assigned identifier from the secure database, further includes:
uploading the trained neural network to a platform assigned to the secure database;
assigning the identifier to the trained neural network;
determining the one or more characteristic features of the trained neural network;
calculating the second hash code using the predetermined hash function from the one or more characteristic features of the trained neural network; and
storing the calculated second hash code of the trained neural network together with the identifier of the trained neural network in the secure database;
wherein the one or more characteristic features of the trained neural network which are stored or mapped in the calculated second hash code are displayed during a download of the trained neural network from the secure database or the platform.
8. A non-transitory computer-readable medium that, when executed by a processor of a computer, cause the processor to execute operations for verifying characteristic features of a trained neural network, the operations comprising:
obtaining the trained neural network;
obtaining an identifier assigned to the trained neural network;
determining one or more characteristic features of the trained neural network;
calculating a first hash code using a predetermined hash function from the one or more characteristic features;
obtaining a second hash code assigned to the identifier from a secure database; and
verifying or rejecting the trained neural network by comparing the first hash code to the second hash code, wherein the one or more characteristic features of the trained neural network include one or more details on a training method used for training the trained neural network.
9. The non-transitory computer-readable medium of claim 8 , wherein the secure database is part of a blockchain network.
10. The non-transitory computer-readable medium of claim 8 , wherein the one or more characteristic features further include at least one of an architecture of the trained neural network, a provider of the trained neural network, or a weight of the trained neural network.
11. The non-transitory computer-readable medium of claim 8 , wherein the predetermined hash function includes a collision resistant hash function.
12. The non-transitory computer-readable medium of claim 8 , wherein the trained neural network is obtained by downloading the trained neural network from an internet connected source, from an internet-based distribution platform, or as an attachment to an email.
13. The non-transitory computer-readable medium of claim 8 , the operations further comprising:
uploading the trained neural network to a platform assigned to the secure database;
assigning the identifier to the trained neural network;
determining the one or more characteristic features of the trained neural network;
calculating the second hash code using the predetermined hash function from the one or more characteristic features of the trained neural network; and
storing the calculated second hash code together with the identifier of the trained neural network in the secure database.
14. The non-transitory computer-readable medium of claim 13 , wherein the one or more characteristic features of the trained neural network are stored or mapped in the calculated second hash code and displayed during a download of the trained neural network from the secure database or the platform.
15. A method for verifying characteristic features of a neural network, the method comprising:
uploading a trained neural network to a platform assigned to a secure database;
assigning an identifier to the trained neural network;
determining one or more characteristic features of the trained neural network;
calculating a second hash code using a predetermined hash function from the one or more characteristic features of the trained neural network;
storing the calculated second hash code together with the identifier of the trained neural network in a secure database;
obtaining the trained neural network;
obtaining the identifier assigned to the trained neural network;
calculating a first hash code using the predetermined hash function from the one or more characteristic features of the obtained trained neural network; and
verifying or rejecting the trained neural network by comparing the calculated first hash code to the second hash code, when the one or more characteristic features of the trained neural network include one or more details on a training method used for training the trained neural network.
16. The method of claim 15 , wherein the trained neural network is obtained by downloading from at least one of an internet website, an internet-based distribution platform, or an email attachment.
17. The method of claim 15 , wherein the secure database is part of a blockchain network.
18. The method of claim 15 , wherein the one or more characteristic features further include an architecture of the trained neural network.
19. The method of claim 15 , therein the one or more characteristic features further include a weight of the trained neural network.
20. The method of claim 15 , wherein the predetermined hash function includes a collision resistant hash function.