Technical Specification: Neural Network Traceability (MPAI-NNT) V1.0, in the following also called MPAI-NNT V1.0 or MPAI-NNT, specifies methods to evaluate the following aspects of Active (Watermarking) and Passive (Fingerprinting) Neural Network Traceability Methods in terms of:

  • Ability of a Neural Network Traceability Detector/Decoder to detect/decode/match Traceability Data when the traced Neural Network has been modified.
  • Computational cost of injecting, extracting, detecting, decoding, or matching Traceability Data.
  • Specifically for active tracing methods, impact on the performance of a neural network with inserted Traceability Data and its inference.

The standard assumes that:

  • The Neural Network Traceability Method to be evaluated according to this standard is known to the Tester.
  • The performance of the Neural Network Traceability Method does not depend on specific Secret Keys.

This Technical Specification has been developed by the MPAI Neural Network Watermarking Development Committee (NNW-DC). As the Neural Network Traceability area is fast-evolving, MPAI expects it will produce future MPAI-NNT versions providing methods to cope with technology evolution.