US 12,254,120 B2
Machine learning model fingerprinting
David Beveridge, Tillamook, OR (US); and Andrew Davis, Portland, OR (US)
Assigned to HiddenLayer, Inc., Austin, TX (US)
Filed by HiddenLayer, Inc., Austin, TX (US)
Filed on Sep. 20, 2023, as Appl. No. 18/471,163.
Application 18/471,163 is a continuation of application No. 18/327,755, filed on Jun. 1, 2023, granted, now 11,921,903.
Prior Publication US 2024/0403492 A1, Dec. 5, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 21/64 (2013.01); G06F 21/00 (2013.01); G06F 21/57 (2013.01); G06F 21/62 (2013.01)
CPC G06F 21/64 (2013.01) [G06F 21/577 (2013.01); G06F 21/629 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving data characterizing artefacts associated with each of a plurality of layers of a first machine learning model;
generating, for each layer of the first machine learning model, a plurality of fingerprints corresponding to the artefacts in such layer, the generated fingerprints for all of the layers of the first machine learning model collectively forming a model indicator for the first machine learning model;
determining whether the first machine learning model is derived from another machine learning model by performing a similarity analysis between one or more value distributions of the model indicator for the first machine learning model and one or more value distributions of model indicators generated for each of a plurality of reference machine learning models each of which comprise a respective set of fingerprints, the similarity analysis identifying one of the plurality of reference machine learning models having a closet set of fingerprints; and
providing data characterizing the determining to a consuming application or process.