US 12,326,920 B2
Embedding concealed meta-data into deep neural networks (DNNs)
Firouzeh Golaghazadeh, Ottawa (CA); and Petar Djukic, Ottawa (CA)
Assigned to Ciena Corporation, Hanover, MD (US)
Filed by Ciena Corporation, Hanover, MD (US)
Filed on Jun. 10, 2024, as Appl. No. 18/738,379.
Application 18/738,379 is a continuation of application No. 17/893,648, filed on Aug. 23, 2022, granted, now 12,056,220.
Prior Publication US 2024/0330417 A1, Oct. 3, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 21/16 (2013.01); G06N 3/10 (2006.01)
CPC G06F 21/16 (2013.01) [G06N 3/10 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A non-transitory computer-readable medium comprising instructions that, when executed, cause one or more processors to perform steps of:
receiving a first trained neural network that includes one or more layers each having weights, at least one layer of the one or more layers having an embedded watermark;
transforming the at least one layer to a transformed domain, wherein the transforming includes sorting the weights in the at least one layer and performing a domain transfer on the sorted weights;
extracting first information from the at least one layer in the transformed domain; and
determining, based on the extracted first information, that the at least one layer has the embedded watermark.