US 11,914,373 B2
Artificial neural network integrity verification
Alberto Troia, Munich (DE); Antonino Mondello, Messina (IT); and Michelangelo Pisasale, Catania (IT)
Assigned to Micron Technology, Inc., Boise, ID (US)
Filed by Micron Technology, Inc., Boise, ID (US)
Filed on Sep. 26, 2022, as Appl. No. 17/953,266.
Application 17/953,266 is a continuation of application No. 16/229,044, filed on Dec. 21, 2018, granted, now 11,454,968.
Claims priority of provisional application 62/636,214, filed on Feb. 28, 2018.
Prior Publication US 2023/0021289 A1, Jan. 19, 2023
Int. Cl. G06N 3/063 (2023.01); G06F 17/16 (2006.01); G05D 1/00 (2006.01); H04L 9/32 (2006.01); G06F 11/10 (2006.01); G06F 21/00 (2013.01)
CPC G05D 1/0088 (2013.01) [G06F 11/1012 (2013.01); G06F 17/16 (2013.01); G06F 21/00 (2013.01); G06N 3/063 (2013.01); H04L 9/3239 (2013.01); H04L 9/3242 (2013.01)] 18 Claims
OG exemplary drawing
 
1. An apparatus, comprising:
a memory storing weight matrix data and bias data corresponding to a number of artificial neural network (ANN) partitions; and
responsive to receiving a number of inputs to an artificial intelligence (AI) controller of a system comprising the memory and the ANN:
determine a particular ANN partition to which the number of inputs correspond;
read, from the memory:
the weight matrix data and bias data corresponding to the particular ANN partition; and
a first cryptographic code corresponding to the particular ANN partition;
generate, via a safety controller, using the weight matrix data and bias data read from the memory, a second cryptographic code corresponding to the particular ANN partition; and
perform, via the safety controller, a comparison of the first cryptographic code and the second cryptographic code;
upon the comparison resulting in a determined mismatch between the first cryptographic code and the second cryptographic code, issue, from the safety controller, an indication of the mismatch to a primary controller of the system; and
upon the comparison resulting in a determined match between the first cryptographic code and the second cryptographic code, determine an output by evaluating, via the AI controller, the number of inputs to the system using the weight matrix and bias data read from the memory; and
wherein the safety controller is configured to issue the indication of the mismatch to the primary controller responsive to the determined mismatch.