US 12,592,961 B2Granted

Quantum-based adaptive deep learning framework for securing network files

AssigneeBank of America Corporation

InventorsUsha Rani Alluru, Saurabh Gupta, Swathi Bussa, Shailendra Singh

Life of this patent

Estimating the term…

Loading patent…

Abstract

Apparatus and methods for quantum-computing based file remediation are provided. A quantum file remediation program on a computer system with a standard processor and an “N”-qubit processor may receive a network file. When a file attribute score determined on the standard processor deviates by more than a predetermined amount from a baseline score, the program may initialize a quantum circuit. A deep learning framework using a quantum generative adversarial network (QGAN) may run on the quantum circuit and generate a remediated file. When the QGAN session ends, the quantum circuit may be collapsed and the original network file may be replaced by the remediated file.

Claims21

Dependent claims

A quantum file remediation computer program product, the computer program product comprising a non-transitory computer-readable medium storing instructions that, when executed by a computer system that comprises a standard processor and an n-qubit processor, cause the computer system to:

receive an original file, the file comprising a set of attributes;
determine, on the standard processor, when a file attribute score deviates by more than a predetermined amount from a first baseline score, the file attribute score based at least in part on the set of attributes;
and
when the file attribute score deviates by more than predetermined amount from the first baseline score:
initialize a quantum circuit with the n-qubit processor;
and
operate a quantum generative adversarial network (QGAN) on the quantum circuit, the QGAN comprising a generator neural network and a discriminator neural network;
detect a security threat associated with the file;
at the generator neural network, synthesize a version of a replacement file comprising a remediation associated with the security threat;
at the discriminator neural network, determine when the version of the replacement file is associated with a file attribute score that deviates by more than a predetermined amount from a second baseline score, the second baseline score dynamically determined based at least in part on a set of reference files;
at the generator neural network, in response to output from the discriminator neural network, synthesize a new version of the replacement file, based at least in part on output from the discriminator neural network;
and
when the QGAN reaches equilibrium:
collapse the quantum circuit;
and
replace the original file with a version of the replacement file;
wherein:
n is a number between two and one thousand;
and
the QGAN utilizes a superposition property of the n-qubit processor.

17 more elements

A method for quantum-based file remediation, the method comprising:

receiving, at a quantum file remediation computer program running on a computer system comprising a standard processor and an “N”-qubit processor, a first file comprising a set of attributes;
determining, on the standard processor, when a file attribute score deviates by more than a predetermined amount from a first baseline score, the file attribute score based at least in part on the set of attributes;
detecting a security threat associated with the first file based on the file attribute score;
and
when the file attribute score deviates by more than a predetermined amount from the first baseline score:
initializing a quantum circuit with the n-qubit processor;
operating a quantum generative adversarial network (QGAN) session on the quantum circuit, the QGAN comprising a generative network and a discriminator network;
at the generator network, synthesizing a second file, the second file comprising a version of the first file, the second file remediating the security threat;
at the discriminator network, determining when the second file is associated with a quantum file attribute score that deviates by more than a predetermined amount from a second baseline score, the second baseline score dynamically determined based at least in part on a set of training files;
and
when the QGAN session ends:
collapsing the quantum circuit;
and
replacing the first file with the second file;
wherein:
n is a number between two and one thousand;
and
the QGAN utilizes a superposition property of the n-qubit processor.

15 more elements

Dependencies are inferred from the claim text.

Facts

Application: 18/663,351Filed 14 May 2024
Published: US 2025/0358311 A120 Nov 2025
Granted: 31 Mar 2026
Term ends: Not estimatedThe source records don’t support an estimate.

Classification10 codes

H04L 63/145
Countermeasures against malicious traffic: the attack involving the propagation of malware through the network, e.g. viruses, trojans or worms
G06N 3/045
Architecture, e.g. interconnection topology: combinations of networks
G06N 3/047
Architecture, e.g. interconnection topology: probabilistic or stochastic networks
G06N 3/0475
Architecture, e.g. interconnection topology: generative networks