	US 12,411,969 B2
	On screen data protection using camera and narrow artificial intelligence
Abhay Kumar, Haryana (IN); Navin R Poojari, Mumbai (IN); Kapil Sudhir Karkhanis, Mumbai (IN); and Syed Luqman Ahmed, New Delhi (IN)
Assigned to Bank of America Corporation, Charlotte, NC (US)
Filed by Bank of America Corporation, Charlotte, NC (US)
Filed on Aug. 29, 2023, as Appl. No. 18/239,261.
Prior Publication US 2025/0077693 A1, Mar. 6, 2025
Int. Cl. G06F 21/62 (2013.01)

CPC G06F 21/6218 (2013.01)

20 Claims

1. A method for using narrow artificial intelligence (“AI”) to detect and secure confidential data, the method comprising:

using a camera on a user computer screen to capture a live camera feed of an area in front of the user computer screen;

pre-processing the live camera feed by categorizing data captured in the live camera feed into categorized data;

identifying one or more unverified data elements in the categorized data using a machine learning algorithm;

determining a security-breach score for each unverified data element;

labeling each unverified data element with the determined security-breach score;

in response to determining that an unverified data element is labeled with a security-breach score that is greater than a predetermined security-breach score, securing the user computer screen, the securing comprising:

capturing a screenshot of the user computer screen;

extracting data from the screenshot;

transmitting the extracted data to an object identification algorithm, the object identification algorithm comprising:

an input layer for determining data values corresponding to the extracted data from the screenshot;

a pattern layer for determining a pattern of the data values;

a summation layer for determining an output signal for the pattern of data values; and

an output layer for outputting the output signal to a data classification model, the data classification model comprising:

identifying pixel values corresponding to each of the data values;

segmenting the data values included in the outputted signal into segmented data values based on identified pixel values for each data value;

analyzing the segmented data values to identify data specific features for each data value segment; and

labeling the data value segments according to the identified data specific features;

recreating the screenshot using a narrow AI model by:

inputting the labeled data value segments;

converting the data value segments into pixels; and

combining the pixels to recreate the screenshot;

identifying confidential data within the screenshot using a pattern analysis model;

blurring the confidential data in the recreated screenshot; and

overwriting the screenshot of the data displayed on the user computer screen to display the recreated screenshot.