Device anti-surveillance system
A method comprises receiving one or more inputs captured by a camera of a device, and determining, using one or more machine learning models, whether the one or more inputs depict at least one object configured to capture a visual representation of a screen of the device. A recommendation is generated responsive to an affirmative determination, the recommendation comprising at least one action to prevent the capture of the visual representation of the screen of the device.
1. A method comprising:
receiving one or more inputs captured by a camera of a device;
determining, using one or more machine learning models, whether the one or more inputs depict at least one object configured to capture a visual representation of a screen of the device; and
generating a recommendation comprising at least one action to prevent the capture of the visual representation of the screen of the device responsive to an affirmative determination;
wherein determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device comprises determining an angle of the at least one object with respect to the screen, and using historical data to determine whether the angle orients the at least one object in a position to enable the capture of the visual representation of the screen; and
wherein the steps of the method are executed by a processing device operatively coupled to a memory.
2. The method claim 1 , wherein determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device further comprises:
generating one or more bounding boxes in the one or more inputs, wherein the one or more bounding boxes are formed around objects in the one or more inputs potentially configured to capture the visual representation of the screen of the device; and
classifying the objects in respective ones of the one or more bounding boxes;
wherein the generating and the classifying are performed using the one or more machine learning models.
3. The method claim 2 , wherein the one or more machine learning models comprises a convolutional neural network (CNN).
4. The method of claim 1 , wherein determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device further comprises:
determining positions of the at least one object at different times; and
generating bounding boxes around the at least one object at the different times.
5. The method of claim 4 , wherein determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device further comprises plotting the determined positions of the at least one object on a confidence map.
6. The method of claim 4 , wherein the one or more inputs comprise one or more video streams including the at least one object.
7. The method of claim 1 , wherein determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device further comprises computing a distance of the at least one object from at least one of a given point and a given line.
8. The method of claim 1 , wherein the one or more inputs comprise at least one of one or more video streams and one or more still images including the at least one object.
9. The method of claim 1 , wherein the at least one action to prevent the capture of the visual representation of the screen comprises concealing a display on the screen.
10. The method of claim 9 , wherein concealing the display on the screen comprises at least one of blurring the display, covering the display with an opaque image, replacing the display with another display and replacing the display with a lock screen.
11. The method of claim 1 , wherein the at least one action to prevent the capture of the visual representation of the screen comprises closing one or more applications on the device.
12. The method of claim 1 , wherein the at least one action to prevent the capture of the visual representation of the screen comprises transmitting an alert based at least in part on the affirmative determination to one or more users.
13. The method of claim 1 , further comprising transmitting the recommendation comprising the at least one action to the device, wherein the recommendation is generated and transmitted to the device as a real-time response to the affirmative determination.
14. The method of claim 1 , further comprising maintaining a registry of one or more applications of the device, wherein opening of the one or more applications automatically triggers the camera of the device to capture the one or more inputs.
15. The method of claim 1 , wherein the at least one object comprises a camera separate from the device.
16. An apparatus comprising:
a processing device operatively coupled to a memory and configured to:
receive one or more inputs captured by a camera of a device;
determine, using one or more machine learning models, whether the one or more inputs depict at least one object configured to capture a visual representation of a screen of the device;
generate a recommendation comprising at least one action to prevent the capture of the visual representation of the screen of the device responsive to an affirmative determinations;
wherein, in determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device, the processing device is further configured to determine an angle of the at least one object with respect to the screen, and use historical data to determine whether the angle orients the at least one object in a position to enable the capture of the visual representation of the screen.
17. The apparatus of claim 16 , wherein, in determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device, the processing device is further configured to:
generate one or more bounding boxes in the one or more inputs, wherein the one or more bounding boxes are formed around objects in the one or more inputs potentially configured to capture the visual representation of the screen of the device; and
classify the objects in respective ones of the one or more bounding boxes;
wherein the generating and the classifying are performed using the one or more machine learning models.
18. The apparatus of claim 16 , wherein, in determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device, the processing device is further configured to:
determine positions of the at least one object at different times; and
generate bounding boxes around the at least one object at the different times.
19. An article of manufacture comprising a non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes said at least one processing device to perform the steps of:
receiving one or more inputs captured by a camera of a device;
determining, using one or more machine learning models, whether the one or more inputs depict at least one object configured to capture a visual representation of a screen of the device;
generating a recommendation comprising at least one action to prevent the capture of the visual representation of the screen of the device responsive to an affirmative determination;
wherein, in determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device, the program code further causes said at least one processing device to perform the steps of determining an angle of the at least one object with respect to the screen, and using historical data to determine whether the angle orients the at least one object in a position to enable the capture of the visual representation of the screen.
20. The article of manufacture of claim 19 , wherein, in determining whether the one or more inputs depict at least one object configured to capture the visual representation of the screen of the device, the program code further causes said at least one processing device to perform the steps of:
generating one or more bounding boxes in the one or more inputs, wherein the one or more bounding boxes are formed around objects in the one or more inputs potentially configured to capture the visual representation of the screen of the device; and
classifying the objects in respective ones of the one or more bounding boxes;
wherein the generating and the classifying are performed using the one or more machine learning models.