| CPC G06F 21/566 (2013.01) [G06F 21/554 (2013.01); G06V 10/764 (2022.01); G06V 10/776 (2022.01); G06V 20/49 (2022.01); G06V 20/52 (2022.01); G06F 2221/034 (2013.01)] | 12 Claims |

|
1. A method, comprising:
dynamically monitoring a stream of image portions that have been classified by a segmentation model of a video surveillance system;
evaluating the image portions;
based on the evaluating, determining that an attack on the video surveillance system is occurring, or has occurred; and
implementing, or causing the implementation of, a remedial action with regard to the attack,
wherein the evaluating comprises classifying each of the image portions as either ‘noise’ or ‘conformant’,
wherein the image portions are evaluated on a rolling time window basis,
the rolling time window basis comprising:
1. splitting a time window into a plurality of time frames;
2. creating a linked list from the plurality of time frames, where a head of the linked list corresponds to a current time frame of the plurality of time frames, and the tail of the linked list is an oldest time frame of the plurality of time frames;
3. predicting a probability of each image portion in the time frame as either a ‘noise’ image or a ‘conformant’ image using a discriminator; and
4. for each time frame within the linked list, storing a respective counter of ‘conformant’ image classifications and another counter of ‘noise’ image classifications based on the predicted probability.
|