Method and system for identifying emerging threats in real-time
The disclosure relates to a method and system for identifying emerging threats in real-time using Artificial Intelligence (AI) model. The method includes receiving first set of feature vectors created from content; determining first set of dimensions for each of first set of feature vectors; comparing first set of dimensions, for each of first set of feature vectors, with second set of dimensions associated with each of second sets of feature vectors created for historical events; computing degree of proximity of first set of feature vectors relative to each of second sets of feature vectors through proximity analysis; identifying contemporaneous to receiving first set of feature vectors, second set of feature vectors from second sets of feature vectors; classifying event into event category from predefined event categories based on computed degree of proximity and predefined threshold; and identifying event as emerging threat.
1 . A method for identifying emerging threats in real-time, the method comprising:
receiving, by an Artificial Intelligence (AI) model, a first set of feature vectors created from content comprising an event comprising one or more entities;
determining, by the AI model, a first set of dimensions for each of the first set of feature vectors;
comparing the first set of dimensions, for each of the first set of feature vectors, by the AI model, with a second set of dimensions associated with each of a plurality of second sets of feature vectors created for historical events, wherein the second set of dimensions and the plurality of second sets of feature vectors are stored in an associated primary repository, wherein the first set of dimensions and the second set of dimensions comprise at least one of a frequency dimension, a recency dimension, a pattern dimension, and an intensity dimension;
computing, by the AI model, a degree of proximity of the first set of feature vectors relative to each of the plurality of second sets of feature vectors based on the comparison, through a proximity analysis;
identifying, by the AI model, contemporaneous to receiving the first set of feature vectors, a second set of feature vectors from the plurality of second sets of feature vectors, wherein the second set of feature vectors has the highest degree of proximity to the first set of feature vectors;
classifying, by the AI model, the event into an event category from a plurality of predefined event categories based on the degree of proximity of the first set of feature vectors relative to the second set of feature vectors and at least one predefined threshold, wherein the plurality of predefined event categories comprises an existing event, a new event, and a partially existing event; and
identifying, by the AI model, the event as at least one emerging threat, in response to the classifying.
2 . The method of claim 1 , wherein the computing further comprises:
generating one or more clusters of feature vectors based on a similarity analysis performed on the plurality of second sets of feature vectors, wherein each of the one or more clusters comprises a visual word associated with a corresponding centroid;
generating a second histogram representation for each of the one or more clusters based on the corresponding visual word;
generating a first histogram representation for the first set of feature vectors; and
comparing the first histogram representation with each of the second histogram representation generated for the one or more clusters through the proximity analysis.
3 . The method of claim 1 , further comprising:
assigning a weight to each of the first set of feature vectors based on a predefined weight allocation criteria.
4 . The method of claim 3 , wherein the computing further comprises determining the degree of proximity based on the weight assigned to each of the first set of feature vectors.
5 . The method of claim 1 , wherein classifying the event further comprises comparing the degree of proximity of the first set of feature vectors relative to the second set of feature vectors with the at least one predefined threshold, and wherein the at least one predefined threshold comprises a first predefined threshold and a second predefined threshold.
6 . The method of claim 5 , wherein the event is classified as the existing event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is above or equal to the first predefined threshold, wherein the event is classified as the partially existing event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is below the first predefined threshold and above or equal to the second predefined threshold, and wherein the event is classified as the new event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is below the second predefined threshold.
7 . The method of claim 1 , further comprising:
classifying the event as a new event;
comparing, for the event, each of the first set of dimensions with a corresponding predefined threat threshold; and
identifying the event as the at least one emerging threat when each of the first set of dimensions is above the corresponding predefined threat threshold.
8 . The method of claim 7 , further comprising storing the first set of dimensions and the first set of feature vectors associated with the event in a second repository.
9 . The method of claim 7 , further comprising:
notifying, in real-time, a user about the at least one emerging threat in response to identifying the at least one emerging threat; and
automatically updating the primary repository with the first set of feature vectors and the first set of dimensions associated with the event.
10 . The method of claim 9 , further comprising training the AI model using the updated primary repository, wherein training the AI model further comprises modifying in real-time or near real-time, a set of parameters of the AI model based on the updated primary repository.
11 . The method of claim 1 , wherein:
the frequency dimension corresponds to a frequency of occurrence of the event captured within the content;
the recency dimension corresponds to time-based proximity with a timestamp associated with the content;
the pattern dimension corresponds to a modification in one or more attributes of the one or more entities; and
the intensity dimension corresponds to a magnitude of an impact of event.
12 . The method of claim 11 , wherein the pattern dimension is determined based on motion signatures of the one or more entities during occurrence of the event.
13 . A system for identifying emerging threats in real-time, the system comprising:
a processor; and
a memory communicatively coupled to the processor, wherein the memory stores processor-executable instructions, which, on execution, causes the processor to:
receive a first set of feature vectors created from content comprising an event comprising one or more entities;
determine a first set of dimensions for each of the first set of feature vectors;
compare the first set of dimensions, for each of the first set of feature vectors, with a second set of dimensions associated with each of a plurality of second sets of feature vectors created for historical events, wherein the second set of dimensions and the plurality of second sets of feature vectors are stored in an associated primary repository, wherein the first set of dimensions and the second set of dimensions comprise at least one of a frequency dimension, a recency dimension, a pattern dimension, and an intensity dimension;
compute a degree of proximity of the first set of feature vectors relative to each of the plurality of second sets of feature vectors based on the comparison, through a proximity analysis;
identify, contemporaneous to receiving the first set of feature vectors, a second set of feature vectors from the plurality of second sets of feature vectors, wherein the second set of feature vectors has the highest degree of proximity to the first set of feature vectors;
classify the event into an event category from a plurality of predefined event categories based on the degree of proximity of the first set of feature vectors relative to the second set of feature vectors and at least one predefined threshold, wherein the plurality of predefined event categories comprises an existing event, a new event, and a partially existing event; and
identify the event as at least one emerging threat, in response to the classifying.
14 . The system of claim 13 , wherein the processor-executable instructions further cause the processor to:
generate one or more clusters of feature vectors based on a similarity analysis performed on the plurality of second sets of feature vectors, wherein each of the one or more clusters comprises a visual word associated with a corresponding centroid;
generate a second histogram representation for each of the one or more clusters based on the corresponding visual word;
generate a first histogram representation for the first set of feature vectors; and
compare the first histogram representation with each of the second histogram representation generated for the one or more clusters through the proximity analysis.
15 . The system of claim 13 , wherein the processor-executable instructions further cause the processor to:
assign a weight to each of the first set of feature vectors based on a predefined weight allocation criteria.
16 . The system of claim 15 , wherein the processor-executable instructions further cause the processor to compute the degree of proximity by determining the degree of proximity based on the weight assigned to each of the first set of feature vectors.
17 . The system of claim 13 , wherein the processor-executable instructions further cause the processor to classify the event by comparing the degree of proximity of the first set of feature vectors relative to the second set of feature vectors with the at least one predefined threshold, and wherein the at least one predefined threshold comprises a first predefined threshold and a second predefined threshold.
18 . The system of claim 17 , wherein the event is classified as the existing event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is above or equal to the first predefined threshold, wherein the event is classified as the partially existing event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is below the first predefined threshold and above or equal to the second predefined threshold, and wherein the event is classified as the new event when the degree of proximity of the first set of feature vectors relative to the second set of feature vectors is below the second predefined threshold.
19 . The system of claim 13 , wherein the processor-executable instructions further cause the processor to:
classify the event as a new event;
compare, for the event, each of the first set of dimensions with a corresponding predefined threat threshold; and
identify the event as the at least one emerging threat when each of the first set of dimensions is above the corresponding predefined threat threshold.
20 . The system of claim 19 , wherein the processor-executable instructions further cause the processor to store the first set of dimensions and the first set of feature vectors associated with the event in a second repository.
21 . The system of claim 19 , wherein the processor-executable instructions further cause the processor to:
notify, in real-time, a user about the at least one emerging threat in response to identifying the at least one emerging threat; and
automatically update the primary repository with the first set of feature vectors and the first set of dimensions associated with the event.
22 . The system of claim 21 , wherein the processor-executable instructions further cause the processor to train an Artificial Intelligence (AI) model using the updated primary repository, wherein training the AI model further comprises modifying in real-time or near real-time, a set of parameters of the AI model based on the updated primary repository.
23 . The system of claim 13 , wherein:
the frequency dimension corresponds to a frequency of occurrence of the event captured within the content;
the recency dimension corresponds to time-based proximity with a timestamp associated with the content;
the pattern dimension corresponds to a modification in one or more attributes of the one or more entities; and
the intensity dimension corresponds to a magnitude of an impact of event.
24 . The system of claim 23 , wherein the pattern dimension is determined based on motion signatures of the one or more entities during occurrence of the event.
25 . A non-transitory computer-readable medium storing computer-executable instructions for identifying emerging threats in real-time, the stored instructions, when executed by a processor, cause the processor to perform operations comprising:
receiving a first set of feature vectors created from content comprising an event comprising one or more entities;
determining a first set of dimensions for each of the first set of feature vectors;
comparing the first set of dimensions, for each of the first set of feature vectors, with a second set of dimensions associated with each of a plurality of second sets of feature vectors created for historical events, wherein the second set of dimensions and the plurality of second sets of feature vectors are stored in an associated primary repository, wherein the first set of dimensions and the second set of dimensions comprise at least one of a frequency dimension, a recency dimension, a pattern dimension, and an intensity dimension;
computing a degree of proximity of the first set of feature vectors relative to each of the plurality of second sets of feature vectors based on the comparison, through a proximity analysis;
identifying, contemporaneous to receiving the first set of feature vectors, a second set of feature vectors from the plurality of second sets of feature vectors, wherein the second set of feature vectors has the highest degree of proximity to the first set of feature vectors;
classifying the event into an event category from a plurality of predefined event categories based on the degree of proximity of the first set of feature vectors relative to the second set of feature vectors and at least one predefined threshold, wherein the plurality of predefined event categories comprises an existing event, a new event, and a partially existing event; and
identifying the event as at least one emerging threat, in response to the classifying.