Approaches to learning behavioral norms through an analysis of digital activities performed across different services and using the same for detecting threats
Introduced here is a network-accessible platform (or simply “platform”) that is designed to monitor digital activities that are performed across different services to ascertain, in real time, threats to the security of an enterprise. In order to surface insights into the threats posed to an enterprise, the platform can apply machine learning models to data that is representative of digital activities performed on different services with respective accounts. Each model may be trained to understand what constitutes normal behavior for a corresponding employee with respect to a single service or multiple services. Not only can these models be autonomously trained for the employees of the enterprise, but they can also be autonomously applied to detect, characterize, and catalog those digital activities that are indicative of a threat.
1 . A method performed by a threat detection platform for learning what constitutes normal behavior for an employee of an enterprise across multiple Software-as-a-Service (SaaS) services, the method comprising:
acquiring, via a first application programming interface (API), first data related to past digital activities performed with a first account that is associated with the employee on a first SaaS service;
segmenting the first data into a first pair of segments,
wherein the first pair of segments includes:
(i) a first segment that includes a portion of the past digital activities performed with the first account on the first SaaS service, and
(ii) a second segment that includes another portion of the past digital activities performed with the first account on the first SaaS service;
acquiring, via a second API, second data related to past digital activities performed with a second account that is associated with the employee on a second SaaS service;
segmenting the second data into a second pair of segments,
wherein the second pair of segments includes:
(i) a third segment that includes a portion of the past digital activities performed with the second account on the second SaaS service, and
(ii) a fourth segment that includes another portion of the past digital activities performed with the second account on the second SaaS service;
providing the first and third segments to a machine learning algorithm that produces, as output, a machine learning model that is able to predict a likelihood of the first and second accounts being compromised based on an analysis of digital activities performed with the first and second accounts on the first and second SaaS services;
validating the machine learning model with the second and fourth segments;
in response to a determination that performance of the machine learning model exceeds a threshold during said validating,
storing the machine learning model in a data structure that is associated with the employee or the enterprise;
acquiring, via the first API, third data that relates to one or more digital activities performed with the first account on the first SaaS service;
acquiring, via the second API, fourth data that relates to one or more digital activities performed with the second account on the second SaaS service; and
providing the third and fourth data to the machine learning model that produces, as output, a prediction as to whether the first account or the second account is compromised.
2 . The method of claim 1 , wherein the machine learning model is a neural network with weights that are tuned as the machine learning algorithm learns from an analysis of the first segment, the third segment, and the first and third segments together.
3 . The method of claim 1 , further comprising:
determining an appropriate remediation action based on the prediction output by the machine learning model.
4 . The method of claim 1 , wherein the enterprise has multiple employees, each of whom is associated with a different one of multiple machine learning models stored in the data structure.
5 . The method of claim 1 , further comprising:
canonicalizing the first and second data by populating the first data into a first data structure having predetermined fields and populating the second data into a second data structure having a same form as the first data structure.
6 . A non-transitory medium with instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
acquiring, via a first application programming interface (API), first data related to past digital activities performed with a first account that is associated with an employee of an enterprise on a first Software-as-a-Service (Saas) service;
segmenting the first data into a first pair of segments,
wherein the first pair of segments includes:
(i) a first segment that includes a portion of the past digital activities performed with the first account on the first SaaS service, and
(ii) a second segment that includes another portion of the past digital activities performed with the first account on the first SaaS service;
acquiring, via a second API, second data related to past digital activities performed with a second account that is associated with the employee on a second SaaS service;
segmenting the second data into a second pair of segments,
wherein the second pair of segments includes:
(i) a third segment that includes a portion of the past digital activities performed with the second account on the second SaaS service, and
(ii) a fourth segment that includes another portion of the past digital activities performed with the second account on the second SaaS service;
providing the first and third segments to a machine learning algorithm that produces, as output, a machine learning model that is able to predict a likelihood of the first and second accounts being compromised based on an analysis of digital activities performed with the first and second accounts on the first and second SaaS services;
validating the machine learning model with the second and fourth segments;
in response to a determination that performance of the machine learning model exceeds a threshold during said validating,
storing the machine learning model in a data structure that is associated with the employee or the enterprise;
acquiring, via the first API, third data that relates to one or more digital activities performed with the first account on the first SaaS service;
acquiring, via the second API, fourth data that relates to one or more digital activities performed with the second account on the second SaaS service; and
providing the third and fourth data to the machine learning model that produces, as output, a prediction as to whether the first account or the second account is compromised.
7 . The non-transitory medium of claim 6 , wherein the machine learning model is a neural network with weights that are tuned as the machine learning algorithm learns from an analysis of the first segment, the third segment, and the first and third segments together.
8 . The non-transitory medium of claim 6 , wherein the operations further comprise:
determining an appropriate remediation action based on the prediction output by the machine learning model.
9 . The non-transitory medium of claim 6 , wherein the enterprise has multiple employees, each of whom is associated with a different one of multiple machine learning models stored in the data structure.
10 . The non-transitory medium of claim 6 , further comprising:
canonicalizing the first and second data by populating the first data into a first data structure having predetermined fields and populating the second data into a second data structure having a same form as the first data structure.
11 . A system comprising:
one or more processors; and
one or more non-transitory media with instructions stored thereon that, when executed by the one or more processors, cause the system to:
acquire, via a first application programming interface (API), first data related to past digital activities performed with a first account that is associated with an employee of an enterprise on a first Software-as-a-Service (SaaS) service;
segment the first data into a first pair of segments,
wherein the first pair of segments includes:
(i) a first segment that includes a portion of the past digital activities performed with the first account on the first SaaS service, and
(ii) a second segment that includes another portion of the past digital activities performed with the first account on the first SaaS service;
acquire, via a second API, second data related to past digital activities performed with a second account that is associated with the employee on a second SaaS service;
segment the second data into a second pair of segments,
wherein the second pair of segments includes:
(i) a third segment that includes a portion of the past digital activities performed with the second account on the second SaaS service, and
(ii) a fourth segment that includes another portion of the past digital activities performed with the second account on the second SaaS service;
provide the first and third segments to a machine learning algorithm that produces, as output, a machine learning model that is able to predict a likelihood of the first and second accounts being compromised based on an analysis of digital activities performed with the first and second accounts on the first and second SaaS services;
validate the machine learning model with the second and fourth segments;
in response to a determination that performance of the machine learning model exceeds a threshold,
store the machine learning model in a data structure that is associated with the employee or the enterprise;
acquire, via the first API, third data that relates to one or more digital activities performed with the first account on the first SaaS service;
acquire, via the second API, fourth data that relates to one or more digital activities performed with the second account on the second SaaS service; and
provide the third and fourth data to the machine learning model that produces, as output, a prediction as to whether the first account or the second account is compromised.
12 . The system of claim 11 , wherein the machine learning model is a neural network with weights that are tuned as the machine learning algorithm learns from an analysis of the first segment, the third segment, and the first and third segments together.
13 . The system of claim 11 , wherein execution of the instructions further causes the system to:
determine an appropriate remediation action based on the prediction output by the machine learning model.
14 . The system of claim 11 , wherein the enterprise has multiple employees, each of whom is associated with a different one of multiple machine learning models stored in the data structure.
15 . The system of claim 11 , wherein execution of the instructions further causes the system to:
canonicalize the first and second data by populating the first data into a first data structure having predetermined fields and populating the second data into a second data structure having a same form as the first data structure.