Programmatic discovery, retrieval, and analysis of communications to identify abnormal communication activity
Conventional email filtering services are not suitable for recognizing sophisticated malicious emails, and therefore may allow sophisticated malicious emails to reach inboxes by mistake. Introduced here are threat detection platforms designed to take an integrative approach to detecting security threats. For example, after receiving input indicative of an approval from an individual to access past email received by employees of an enterprise, a threat detection platform can download past emails to build a machine learning (ML) model that understands the norms of communication with internal contacts (e.g., other employees) and/or external contacts (e.g., vendors). By applying the ML model to incoming email, the threat detection platform can identify security threats in real time in a targeted manner.
1. A method comprising:
acquiring an email addressed to an employee of an enterprise;
applying multiple machine learning (ML) models to the email to identify linguistic features that are indicative of content, sentiment, or tone of the email, wherein each ML model is able to detect linguistic features of a different type, such that the multiple ML models are able to collectively surface information regarding a goal of the email;
determining that the email poses a first type of threat included in a plurality of types of potentially posed threats, based at least in part on the linguistic features identified by the multiple ML models; and
causing display of a visualization component that indicates how a determination of the first type of threat was made on an interface, including by indicating a type of the first type of threat, and a plurality of reasons supporting the determination, wherein the plurality of reasons includes, for a specific reason included in the plurality of reasons, a visually displayed metric comparing specific content included in the email with respect to content of other emails that provides a numerical explanation of the completed determination of the first type of threat for the specific email.
2. The method of claim 1 , wherein the visualization component includes the email with digital elements embedded therein to visually highlight the linguistic features.
3. The method of claim 2 , wherein each digital element includes (i) one or more words that are representative of the linguistic feature and (ii) a tag that specifies the type of linguistic feature.
4. The method of claim 2 , wherein each type of linguistic feature is represented by digital elements in a different color.
5. The method of claim 1 , wherein the visualization component includes a graphic within a user interface window that provides the visually displayed metric used in a text description of the reasons supporting the determination of the first type of threat different from an overall numerical threat indicator displayed on the same user interface window.
6. The method of claim 1 , wherein at least one of the multiple ML models is unable to identify any instances of the corresponding type of linguistic feature.
7. The method of claim 1 , wherein the multiple ML models are able to identify financial terminology, credential theft terminology, engagement terminology, request terminology, or any combination thereof.
8. The method of claim 1 , wherein at least one of the multiple ML models is able to identify terminology indicative of attachments or links included in the email.
9. A non-transitory medium with instructions stored thereon that, when executed by a processor of a computing device, cause the computing device to perform operations comprising:
acquiring an email addressed to an employee of an enterprise;
applying multiple machine learning (ML) models to the email to identify linguistic features that are indicative of content, sentiment, or tone of the email, wherein each ML model is able to detect linguistic features of a different type, such that the multiple ML models are able to collectively surface information regarding a goal of the email;
determining that the email poses a first type of threat included in a plurality of types of potentially posed threats, based at least in part on the linguistic features identified by the multiple ML models; and
causing display of a visualization component that indicates how a determination of the first type of threat was made on an interface, including by indicating a type of the first type of threat, and a plurality of reasons supporting the determination, wherein the plurality of reasons includes, for a specific reason included in the plurality of reasons, a visually displayed metric comparing specific content included in the email with respect to content of other emails that provides a numerical explanation of the completed determination of the first type of threat for the specific email.
10. The non-transitory medium of claim 9 , wherein the visualization component includes the email with digital elements embedded therein to visually highlight the linguistic features.
11. The non-transitory medium of claim 10 , wherein each digital element includes (i) one or more words that are representative of the linguistic feature and (ii) a tag that specifies the type of linguistic feature.
12. The non-transitory medium of claim 10 , wherein each type of linguistic feature is represented by digital elements in a different color.
13. The non-transitory medium of claim 9 , wherein the visualization component includes a graphic within a user interface window that provides the visually displayed metric used in a text description of the reasons supporting the determination of the first type of threat different from an overall numerical threat indicator displayed on the same user interface window.
14. The non-transitory medium of claim 9 , wherein at least one of the multiple ML models is unable to identify any instances of the corresponding type of linguistic feature.
15. The non-transitory medium of claim 9 , wherein the multiple ML models are able to identify financial terminology, credential theft terminology, engagement terminology, request terminology, or any combination thereof.
16. The non-transitory medium of claim 9 , wherein at least one of the multiple ML models is able to identify terminology indicative of attachments or links included in the email.
17. A system comprising:
a processor configured to:
acquire an email addressed to an employee of an enterprise;
apply multiple machine learning (ML) models to the email to identify linguistic features that are indicative of content, sentiment, or tone of the email, wherein each ML model is able to detect linguistic features of a different type, such that the multiple ML models are able to collectively surface information regarding a goal of the email;
determine that the email poses a first type of threat included in a plurality of types of potentially posed threats, based at least in part on the linguistic features identified by the multiple ML models; and
cause display of a visualization component that indicates how a determination of the first type of threat was made on an interface, including by indicating a type of the first type of threat, and a plurality of reasons supporting the determination, wherein the plurality of reasons includes, for a specific reason included in the plurality of reasons, a visually displayed metric comparing specific content included in the email with respect to content of other emails that provides a numerical explanation of the completed determination of the first type of threat for the specific email; and
a memory coupled to the processor and configured to provide the processor with instructions.
18. The system of claim 17 , wherein the visualization component includes the email with digital elements embedded therein to visually highlight the linguistic features.
19. The system of claim 18 , wherein each digital element includes (i) one or more words that are representative of the linguistic feature and (ii) a tag that specifies the type of linguistic feature.
20. The system of claim 18 , wherein each type of linguistic feature is represented by digital elements in a different color.
21. The system of claim 17 , wherein the visualization component includes a graphic within a user interface window that provides the visually displayed metric used in a text description of the reasons supporting the determination of the first type of threat different from an overall numerical threat indicator displayed on the same user interface window.
22. The system of claim 17 , wherein at least one of the multiple ML models is unable to identify any instances of the corresponding type of linguistic feature.
23. The system of claim 17 , wherein the multiple ML models are able to identify financial terminology, credential theft terminology, engagement terminology, request terminology, or any combination thereof.
24. The system of claim 17 , wherein at least one of the multiple ML models is able to identify terminology indicative of attachments or links included in the email.