Trust monitoring for input to messaging group

1 . A computing device comprising:

one or more processors; and

memory storing instructions that, when executed by the one or more processors, cause the computing device to:

detect text, input by a user of a messaging application, associated with a messaging group comprising a plurality of different users;

process the text to determine at least one topic;

generate a trust graph that comprises a plurality of connections between the user and each of the plurality of different users by processing a history of messages between the user and the plurality of different users, where each of the plurality of connections indicates a degree of trust, corresponding to the at least one topic, between the user and each of the plurality of different users, the degree of trust being based on a number of times the at least one topic had been discussed between the user and a specific user of the plurality of different users, the number of times being higher indicating a higher degree of trust and the number of times being lower indicating a lower degree of trust;

detect that a new user has joined the messaging group:

generate an updated trust graph that comprises a second plurality of connections between the user, each of the plurality of different users, and the new user:

determine, based on the updated trust graph, that a first degree of trust, associated with the new user, falls below a threshold, wherein the threshold is based on a security level of the at least one topic; and

cause output of a notification comprising an identity of the new user as untrustworthy; and

prevent the user from adding the text to the messaging group until the user acknowledges the notification by pressing a button or closing the notification.

2 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic using a topic model.

3 . The computing device of claim 2 , wherein the topic model is a Latent Dirichlet allocation generative statistical model.

4 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic by causing the computing device to:

provide, to a trained machine learning model, the text, wherein the trained machine learning model has been trained, using training data, to determine one or more topics based on input data, and wherein the training data comprises messaging data tagged based on the one or more topics; and

receive, as output from the trained machine learning model and based on the text, the at least one topic.

5 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic by causing the computing device to:

remove, from the text, one or more stop words.

6 . The computing device of claim 1 , wherein the security level comprises a sensitivity level, and wherein:

based on a determination that the at least one topic corresponds to a higher sensitivity level, a higher threshold is selected for the first degree of trust, and

based on a determination that the at least one topic corresponds to a lower sensitivity level, a lower threshold is selected for the first degree of trust.

7 . The computing device of claim 1 , wherein the text comprises a first topic and a second topic, and wherein the first topic is associated with a first threshold, the second topic is associated with a second threshold, and wherein the first threshold is different from the second threshold.

8 . A method comprising:

detecting, by a computing device, text, input by a user of a messaging application, associated with a messaging group comprising a plurality of different users;

processing, by the computing device, the text to determine at least one topic;

generating, by the computing device, a trust graph that comprises a plurality of connections between the user and each of the plurality of different users by processing a history of messages between the user and the plurality of different users, where each of the plurality of connections indicates a degree of trust, corresponding to the at least one topic, between the user and each of the plurality of different users, the degree of trust being based on a number of times the at least one topic had been discussed between the user and a specific user of the plurality of different users, the number of times being higher indicating a higher degree of trust and the number of times being lower indicating a lower degree of trust;

detecting that a new user has joined the messaging group;

generating an updated trust graph that comprises a second plurality of connections between the user, each of the plurality of different users, and the new user;

determining, by the computing device and based on the updated trust graph, that a first degree of trust, associated with the new user, falls below a threshold, wherein the threshold is based on a security level of the at least one topic;

causing, by the computing device, output of a notification comprising an identity of the new user as untrustworthy; and

preventing the user from adding the text to the messaging group until the user acknowledges the notification by pressing a button or closing the notification.

9 . The method of claim 8 , wherein processing the text to determine the at least one topic comprises use of a topic model.

10 . The method of claim 9 , wherein the topic model is a Latent Dirichlet allocation generative statistical model.

11 . The method of claim 8 , wherein the security level comprises a sensitivity level, and wherein:

based on a determination that the at least one topic corresponds to a higher sensitivity level, a higher threshold is selected for the first degree of trust, and

based on a determination that the at least one topic corresponds to a lower sensitivity level, a lower threshold is selected for the first degree of trust.

12 . The method of claim 8 , wherein the text comprises a first topic and a second topic, and wherein the first topic is associated with a first threshold, the second topic is associated with a second threshold, and wherein the first threshold is different from the second threshold.

13 . The method of claim 8 , wherein processing the text to determine the at least one topic comprises:

providing, to a trained machine learning model, the text, wherein the trained machine learning model has been trained, using training data, to determine one or more topics based on input data, and wherein the training data comprises messaging data tagged based on the one or more topics; and

receiving, as output from the trained machine learning model and based on the text, the at least one topic.

14 . The method of claim 8 , wherein processing the text to determine the at least one topic comprises:

remove, from the text, one or more stop words.

15 . One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors of a computing device, cause the computing device to:

detect text, input by a user of a messaging application, associated with a messaging group comprising a plurality of different users;

process the text to determine at least one topic;

generate a trust graph that comprises a plurality of connections between the user and each of the plurality of different users by processing a history of messages between the user and the plurality of different users, where each of the plurality of connections indicates a degree of trust, corresponding to the at least one topic, between the user and each of the plurality of different users, the degree of trust being based on a number of times the at least one topic had been discussed between the user and a specific user of the plurality of different users, the number of times being higher indicating a higher degree of trust and the number of times being lower indicating a lower degree of trust;

detect that a new user has joined the messaging group:

generate an updated trust graph that comprises a second plurality of connections between the user, each of the plurality of different users, and the new user:

determine, based on the updated trust graph, that a first degree of trust, associated with the new user, falls below a threshold, wherein the threshold is based on a security level of the at least one topic; and

cause output of a notification comprising an identity of the new user as untrustworthy and prevent the user from adding the text to the messaging group until the user acknowledges the notification by pressing a button or closing the notification.

16 . The computer-readable media of claim 15 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic using a topic model.

17 . The computer-readable media of claim 16 , wherein the topic model is a Latent Dirichlet allocation generative statistical model.

18 . The computer-readable media of claim 15 , wherein the security level comprises a sensitivity level, and wherein:

based on a determination that the at least one topic corresponds to a higher sensitivity level, a higher threshold is selected for the first degree of trust, and

based on a determination that the at least one topic corresponds to a lower sensitivity level, a lower threshold is selected for the first degree of trust.

19 . The computer-readable media of claim 15 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic by causing the computing device to:

provide, to a trained machine learning model, the text, wherein the trained machine learning model has been trained, using training data, to determine one or more topics based on input data, and wherein the training data comprises messaging data tagged based on the one or more topics; and

receive, as output from the trained machine learning model and based on the text, the at least one topic.

20 . The computer-readable media of claim 15 , wherein the instructions, when executed by the one or more processors, cause the computing device to process the text to determine the at least one topic by causing the computing device to:

remove, from the text, one or more stop words.