Systems and methods for real time deepfake identification using knowledge graphs in virtual meetings

1 . A method for real time deepfake identification using a knowledge graph in a virtual meeting, the method comprising:

calculating trust scores of users participating in a virtual meeting;

responsive to determining that a first user among the users has a trust score below a trust score threshold, instantiating a knowledge graph representing known associations between the users;

selecting content in the knowledge graph by an Artificial Intelligence (AI) engine using machine learning, the content having a higher probability of being recognized by the first user and a second user among the users, wherein the machine learning is trained to identify content in the knowledge graph and to calculate probabilities of recognition of the content by the users;

generating, by the AI engine, a question comprising at least a portion of the content in a tone determined by a balancing algorithm that considers an organizational hierarchy of the first user and the second user;

generating instructions to a meeting-bot executing in the virtual meeting to prompt the second user to ask the question in the virtual meeting;

responsive to a reply from the second user that an answer to the question by the first user is satisfactory, allowing the virtual meeting to continue; and

responsive to a reply from the second user that the answer to the question by the first user is not satisfactory, categorizing the first user as inauthentic, and performing at least one of: (i) providing another authentication step to the first user for entry into the virtual meeting; (ii) preventing entry by the first user to the virtual meeting; and (iii) terminating the virtual meeting.

2 . The method of claim 1 , wherein the second user has a higher trust score than all other users.

3 . The method of claim 1 , wherein at least one authentication step completed by the users comprises:

sending instructions to a respective device associated with each user to generate and emit at least one of: (i) a coded ultrasonic signal, and (ii) a coded infrared signal; and

receiving in real time from the virtual meeting, at least one of: (i) audio data associated with the users and (ii) video data associated with the users.

4 . The method of claim 3 , further comprising:

performing a search of at least one of: (i) the audio data for the coded ultrasonic signal, and (ii) the video data for the coded infrared signal; and

responsive to not finding at least one of (i) the coded ultrasonic signal in the audio data and (ii) the coded infrared signal in the video data, generating a trust score that is lower than the trust score threshold.

5 . The method of claim 1 , wherein the trust score of the first user is generated based on input by any other user to the meeting-bot about possible inauthenticity of the first user.

6 . The method of claim 1 , wherein the balancing algorithm further considers a plurality of factors according to preconfigured settings.

7 . The method of claim 1 , wherein:

the knowledge graph is generated from organizational data, and the organizational data is updated in real time.

8 . Non-transitory computer-readable tangible media that includes instructions for execution, which when executed by a processor of a computing device, is operable to perform operations comprising:

calculating trust scores of users participating in a virtual meeting;

responsive to determining that a first user among the users has a trust score below a trust score threshold, instantiating a knowledge graph representing known associations between the users;

selecting content in the knowledge graph by an Artificial Intelligence (AI) engine using machine learning, the content having a higher probability of being recognized by the first user and a second user among the users, wherein the machine learning is trained to identify content in the knowledge graph and to calculate probabilities of recognition of the content by the users;

generating, by the AI engine, a question comprising at least a portion of the content in a tone determined by a balancing algorithm that considers an organizational hierarchy of the first user and the second user;

generating instructions to a meeting-bot executing in the virtual meeting to prompt the second user to ask the question in the virtual meeting;

responsive to a reply from the second user that an answer to the question by the first user is satisfactory, allowing the virtual meeting to continue; and

responsive to a reply from the second user that the answer to the question by the first user is not satisfactory, categorizing the first user as inauthentic, and performing at least one of: (i) providing another authentication step to the first user for entry into the virtual meeting; (ii) preventing entry by the first user to the virtual meeting; and (iii) terminating the virtual meeting.

9 . The non-transitory computer-readable tangible media of claim 8 , wherein the second user has a higher trust score than all other users.

10 . The non-transitory computer-readable tangible media of claim 8 , wherein at least one authentication step completed by the users comprises:

sending instructions to a respective device associated with each user to generate and emit at least one of: (i) a coded ultrasonic signal, and (ii) a coded infrared signal; and

receiving in real time from the virtual meeting, at least one of: (i) audio data associated with the users and (ii) video data associated with the users.

11 . The non-transitory computer-readable tangible media of claim 10 , wherein the operations further comprise:

performing a search of at least one of: (i) the audio data for the coded ultrasonic signal, and (ii) the video data for the coded infrared signal; and

responsive to not finding at least one of (i) the coded ultrasonic signal in the audio data and (ii) the coded infrared signal in the video data, generating a trust score that is lower than the trust score threshold.

12 . The non-transitory computer-readable tangible media of claim 8 , wherein the trust score of the first user is generated based on input by any other user to the meeting-bot about possible inauthenticity of the first user.

13 . The non-transitory computer-readable tangible media of claim 8 , wherein the balancing algorithm further considers a plurality of factors according to preconfigured settings.

14 . The non-transitory computer-readable tangible media of claim 8 , wherein:

the knowledge graph is generated from organizational data, and the organizational data is updated in real time.

15 . An apparatus comprising:

a processing circuitry;

a memory storing data; and

a communication circuitry, wherein the processing circuitry executes instructions associated with the data, the processing circuitry is coupled to the communication circuitry and the memory, and the processing circuitry and the memory cooperate, such that the apparatus is configured for:

calculating trust scores of users participating in a virtual meeting;

responsive to determining that a first user among the users has a trust score below a trust score threshold, instantiating a knowledge graph representing known associations between the users;

selecting content in the knowledge graph by an Artificial Intelligence (AI) engine using machine learning, the content having a higher probability of being recognized by the first user and a second user among the users, wherein the machine learning is trained to identify content in the knowledge graph and to calculate probabilities of recognition of the content by the users;

generating, by the AI engine, a question comprising at least a portion of the content in a tone determined by a balancing algorithm that considers an organizational hierarchy of the first user and the second user;

generating instructions to a meeting-bot executing in the virtual meeting to prompt the second user to ask the question in the virtual meeting;

responsive to a reply from the second user that an answer to the question by the first user is satisfactory, allowing the virtual meeting to continue; and

responsive to a reply from the second user that the answer to the question by the first user is not satisfactory, categorizing the first user as inauthentic, and performing at least one of: (i) providing another authentication step to the first user for entry into the virtual meeting; (ii) preventing entry by the first user to the virtual meeting; and (iii) terminating the virtual meeting.

16 . The apparatus of claim 15 , wherein the second user has a higher trust score than all other users.

17 . The apparatus of claim 15 , wherein at least one authentication step completed by the users comprises:

sending instructions to a respective device associated with each user to generate and emit at least one of: (i) a coded ultrasonic signal, and (ii) a coded infrared signal; and

receiving in real time from the virtual meeting, at least one of: (i) audio data associated with the users and (ii) video data associated with the users.

18 . The apparatus of claim 17 , wherein the apparatus is further configured for:

performing a search of at least one of: (i) the audio data for the coded ultrasonic signal, and (ii) the video data for the coded infrared signal; and

responsive to not finding at least one of (i) the coded ultrasonic signal in the audio data and (ii) the coded infrared signal in the video data, generating a trust score that is lower than the trust score threshold.

19 . The apparatus of claim 15 , wherein the trust score of the first user is generated based on input by any other user to the meeting-bot about possible inauthenticity of the first user.

20 . The apparatus of claim 15 , wherein the balancing algorithm further considers a plurality of factors according to preconfigured settings.