Digital media authentication

Abstract

A method, system and product including obtaining a media stream depicting a real-time communication of a participant in a communication context; identifying the communication context; obtaining a personalized model of the participant when communicating in the communication context, wherein the personalized model is configured to identify a behavioral pattern of the participant; executing the personalized model on at least a portion of the media stream to determine whether a behavioral pattern of the participant in the media stream matches the behavioral pattern of the participant according to the personalized model; and upon identifying a mismatch between the behavioral pattern of the participant in the media stream and the behavioral pattern of the participant according to the personalized model, performing a responsive action.

What is claimed is: 
     
         1 . A method comprising:
 obtaining a media stream associated with a participant, wherein the media stream depicting a real-time communication of the participant in a communication context;   identifying the communication context;   obtaining a personalized model of the participant when communicating in the communication context, wherein the personalized model is configured to identify a behavioral pattern of the participant;   executing the personalized model on at least a portion of the media stream to determine whether a behavioral pattern of the participant in the media stream matches the behavioral pattern of the participant according to the personalized model; and   upon identifying a mismatch between the behavioral pattern of the participant in the media stream and the behavioral pattern of the participant according to the personalized model, performing a responsive action.   
     
     
         2 . The method of  claim 1 , wherein the responsive action comprises generating an alert or blocking the real-time communication, wherein the alert indicates that the media stream is forged. 
     
     
         3 . The method of  claim 1 , wherein said identifying the mismatch comprises determining that a difference between the behavioral pattern of the participant in the media stream and the behavioral pattern of the participant according to the personalized model exceeds a threshold. 
     
     
         4 . The method of  claim 1 , wherein the personalized model comprises a classifier that is trained on a dataset, wherein the dataset comprises media records depicting communications of the participant in the communication context. 
     
     
         5 . The method of  claim 4 , wherein the dataset comprises a first class of media and a second class of media, wherein the first class comprises media records originally depicting the participant in a communication context, wherein the second class comprises media records originally depicting other people excluding the participant in the communication context, the method comprising training the personalized model to classify media as belonging to the first class or to the second class. 
     
     
         6 . The method of  claim 5  comprising:
 implementing media fabrication techniques on the first class, thereby obtaining processed records of the participant, wherein said media fabrication techniques are configured to replace the participant with different people excluding the participant, 
 implementing media fabrication techniques on the second class, thereby obtaining processed records of the other people, wherein said media fabrication techniques are configured to replace the other people, 
 adding the processed records of the participant to the first class, and 
 adding the processed records of the other people to the second class. 
 
     
     
         7 . The method of  claim 6 , wherein said implementing the media fabrication techniques on the second class comprises superimposing the participant over at least some of the other people. 
     
     
         8 . The method of  claim 1  comprising training a first personalized model of the participant under a first communication context, and training a second personalized model of the participant under a second communication context. 
     
     
         9 . The method of  claim 1 , wherein the communication context is selected from a group consisting of: a friendship relationship, a co-working relationship, a family relationship, a business relationship, a customer-client relationship, and a romantic relationship. 
     
     
         10 . The method of  claim 1 , wherein the communication context comprises a topic of the real-time communication. 
     
     
         11 . The method of  claim 1  comprising determining an identity of the participant based on at least one of:
 a facial recognition method implemented on the media stream, 
 an audio recognition method implemented on the media stream, 
 metadata of the media stream, and 
 tags relating to the participant that are attached to the media stream, 
 wherein the communication context comprises the identity of the participant. 
 
     
     
         12 . The method of  claim 1 ,
 wherein said identifying the communication context comprises determining a second participant in the real-time communication, wherein the media stream depicts the real-time communication between the participant and the second participant;   wherein the communication context is a context of the participant communicating with the second participant; and   said obtaining the personalized model comprises: obtaining a private model generated based on past communications between the participant and the second participant, wherein the past communications are not publicly accessible.   
     
     
         13 . The method of  claim 1 , wherein the behavioral pattern of the participant comprises at least one of: face movements of the participant, face gestures of the participant, a gait of the participant, a walking pattern of the participant, hand movements of the participant, frequently used phrases of the participant, a talking manner of the participant, or a voice pattern of the participant. 
     
     
         14 . The method of  claim 1  implemented on a communication system used by a second participant, wherein the communication context is a communication between the participant and the second participant, wherein the communication system is configured to retain communications between the participant and the second participant and to generate a private model for the communication context based on the retained communications. 
     
     
         15 . A computer program product comprising a non-transitory computer readable storage medium retaining program instructions, which program instructions when read by a processor, cause the processor to:
 obtain a media stream associated with a participant, wherein the media stream depicts a real-time communication of the participant in a communication context;   identify the communication context;   obtain a personalized model of the participant when communicating in the communication context, wherein the personalized model is configured to identify a behavioral pattern of the participant;   execute the personalized model on at least a portion of the media stream to determine whether a behavioral pattern of the participant in the media stream matches the behavioral pattern of the participant according to the personalized model; and   upon identifying a mismatch between the behavioral pattern of the participant in the media stream and the behavioral pattern of the participant according to the personalized model, perform a responsive action.   
     
     
         16 . The computer program product of  claim 15 , wherein the instructions, when read by the processor, cause the processor to train a first personalized model of the participant under a first communication context, and to train a second personalized model of the participant under a second communication context. 
     
     
         17 . The computer program product of  claim 15 , wherein the communication context is selected from a group consisting of: a friendship relationship, a co-working relationship, a family relationship, a business relationship, a customer-client relationship, and a romantic relationship. 
     
     
         18 . The computer program product of  claim 15 , wherein the communication context comprises a topic of the real-time communication. 
     
     
         19 . The computer program product of  claim 15 , wherein the instructions, when read by the processor, cause the processor to determine an identity of the participant based on at least one of:
 a facial recognition method implemented on the media stream,   an audio recognition method implemented on the media stream,   metadata of the media stream, and   tags relating to the participant that are attached to the media stream,   wherein the communication context comprises the identity of the participant.   
     
     
         20 . A system, the system comprising a processor and coupled memory, the processor being adapted to:
 obtain a media stream associated with a participant, wherein the media stream depicts a real-time communication of the participant in a communication context;   identify the communication context;   obtain a personalized model of the participant when communicating in the communication context, wherein the personalized model is configured to identify a behavioral pattern of the participant;   execute the personalized model on at least a portion of the media stream to determine whether a behavioral pattern of the participant in the media stream matches the behavioral pattern of the participant according to the personalized model; and   upon identifying a mismatch between the behavioral pattern of the participant in the media stream and the behavioral pattern of the participant according to the personalized model, perform a responsive action.