Automatic fraud engagement adjustment

Abstract

A method, product and system for automatic fraud engagement adjustment. The method comprises, based on monitoring of engagements and conversions, determining a segment-specific estimated quality score of a traffic segment; determining an observation-based pair-specific quality score for the traffic segment and for a specific campaign; and automatically performing mitigating fraudulent engagements in the traffic segment by reducing, in real-time, a reward for engagements in the traffic segment with respect to the specific campaign, whereby aggregative reward mitigates rewards for fraudulent engagements without specifically identifying which engagements are fraudulent.

What is claimed is: 
     
         1 . A method for mitigating fraudulent engagements in a digital traffic segment based on a limited amount of information, the digital traffic segment is one of a plurality of digital traffic segments in which a plurality of campaigns is presented, the method comprising:
 monitoring engagements and conversions of each campaign of the plurality of campaigns when presented in the plurality of digital traffic segments;   determining a segment-specific estimated quality score of the digital traffic segment based on presentation of one or more campaigns of the plurality of campaigns in the digital traffic segment, wherein the segment-specific estimated quality score below a threshold is indicative of fraudulent engagements in the digital traffic segment;   determining an observation-based pair-specific quality score for the digital traffic segment and a specific campaign, wherein the observation-based pair-specific quality score is calculated based on a number of observed engagements with the specific campaign when presented in the digital traffic segment and based on a number of observed conversions in the specific campaign when presented in the digital traffic segment; and   automatically executing one or more operations for mitigation fraudulent engagement in the digital traffic segment, wherein said automatically executing includes reducing, in real-time, a reward for engagements in the digital traffic segment with respect to the specific campaign, whereby aggregative reward mitigates rewards for fraudulent engagements without specifically identifying which engagements are fraudulent.   
     
     
         2 . The method of  claim 1  the method further comprises determining a pair-specific estimated quality score of the specific campaign for the digital traffic segment indicating an estimated quality score of the specific campaign when presented in the digital traffic segment, wherein the pair-specific estimated quality score is determined based on the segment-specific estimated quality score of the given segment and based on the observation-based pair-specific quality score for the digital traffic segment and the specific campaign, wherein said reducing the reward for the engagements in the digital traffic segment is performed based on the pair-specific estimated quality score of the specific campaign for the digital traffic segment, wherein the pair-specific estimated quality score of the specific campaign for the digital traffic segment is determined based on a weighted value of the segment-specific estimated quality score of the digital traffic segment and based on a weighted value of the observation-based pair-specific quality score for the digital traffic segment and the specific campaign. 
     
     
         3 . The method of  claim 2 , wherein the weighted value of the segment-specific estimated quality score of the digital traffic segment is based on a first weight, wherein the weighted value of the observation-based pair-specific quality score for the digital traffic segment and the specific campaign is based on a second weight, the method further comprises: updating the first and second weights, thereby adapting the pair-specific estimated quality score of the specific campaign for the digital traffic segment. 
     
     
         4 . The method of  claim 3 , wherein said updating is based on amount of available information to compute the observation-based pair-specific quality score, whereby as more information becomes available, the first weight is decreased and the second weight is increased. 
     
     
         5 . The method of  claim 2 , wherein the weighted value of the segment-specific estimated quality score of the given segment is based on a first weight, wherein the weighted value of the observation-based pair-specific quality score for the given traffic segment and the specific campaign is based on a second weight, wherein the first and second weights are computed based on at least one of:
 the number of observed engagements with the specific campaign when presented in the digital traffic segment, and   the number of observed conversions in the specific campaign when presented in the digital traffic segment.   
     
     
         6 . The method of  claim 5 , wherein the number of observed engagements with the specific campaign when presented in the digital traffic segment and the number of observed conversions in the specific campaign when presented in the digital traffic segment are limited to number of observed engagements and conversions, respectively, within a sliding time window. 
     
     
         7 . The method of  claim 1 , wherein each traffic segment from the plurality of traffic segments is associated with a different content publisher. 
     
     
         8 . The method of  claim 1 , wherein the reward is given per a user engagement with content of the one or more campaigns presented in the digital traffic segment. 
     
     
         9 . The method of  claim 1 , wherein the digital traffic segment is associated to any one of: a website; a time frame; and a group of users having one or more common characteristics. 
     
     
         10 . The method of  claim 1 , wherein the method is performed with a processing complexity that is bound to an order of magnitude of O(|S|+|C|) Central Processing Unit (CPU) operations, wherein |S| is a number of the plurality of digital traffic segments, wherein |C| is a number of the plurality of campaigns. 
     
     
         11 . An apparatus for mitigating fraudulent engagements in a digital traffic segment based on a limited amount of information, the digital traffic segment is one of a plurality of digital traffic segments in which a plurality of campaigns is presented, the apparatus comprising one or more computer processors configured and operable to perform:
 monitoring engagements and conversions of each campaign of the plurality of campaigns when presented in the plurality of digital traffic segments;   determining a segment-specific estimated quality score of the digital traffic segment based on presentation of one or more campaigns of the plurality of campaigns in the digital traffic segment, wherein the segment-specific estimated quality score below a threshold is indicative of fraudulent engagements in the digital traffic segment;   determining an observation-based pair-specific quality score for the digital traffic segment and a specific campaign, wherein the observation-based pair-specific quality score is calculated based on a number of observed engagements with the specific campaign when presented in the digital traffic segment and based on a number of observed conversions in the specific campaign when presented in the digital traffic segment; and   automatically executing one or more operations for mitigation fraudulent engagement in the digital traffic segment, wherein said automatically executing includes reducing, in real-time, a reward for engagements in the digital traffic segment with respect to the specific campaign, whereby aggregative reward mitigates rewards for fraudulent engagements without specifically identifying which engagements are fraudulent.   
     
     
         12 . The apparatus of  claim 11 , said one or more computer processors are further configured and operable to perform: determining a pair-specific estimated quality score of the specific campaign for the digital traffic segment indicating an estimated quality score of the specific campaign when presented in the digital traffic segment, wherein the pair-specific estimated quality score is determined based on the segment-specific estimated quality score of the given segment and based on the observation-based pair-specific quality score for the digital traffic segment and the specific campaign, wherein said reducing the reward for the engagements in the digital traffic segment is performed based on the pair-specific estimated quality score of the specific campaign for the digital traffic segment, wherein the pair-specific estimated quality score of the specific campaign for the digital traffic segment is determined based on a weighted value of the segment-specific estimated quality score of the digital traffic segment and based on a weighted value of the observation-based pair-specific quality score for the digital traffic segment and the specific campaign. 
     
     
         13 . The apparatus of  claim 12 , wherein the weighted value of the segment-specific estimated quality score of the digital traffic segment is based on a first weight, wherein the weighted value of the observation-based pair-specific quality score for the digital traffic segment and the specific campaign is based on a second weight, the method further comprises: updating the first and second weights, thereby adapting the pair-specific estimated quality score of the specific campaign for the digital traffic segment. 
     
     
         14 . The apparatus of  claim 13 , wherein said updating is based on amount of available information to compute the observation-based pair-specific quality score, whereby as more information becomes available, the first weight is decreased and the second weight is increased. 
     
     
         15 . The apparatus of  claim 12 , wherein the weighted value of the segment-specific estimated quality score of the given segment is based on a first weight, wherein the weighted value of the observation-based pair-specific quality score for the given traffic segment and the specific campaign is based on a second weight, wherein the first and second weights are computed based on at least one of:
 the number of observed engagements with the specific campaign when presented in the digital traffic segment, and   the number of observed conversions in the specific campaign when presented in the digital traffic segment.   
     
     
         16 . The apparatus of  claim 15 , wherein the number of observed engagements with the specific campaign when presented in the digital traffic segment and the number of observed conversions in the specific campaign when presented in the digital traffic segment are limited to number of observed engagements and conversions, respectively, within a sliding time window. 
     
     
         17 . The apparatus of  claim 11 , wherein a processing complexity of processing performed by said one or more computer processors is bound to an order of magnitude of O(|S|+|C|) Central Processing Unit (CPU) operations, wherein |S| is a number of the plurality of digital traffic segments, wherein |C| is a number of the plurality of campaigns. 
     
     
         18 . A non-transitory computer readable storage medium for mitigating fraudulent engagements in a digital traffic segment based on a limited amount of information, the digital traffic segment is one of a plurality of digital traffic segments in which a plurality of campaigns is presented, said non-transitory computer readable storage medium tangibly embodying a program of instructions that, when executed by a computer, cause the computer to perform:
 monitoring engagements and conversions of each campaign of the plurality of campaigns when presented in the plurality of digital traffic segments;   determining a segment-specific estimated quality score of the digital traffic segment based on presentation of one or more campaigns of the plurality of campaigns in the digital traffic segment, wherein the segment-specific estimated quality score below a threshold is indicative of fraudulent engagements in the digital traffic segment;   determining an observation-based pair-specific quality score for the digital traffic segment and a specific campaign, wherein the observation-based pair-specific quality score is calculated based on a number of observed engagements with the specific campaign when presented in the digital traffic segment and based on a number of observed conversions in the specific campaign when presented in the digital traffic segment; and   automatically executing one or more operations for mitigation fraudulent engagement in the digital traffic segment, wherein said automatically executing includes reducing, in real-time, a reward for engagements in the digital traffic segment with respect to the specific campaign, whereby aggregative reward mitigates rewards for fraudulent engagements without specifically identifying which engagements are fraudulent.   
     
     
         19 . The non-transitory computer readable storage medium of  claim 18 , wherein a processing complexity of processing performed by a processor of the computer when executing said program instructions is bound to an order of magnitude of O(|S|+|C|) Central Processing Unit (CPU) operations, wherein |S| is a number of the plurality of digital traffic segments, wherein |C| is a number of the plurality of campaigns.