# US 10,484,532 B1 · System and method detecting fraud using machine-learning and recorded voice clips > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,484,532 B1 - Document ID: US10484532B1 - Status: Granted - Kind: B1 - Application number: 16/167,710 - Filed: 2018-10-23 - Granted: 2019-11-19 - Earliest claimed: 2018-10-23 - Assignees: Capital One Services, LLC - Applicants: Capital One Services, LLC - Inventors: Kaitlin Newman; Jeffrey Rule; Rajko Ilincic - Web page: https://patentgrep.com/patent/US10484532B1 ## CPC - H04M 3/2281: {Call monitoring, e.g. for law enforcement purposes; Call tracing; Detection or prevention of malicious calls} - G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning - G06N 3/09: Supervised learning - G06N 20/00: Machine learning - H04L 63/14: {for detecting or protecting against malicious traffic} - H04M 3/436: Arrangements for screening incoming calls {, i.e. evaluating the characteristics of a call before deciding whether to answer it (based on the calling party profile H04M3/42059; based on location H04M3/42348; based on presence H04M3/42365; diversion H04M3/54)} - H04M 3/5183: {Call or contact centers with computer-telephony arrangements} - H04W 12/12: Detection or prevention of fraud - G06N 3/043: based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS] - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/045: Combinations of networks - G06N 3/049: Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - H04M 2201/40: using speech recognition - H04M 2203/6027: Fraud preventions - H04M: TELEPHONIC COMMUNICATION (circuits for controlling other apparatus via a telephone cable and not involving telephone switching apparatus G08) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) - H04W: WIRELESS COMMUNICATION NETWORKS (broadcast communication H04H; communication systems using wireless links for non-selective communication, e.g. wireless extensions H04M1/72) ## Abstract A system and method are disclosed for training a machine-learning model to detect characteristics of fraudulent calls. The machine-learning model is trained using audio clips, voice recognition, call handler feedback and general public knowledge of commercial risks to detect and divert fraudulent calls, thereby alleviating the burdens otherwise placed on call center service representatives. ## Claims ### Claim 1 A system for detecting fraudulent calls received from an incoming call network by a call service center comprising a plurality of call handlers comprises: a fraud training table comprising at least one fraud training entry, the fraud training entry storing information related to at least one characteristic of a fraudulent caller as indicated by a call handler, the at least one characteristic including an audible characteristic; a machine-learning model coupled to receive the at least one fraud training entry, and to use the fraud training entry for training the machine-learning model to provide a trained machine-learning model; and a call center interface, coupled between the incoming call network and the call handler, to apply the trained machine-learning model to data associated with a call and to selectively route the call received from the incoming call network to the call handler in response to applying the machine-learning model. ### Claim 2 (depends on 1) The system of claim 1 further comprising a storage device coupled to the machine-learning model and comprising call data. ### Claim 3 (depends on 1) The system of claim 1 wherein the machine-learning model includes logic for calculating a probability that the call is fraudulent. ### Claim 4 (depends on 3) The system of claim 3 wherein the machine-learning model includes logic for dynamically updating the probability in response to additionally received data for the call. ### Claim 5 (depends on 4) The system of claim 4 further comprising a user interface for displaying the probability to the call handler. ### Claim 6 (depends on 5) The system of claim 5 wherein the user interface comprises an input enabling the call handler to manually indicate that the call is fraudulent. ### Claim 7 (depends on 5) The system of claim 5 wherein the input is selected from a group including a probability input field, a sliding scale input and a fraud button. ### Claim 8 (depends on 1) The system of claim 1 wherein the audible characteristics are selected from a group including background noise, voice attributes and speech patterns. ### Claim 9 (depends on 1) The system of claim 1 wherein the at least one characteristic includes an origin characteristic selected from a group including area code, country code and Internet Protocol (IP) address. ### Claim 10 A method for detecting fraudulent calls received at a call service center includes the steps of: receiving a call at the call service center and selectively forwarding the call to a call handler; monitoring characteristics of the call to locate audible characteristics, including background noises, which are identified in a fraud training table as being associated with fraudulent calls by a call handler; and responsive to the location of an audible characteristic that is identified in the fraud training table, calculating a probability that the call is from a fraudulent caller. ### Claim 11 (depends on 10) The method of claim 10 further including the step of displaying the probability to the call handler. ### Claim 12 (depends on 10) The method of claim 10 wherein the step of calculating the probability includes the step of updating the probability during the call in response to further monitoring of the call. ### Claim 13 (depends on 10) The method of claim 10 wherein the step of calculating the probability operates in response to input from the call handler. ### Claim 14 (depends on 10) The method of claim 10, further including the step of deploying a call agent for each call received by the call center, the call agent comprising a machine-learning model trained using the fraud training table, and wherein the step of monitoring is performed by the call agent. ### Claim 15 (depends on 10) The method of claim 10 further including the steps of: storing an audio file for the call; in response to a determination that the call is from the fraudulent caller, parsing the audio file to extract characteristics of the call; analyzing the extracted characteristics; and updating the fraud training table with the extracted characteristics of the call. ### Claim 16 (depends on 15) The method of claim 15 further including the step of calculating a probability factor for the extracted characteristic and storing the probability factor with the extracted characteristic in the fraud training table. ### Claim 17 (depends on 11) The method of claim 11 wherein the step of displaying includes the step of providing a graphical user interface (GUI) for display on a computing device of the call handler, the GUI comprising a fraud detection window comprising an first mechanism for receiving information from the call handler and a second mechanism for displaying the probability to the call handler. ### Claim 18 (depends on 17) The method of claim 17 wherein the first mechanism is selected from a group including a probability input field, a sliding scale input and a fraud button. ### Claim 19 (depends on 11) The method of claim 11 wherein the step of displaying includes providing one of a visual indication or an audible indication of the probability to the call handler. ### Claim 20 A method for detecting fraudulent calls received at a call service center includes the steps of: training a machine-learning model to recognize characteristics, including background noise characteristics, of fraudulent calls to provide a trained machine-learning model, the training performed using a fraud training table, wherein the fraud training table comprises a plurality of entries corresponding to a plurality of characteristics of fraudulent calls; receiving a call at the call service center; deploying a call agent to handle the call, the call agent comprising a copy of the trained machine-learning model; during the call, the call agent analyzing characteristics of the call to identify characteristics of the call that are associated with fraudulent calls utilizing the trained machine-learning model; calculating a probability that the call is fraudulent in response to the step of analyzing; responsive to the probability, selectively routing the call to a call handler; displaying the probability at a graphical user interface (GUI) of the call handler; and determining, in response to at least one of a call handler input or the probability, that the call is from a fraudulent caller; and diverting the call.