Real-Time Synthetic and Imitation Voice Detection System for Telecommunication Networks

Abstract

The invention relates to a system used in telecommunications networks, cybersecurity, voice signal processing, and artificial intelligence-based fraud prevention systems that enables the detection of synthetic or imitative voices generated using artificial intelligence technologies in voice calls by analyzing voice calls made over an operator core network in real time.

Description

1 TARIFF Real-time synthetic and spoofed voice detection in telecommunication networks. The system Technical Area The invention has implications for telecommunications networks, cybersecurity, voice signal processing, and artificial intelligence. operator core network used in the field of fraud prevention systems based on data transfer. Real-time analysis of voice calls made over the Core Network. By using artificial intelligence technologies, synthetic voice searches are performed. or it relates to a system that enables the detection of imitated voices. State of the Art 10 Today, voice calls are made over telecommunication networks. Calling line information (CLI) is primarily used in verification. However, only Using CLI information for verification purposes, such as number spoofing, is prohibited. synthetic or created using methods and generative artificial intelligence technologies It does not provide adequate protection against imitative voices. 15 The current technology focuses on analyzing voice searches using artificial intelligence. While there are some end-user applications, these applications are mostly It operates on end-user devices. Therefore, the processing power of the devices and Performing real-time audio and spectral analysis within the limits of available resources. This is necessary; this situation leads to delays in the analysis process and a 20% accuracy rate in the detection. This can lead to limitations. However, in device-based solutions, each subscriber needs the relevant software or application. It needs to be used or installed separately. This situation concerns telecommunications. the creation of a centralized and standardized security mechanism at the network level making it more difficult and creating a common fraud prevention structure that covers all subscribers. 25 It limits its implementation. Application number US20240355323A1, the call tone being received by a server, Speech-to-text conversion and textual / semantic anomalies through NLP It is related to the system that detects voice-based synthetic speech through examination. However, The application includes spectral and biometric data of the voice signal captured over the network. 2 the characteristics are evaluated in separate analysis stages and the results obtained are used as a risk assessment. combined in the evaluation unit and analyzed in real time It is not being implemented. In conclusion, due to the negative aspects described above and the current solutions being the subject of discussion... Due to its shortcomings, an improvement in the relevant technical field is necessary. 5 It has been made. Purpose of the Invention The invention was created by drawing inspiration from existing situations and overcoming the aforementioned drawbacks. It aims to solve the problem. The main purpose of the invention is to enable telecommunications operators to transmit signals over their core network to 10 the real-time analysis of the voice calls made and the said voice searches are synthetic or simulated searches created using artificial intelligence technologies. The aim is to determine whether or not it contains sounds. The aim of the invention is to analyze the biometric parameters of the voice signal involved in voice calls. Specifically, by analyzing characteristics such as frequency, timbre, and rhythm, 15 The aim is to enable the detection of synthetic sound tracks that are difficult to distinguish. Another aim of the invention is to operate at the operator level, from the end-user device. to create an independent security mechanism and through this mechanism the operator The goal is to provide standard and highly secure protection for all of its subscribers. Another purpose of the invention is to detect suspicious audio patterns while the search continues 20 the user is notified in real time, thus detecting potential voice-based scams. The aim is to warn the user against these attempts. Another purpose of the invention is manual control and evaluation of voice calls. by reducing the number of processes and carrying out these processes digitally, and This aims to reduce operational costs. 25 Another purpose of the invention is to detect AI-based spurious voices in voice searches. by ensuring the security of subscribers and customer trust in the operator. to increase. 3 The invention is relevant to mobile telecommunications operators, banking voice response (IVR) systems, and remote identification. verification processes, call center security protocols, and corporate voice communication These are networks. They are also used in voice confirmation stages in eSIM and digital subscription management systems. It can also be used. The structural and characteristic features and all the advantages of the invention are given in the figures and 5 below. This becomes clearer thanks to the detailed explanation written with references to these figures. This will be understood, and therefore the evaluation should also take these forms and detailed explanations into consideration. It needs to be done by taking precautions. Figures that will help understand the invention. Figure 1 is the block diagram of the system. 10 Explanation of Part References 1. Audio signal sampling unit 2. Spectral analysis module 3. Biometric verification engine 4. Risk assessment unit 15 5. Network-based alert interface Detailed Description of the Invention This detailed explanation describes the actual situation in telecommunication networks that are the subject of the invention. Synthetic and imitation voice detection systems are preferred configurations, only the subject matter It is explained in order to facilitate a better understanding. 20 The invention enables voice communication during voice calls over telecommunication networks. real-time analysis of data and the use of artificial intelligence technologies It is a system that enables the detection of synthetic or imitative sounds. (Invention) as part of this, every voice call passes through an autonomous security filter. This system provides real-time and 25-hour voice data passing through the network. audio signal sampling unit (1) that captures lossless sound signals from the signal sampling unit (1) spectral analyzes the harmonic structure and artificial production signs of the received audio data. analysis module (2), which compares the characteristic features of the sound with the recorded models biometric authentication engine (3), spectral analysis module (2) and biometric authentication 4 Evaluate the analysis results obtained by the engine (3) and determine whether the sound is synthetic or A risk assessment unit that creates a security score determining the likelihood of an impersonation. (4), the risk level reaches a critical value according to the generated security score or The network informs the subscriber if the data is evaluated as falling below a defined critical threshold. It includes a base alert interface (5). 5 The working process of the invention is via the network by the sound signal sampling unit (1). It begins with the instantaneous and lossless capture of the transmitted audio data. Audio signal Sound data captured by sampling unit (1) is spectrally analyzed for analysis. It is transmitted to module (2). The spectral analysis module (2) analyzes the sound data transmitted to it and determines the harmonics of the sound. It examines the structure and signs of artificial production in the audio signal. Indicators of artificial production, such as synthetic noise, are being analyzed. After spectral analysis, the voice data is processed by the biometric authentication engine (3). is being evaluated. The biometric authentication engine (3) evaluates the characteristic features of the voice. reviewing and, if approved, applying those features to registered models 15 Thus, the characteristic features of the voice related to the human voice are compared. It is ensured that it is evaluated in this respect. Risk assessment unit (4), spectral analysis module (2) and biometric verification engine (3) evaluates the analysis results obtained by bringing them together and It creates a safety score that determines the likelihood of the voice being synthetic or an imitation. 20 The risk level associated with voice calls is assessed based on the generated score. The risk level reaches a critical value or falls below the defined critical threshold. If evaluated, the network-based warning interface (5) is activated. Network-based warning interface (5) provides visual or audible alerts to the relevant subscriber in case of high risk. by sending a warning message that the voice call in question may be risky, the actual 25 It provides timely information. This invention involves capturing the sound signal over the network, using spectral and biometric methods. analyzing the characteristics, evaluating the analysis results as a risk score, and This includes steps to alert the subscriber in real-time in high-risk situations. It functions as an autonomous security filter for voice calls. 30 The invention describes a system that operates at the call signaling protocol (SIP) level, using voice packets. It continuously monitors the data flow between the subscriber and the switchboard by analyzing it via (RTP).

Claims

6 REQUESTS 1. Real-time synthetic and imitative voice detection system in telecommunication networks. and its characteristic is;  Audio capture system that instantly and losslessly captures voice data passing through the network. signal sampling unit (1), 5  the harmonic structure of the sound data received from the signal sampling unit (1) and Spectral analysis module (2) which analyzes signs of artificial production.  Biometrics that compare the characteristic features of the voice with recorded models verification engine (3),  Spectral analysis module (2) and biometric verification engine (3) by 10 By evaluating the analysis results obtained, we can determine whether the voice is synthetic or imitation voice. Risk assessment that creates a safety score determining the probability of occurrence. unit (4),  the risk level reaching a critical value according to the generated security score or if assessed as below the determined critical threshold, the subscriber will be dismissed for 15 It includes an informative network-based alert interface (5).

Images