CN103685613B - A kind of anti-telephone fraud system and method based on speech recognition - Google Patents

Abstract

The invention discloses a phone fraud prevention system and method based on speech recognition. The system includes a start switch, a main control module, a voice acquisition module, a storage module, a remote communication module, a phone fraud keyword database and a prompt module. Among them, the main control module is used for voice feature extraction, keyword recognition, fraudulent keyword comparison and result judgment; the remote communication module is used for remote update of telephone fraud keyword database and data sending and receiving function when real-time upload data is required; The fraud keyword database consists of several keywords extracted from common telephone fraud cases. The system of the present invention is built into a fixed phone or a mobile phone, which can help phone users identify endless phone frauds; when the phone equipment cannot complete the identification in time due to reasons such as cost or its own computing power, the present invention can also be used without leaking the content of the user's call. Under the premise of remote identification of telephone fraud keywords through background services.

Description

A phone fraud prevention system and method based on speech recognition

technical field

The invention belongs to the communication field, in particular to a phone fraud prevention system based on speech recognition and a method thereof.

Background technique

At present, there are more and more cases of telephone fraud, and fraud methods emerge in endlessly, especially for the elderly, people with insufficient knowledge or people who know less about similar fraud information, they are more likely to be deceived and cause economic losses. At present, there is still a lack of timely and effective techniques and methods for helping victims to see through fraudulent traps in time and accurately before being deceived for such telephone fraud cases. The invention is an anti-telephone fraud system based on speech recognition, which provides timely and effective technology and method to help victims see through fraud traps before being deceived.

Homomorphic encryption is a form of encryption that allows people to perform specific algebraic operations on ciphertext to obtain an encrypted result, just like performing the same operation on plaintext and then encrypting the result. In other words, this technology allows people to perform operations such as retrieval and comparison on encrypted data and get correct results without decrypting the data during the entire process. Its significance lies in that it truly fundamentally solves the confidentiality problem when entrusting data and its operations to a third party, such as for various cloud computing applications.

Contents of the invention

In order to overcome the deficiencies of the prior art, the present invention provides a voice recognition-based anti-telephone fraud system and its method. The system is built into a fixed phone or a mobile phone, which can help phone users to identify endless phone frauds; when the phone equipment cannot complete the identification in time due to reasons such as cost or its own computing power, the method of the present invention can also be used without leaking the user's phone calls. Under the premise of the content, remote identification of phone fraud keywords through background services to help users see through phone fraud.

The technical scheme adopted in the present invention is:

A phone fraud prevention system based on language recognition, including a start switch, a main control module, a voice acquisition module, a storage module, a remote communication module and a prompt module, the storage module includes a temporary storage area and a phone fraud keyword library, the The voice acquisition module, the memory module, the remote communication module, and the prompt module are connected to the main control module respectively; the start switch is used to start the system work; To and result judgment; The voice acquisition module is used to obtain the call voice in real time after the system starts; The storage module is used to store the processed voice features and the keywords identified, and the telephone fraud keyword library is also stored in the storage module In; the remote communication module is used for remotely updating the telephone fraud keyword database and the data sending and receiving function when uploading data in real time; the prompt module is used for giving the result prompt of telephone fraud, which can be sound or vibration or screen display or a combination of the above three; the telephone fraud keyword library is composed of several keywords extracted from common telephone fraud cases, which can be dynamically added, and the combination of several keywords is used to judge the possibility of telephone fraud.

As a further improvement of the present invention, it also includes a remote background server, the remote background server includes a homomorphic encryption data processing module and a phone fraud keyword library, and the storage module also includes a homomorphic encryption data storage module.

A method for preventing telephone fraud based on language recognition, specifically including the following processes:

Step 1: Establish the above-mentioned anti-telephone fraud system based on language recognition;

Step 2: The user activates the phone fraud identification system after triggering the start switch while answering the phone;

Step 3: the voice acquisition module acquires the voice of the person talking to the user in real time through the handset of the telephone device;

Step 4: the main control module processes the voice data, and saves the processed data in the storage module;

Step 5: Compare the extracted keywords with the telephone fraud keyword database. When the score of the comparison result is greater than the set identification threshold, the call is identified as telephone fraud; otherwise, it is not identified as fraud;

Step 6: When it is identified as phone fraud, the prompt module prompts the user which type of fraud it may be.

Further, the telephone fraud keyword database can be regularly updated through the remote communication module, and the updated keyword database is provided by government departments or specialized merchants.

Further, when the user is idle on the telephone device, that is, when the user is not talking, the system will trigger the start switch to ask the user whether to update the keyword database by voice or screen display. After the user agrees, the remote communication function will be automatically activated to update the current keyword database. The user can interrupt the process.

A phone fraud prevention method based on language recognition. The method is based on homomorphic encryption technology. When the phone device cannot complete the recognition in time due to reasons such as cost or its own computing power, it can pass the background service without revealing the content of the user's call. The remote identification of telephone fraud keywords specifically includes the following process:

Step 1: set up the anti-telephone fraud system based on language recognition described in claim 2;

Step 2: When the user is answering the phone, the phone fraud identification system will be activated after triggering the activation switch;

Step 3: the voice acquisition module will acquire the voice of the person talking to the user in real time through the handset of the telephone device;

Step 4: the main controller processes the voice data, and saves the processed data in the storage unit;

Step 5: call the remote communication module to transmit the encrypted data to the remote background server;

Step 6: The remote background server uses a special processing program to process the voice data received by the user after homomorphic encryption, extracts and compares the telephone fraud keyword information; in order to ensure that the remote background does not decrypt the user data It can also complete the work of keyword extraction and recognition, which better guarantees the privacy and security of users;

Step 7: After receiving the ciphertext information sent by the remote background server, the user uses the system built-in homomorphic encryption algorithm decryption key to decrypt the ciphertext to obtain the recognition result;

Step 8: When it is identified as phone fraud, the prompt module prompts the user which kind of fraud method it may be; the prompt method can be sound prompt, vibration, screen display or a combination of two or three;

Step 9: The keyword library of remote background fraud means is provided by government departments or specialized merchants, and is regularly updated to ensure the latest status of the keyword library;

Step 10: When the user is idle on the telephone equipment, that is, when the user is not talking, the system will trigger the start switch to ask the user whether to update the keyword database through voice or screen display. After the user agrees, the remote communication function will be automatically activated to update the current keyword database. The user can interrupt the process.

Further, the main controller described in step 4 processes the voice data as follows: the hardware filters out the background noise in the voice, and then performs homomorphic encryption on the voice data; stores the encrypted data in a temporary storage area to for remote transmission.

Further, the method of data transmission in step 5 is wired network transmission, or 3G, 4G, 5G wireless communication methods commonly used by mobile devices; the remote background server is built and operated by government departments or specialized operators, providing users with Remote Data Processing Request.

Further, the special processing program described in step 6 is provided by a trusted third party, which is used to extract and compare keywords from the data homomorphically encrypted by the user, and output the comparison result; using the special processing program to The ciphertext information sent by the user is processed, and the comparison result obtained is ciphertext, which is opaque to the remote background server; the remote background server uses a special processing program to score the comparison result. When the comparison result score is greater than the set recognition threshold, It is recognized that this call is phone fraud; otherwise, it is not recognized as fraud; the remote background server will send the final judgment result to the user in cipher text.

The beneficial effects of the present invention are:

1) The system is built into landline or mobile phones, which can help phone users identify endless phone scams;

2) When the telephone equipment cannot complete the identification in time due to reasons such as cost or its own computing power, the present invention also provides a remote keyword extraction and identification method based on homomorphic encryption technology. The service remotely identifies phone fraud keywords to help users see through phone fraud.

Description of drawings

Fig. 1 is a structural schematic diagram of a telephone embedded anti-telephone fraud system;

Fig. 2 is a schematic structural diagram of a remote speech recognition and comparison telephone fraud system based on homomorphic encryption.

detailed description

The present invention will be described in detail below in conjunction with various embodiments shown in the drawings. However, these embodiments do not limit the present invention, and any structural, method, or functional changes made by those skilled in the art according to these embodiments are included in the protection scope of the present invention.

Example 1:

An anti-telephone fraud system based on language recognition of the present invention is shown in Figure 1, including a start switch, a main control module, a voice acquisition module, a storage module, a remote communication module and a prompt module. The storage module includes a temporary storage area and a telephone fraud keyword library, and the voice acquisition module, storage module, remote communication module and prompt module are respectively connected with the main control module.

The start switch is used to start the system work.

The main control module is used for voice feature extraction, keyword identification, fraudulent keyword comparison and result determination; the voice acquisition module is used for real-time acquisition of call voice after the system is started.

The storage module is used for storing the processed voice features and the recognized keywords, and the telephone fraud keyword library is also stored in the storage module.

The remote communication module is used for remotely updating the telephone fraud keyword database and the data sending and receiving function when real-time uploading of data is required.

The prompting module is used to give the result prompt of whether the phone is fraudulent, which can be sound or vibration or screen display or a combination of the above three.

The telephone fraud keyword database is composed of several keywords extracted from common telephone fraud cases, which can be added dynamically, and the combination of several keywords is used to judge the possibility of telephone fraud.

The anti-telephone fraud method process of the anti-telephone fraud system based on language recognition shown in Figure 1 is as follows:

Step 1 System establishment: As shown in Figure 1, the system includes a start switch, a main control module, a voice acquisition module, a storage module, a remote communication module, a telephone fraud keyword library and a prompt module.

1. The start switch is used to start the system, which is set as a mechanical button or a touch screen button;

2. The main control module is used for voice feature extraction, keyword recognition, fraudulent keyword comparison and result judgment, and is integrated in the telephone device;

3. The voice acquisition module is used to acquire the call voice in real time when the system is started, and use the shared telephone handset as a voice pickup device;

4. The storage module is used to store the processed speech features and identified keywords, and the telephone fraud keyword database is also stored in the storage module and integrated into the telephone device;

5. The remote communication module is used to remotely update the telephone fraud keyword database and the data sending and receiving function when real-time uploading of data is required; the telephone fraud keyword database is composed of several keywords extracted from common telephone fraud cases, which can be added dynamically. The combination of keywords is used to judge the possibility of telephone fraud, and the wired Internet function of the fixed phone or the wireless Internet technology of the mobile phone are used to realize long-distance communication;

6. The prompt module is used to give the result prompt of whether the phone is fraudulent, which can be sound or vibration or screen display or a combination of the above three.

Step2 When the user is answering the phone, the phone fraud identification system will be activated after triggering the activation switch;

Step3 The voice acquisition module will acquire the voice of the person talking to the user in real time through the handset of the telephone device;

Step4 The main controller processes the voice data and saves the processed data in the storage unit;

Step5 Compare the extracted keywords with the telephone fraud keyword database. When the comparison result score is greater than the set identification threshold, the call is identified as telephone fraud; otherwise, it is not identified as fraud;

Step6 When it is identified as phone fraud, the prompt module prompts the user which type of fraud it may be;

Step7 The telephone fraud keyword database can be regularly updated through the remote communication module, so that the latest fraud methods can be included in it in time, and the remote background fraud keyword database is provided by government departments or specialized merchants;

Step71 When the user is idle on the telephone device, that is, when the user is not talking, the trigger switch system will ask the user whether to update the keyword database by voice or screen display. After the user agrees, the remote communication function will be automatically activated to update the current keyword database. The user can interrupt the process.

Example 2:

When the telephone device cannot complete the identification in time due to reasons such as cost or its own computing power, a phone fraud prevention system based on language recognition in the present invention is shown in Figure 2. On the basis of Figure 1, it also includes a remote background server, a remote background The server includes a homomorphic encryption data processing module and a telephone fraud keyword library, and the storage module also includes a homomorphic encryption data storage module.

The anti-telephone fraud method process of the anti-telephone fraud system based on language recognition shown in Figure 2 is as follows:

Step1 system establishment: as shown in Figure 2, the system includes a start switch, a main control module, a voice acquisition module, a storage module, a remote communication module, a telephone fraud keyword library, a prompt module and a remote background server.

1. The start switch is used to start the system, which is set as a mechanical button or a touch screen button;

2. The main control module is used for voice feature extraction, keyword recognition, fraudulent keyword comparison and result judgment, and is integrated in the telephone device;

3. The voice acquisition module is used to acquire the call voice in real time when the system is started, and use the shared telephone handset as a voice pickup device;

4. The storage module is used to store the preprocessed voice for remote transmission, and is integrated in the telephone equipment;

5. The remote communication module is used to transmit the preprocessed voice data to the remote background server in real time after the system is started, and also receive information such as the recognition result data sent from the remote background server; use the wired Internet function of the fixed phone or the mobile phone Wireless Internet access technology to realize long-distance communication;

6. The prompt module is used to give the result prompt of whether the phone is fraudulent, which can be sound or vibration or screen display or a combination of the above three.

Step2 When the user is answering the phone, the phone fraud identification system will be activated after triggering the activation switch;

Step3 The voice acquisition module will acquire the voice of the person talking to the user in real time through the handset of the telephone device;

Step4 The main controller processes the voice data and saves the processed data in the storage unit;

Step41 The processing process is: the hardware filters out the background noise in the sound, and then performs homomorphic encryption on the sound data;

Step42 Store the encrypted data in a temporary storage area for remote transmission;

Step5 calls the remote communication module to transfer the encrypted data to the remote background server;

Step51 The transmission method can be wired network transmission, or 3G, 4G, 5G and other wireless communication methods commonly used by mobile devices;

Step52 The remote background server is built and operated by government departments or specialized operators, and is responsible for providing users with remote data processing requests;

Step6 The remote background server uses a special processing program to process the voice data received by the user after homomorphic encryption, extracts and compares the phone fraud keyword information; in order to ensure that the remote background can also decrypt the user data without decrypting Complete the work of keyword extraction and identification, better guarantee the privacy and security of users;

The special processing program of Step61 is provided by a trusted third party, and its function is to extract and compare keywords from the data homomorphically encrypted by the user, and output the comparison results; the important thing is that this process is invisible to the remote server, In other words, in the process of processing information, the remote server does not know the plaintext information of the processed voice data, and only uses this special processing program to process the ciphertext information sent by the user. The server is also opaque;

Step62 The remote background server uses a special processing program to compare and score the comparison results. When the comparison result score is greater than the set recognition threshold, it will be recognized as a phone call fraud; otherwise, it will not be recognized as a fraud;

Step63 The remote background server sends the final judgment result to the user in ciphertext;

Step7 After receiving the ciphertext information sent by the remote background server, the user uses the system built-in homomorphic encryption algorithm decryption key to decrypt the ciphertext to obtain the recognition result;

Step8 When it is identified as phone fraud, the prompt module prompts the user which type of fraud it may be;

Step81 The prompt method can be sound prompt, vibration, screen display or a combination of two or three;

Step9 The keyword library of remote background fraud means is provided by government departments or specialized merchants, and is regularly updated to ensure the latest status of the keyword library.

Step10 When the user is idle on the telephone device, that is, when the user is not talking, the trigger switch system will ask the user whether to update the keyword database by voice or screen display. After the user agrees, the remote communication function will be automatically activated to update the current keyword database. The user can interrupt the process.

Claims (3)

1. An anti-telephone fraud method based on speech recognition, said method is based on a kind of anti-telephone fraud system based on speech recognition, said system includes start switch, main control module, voice acquisition module, storage module, remote communication module and A prompt module, the storage module includes a temporary storage area and a telephone fraud keyword library, the voice acquisition module, storage module, remote communication module, and prompt module are respectively connected to the main control module; the start switch is used to start the system; The main control module is used for voice feature extraction, keyword identification, fraudulent keyword comparison and result determination; the voice acquisition module is used for real-time acquisition of call voice when the system is started; the storage module is used for storing and processing voice characteristics and identified keywords, while the telephone fraud keyword database is also stored in the storage module; the remote communication module is used for remotely updating the telephone fraud keyword database and the data sending and receiving function when uploading data in real time; The prompt module is used to give the result prompt of telephone fraud, which can be sound or vibration or screen display or a combination of the above three; the telephone fraud keyword library is composed of several keywords extracted from common telephone fraud cases, which can be dynamically Add, combine several keywords to judge the possibility of telephone fraud; also include a remote background server, the remote background server includes a homomorphic encryption data processing module and a telephone fraud keyword library, and the storage module also includes a homomorphic Encrypted data storage module; The method is based on homomorphic encryption technology. When the phone device cannot complete the identification in time due to reasons such as cost or its own computing power, the phone fraud keyword is remotely identified through the background service without revealing the content of the user's call, which specifically includes the following process : Step 1: set up the anti-telephone fraud system based on language recognition described in claim 1; Step 2: When the user is answering the phone, the phone fraud identification system will be activated after triggering the activation switch; Step 3: the voice acquisition module will acquire the voice of the person talking to the user in real time through the handset of the telephone device; Step 4: the main controller processes the voice data, and saves the processed data in the storage unit; Step 5: call the remote communication module to transmit the encrypted data to the remote background server; Step 6: The remote background server uses a special processing program to process the voice data received by the user after homomorphic encryption, extracts and compares the telephone fraud keyword information; in order to ensure that the remote background does not decrypt the user data The work of keyword extraction and identification can also be completed, which better guarantees the user's privacy and security; the special processing program is provided by a trusted third party, and is used to extract and compare keywords after homomorphic encryption by the user , and output the comparison result; use the special processing program to process the ciphertext information sent by the user, and the obtained comparison result is ciphertext, which is opaque to the remote background server; the remote background server uses a special processing program to compare the comparison result Scoring, when the score of the comparison result is greater than the set recognition threshold, it will be recognized as a phone call fraud; otherwise, it will not be recognized as a fraud; the remote background server will send the final judgment result to the user in cipher text; Step 7: After receiving the ciphertext information sent by the remote background server, the user uses the system built-in homomorphic encryption algorithm decryption key to decrypt the ciphertext to obtain the recognition result; Step 8: When it is identified as phone fraud, the prompt module prompts the user which kind of fraud method it may be; the prompt method can be sound prompt, vibration, screen display or a combination of two or three; Step 9: The keyword library of remote background fraud means is provided by government departments or specialized merchants, and is regularly updated to ensure the latest status of the keyword library; Step 10: When the user is idle on the telephone device, that is, when the user is not talking, the system will trigger the start switch and the system will ask the user whether to update the keyword database through voice or screen display. After the user agrees, the remote communication function will be automatically activated to update the current keyword database. The process can be interrupted. 2. a kind of anti-telephone fraud method based on speech recognition according to claim 1 is characterized in that: the process that main controller is processed voice data in the step 4 is: background noise in the hardware filtering sound, Then perform homomorphic encryption on the sound data; store the encrypted data in a temporary storage area for remote transmission. 3. A phone fraud prevention method based on language recognition according to claim 1, characterized in that: the data transmission method in step 5 is wired network transmission, or 3G, 4G, 5G wireless network commonly used by mobile devices Communication method; the remote background server is set up and operated by government departments or specialized operators to provide users with remote data processing requests.