CN105933272A - Voiceprint recognition method capable of preventing recording attack, server, terminal, and system - Google Patents

Abstract

The present invention provides a voiceprint authentication method, server, terminal and system capable of preventing recording attacks. The voiceprint authentication method includes: generating character combinations and character pronunciation rules according to a user's voiceprint authentication request; and the pronunciation rules of the character are sent to the requesting terminal; receiving the user voice input by the requesting terminal according to the character combination and the pronunciation rule of the character; performing voiceprint authentication according to the user voice, the character combination and the pronunciation rule of the character; Send the voiceprint authentication result to the requesting terminal. The invention can effectively prevent recording attacks.

Description

Voiceprint authentication method, server, terminal and system capable of preventing recording attacks

technical field

The invention belongs to the field of voiceprint recognition, in particular to a voiceprint authentication method, server, terminal and system capable of preventing recording attacks.

Background technique

Like fingerprints, voiceprint is a very important biological feature that can represent a person's identity. Compared with traditional methods such as password authentication, voiceprint has the characteristics of high security and convenience. The most commonly used attack methods in voiceprint authentication mainly include recording playback attack, speaker spoofing attack and fake authentication voice attack.

Among them, the recording playback attack refers to that the attacker obtains the user's voice sample through various means through high-fidelity recording equipment, uses the user's original recording or processes it through cropping, splicing, etc. When the user's voice is collected, it is played back through a high-fidelity power amplifier to attack. Speaker spoofing attack means that some attackers who are good at acquiescing to other people's voice attack by imitating the speaker's speaking style and pronunciation characteristics. Forged authenticated voice attack refers to the attack by forging the voice of the victim through synthesis, conversion, splicing and other technical means.

Speaker spoofing attacks require the attacker to have good imitation ability, and fake authentication voice attacks often require high professional skills. Is the attack of these two attacks high? In addition, whether it is imitation voice or fake voice, it is not real voice after all. , the existing voiceprint recognition technology can basically deal with these two types of attacks.

Recording playback attack is a very important problem in voiceprint recognition. The attacker obtains the voice and then uses software synthesis to attack. There are two types of recording attacks. One is that the user's voice is stolen in other circumstances to attack; the other is that the user uses malware to record the user's voice during voiceprint recognition.

For recording attacks, in the prior art, there are mainly the following two solutions:

The first solution is to distinguish whether it is the recording content by analyzing the difference in the channel characteristic mode between the recording and the original voice; the second solution is to verify the speaker's speech content while verifying the speaker's voiceprint, Because the recording attacker does not know the content of this speech.

However, the solution has high requirements on sound signal quality, signal-to-noise ratio, and channel quality, and the effect achieved in practical applications is not very good.

In Scheme 2, if the user is randomly asked to read and write a large piece of text each time, the user experience is poor. If the user’s voice input is reduced, such as a patent (application number: 201310123555.0; invention name: identity confirmation system and method based on dynamic password voice) , select a combination from 26 English letters and 10 numbers, and after each random combination produces a dynamic password, let the user input by voice. Since the dynamic password produced each time is not known in advance, it can resist simple recording attacks. is a better solution. However, since the patent only randomly combines 26 English letters and 10 numbers with a total of 36 characters, if the attacker separates the 36 characters by recording, no matter what kind of random string is obtained, the attacker only needs to Simply attack by concatenating 36 characters.

Contents of the invention

The invention provides a voiceprint authentication method, a server and a terminal with the function of preventing recording attacks, which are used to solve the defects that there are loopholes in the methods for preventing recording attacks in the prior art and cannot effectively prevent recording attacks.

In order to solve the above technical problems, the present invention provides a voiceprint authentication method capable of preventing recording attacks,

Generate character combinations and character pronunciation rules according to a user's voiceprint authentication request;

Send the pronunciation rules of the character combination and characters to the requesting terminal;

receiving the user voice input by the requesting terminal according to the pronunciation rules of the character combination and characters;

Perform voiceprint authentication according to the user's voice, the character combination and the pronunciation rules of the characters;

Send the voiceprint authentication result to the requesting terminal.

The present invention also provides a voiceprint authentication method capable of preventing recording attacks,

Send a user's voiceprint authentication request to the server;

receiving and displaying the character combination and the pronunciation rules of the characters sent by the server;

receiving the user's voice input by the user according to the pronunciation rules of the character combination and characters;

sending the user voice to the server;

Receive the voiceprint authentication result sent by the server.

The present invention also provides a voiceprint authentication server capable of preventing recording,

A generating unit, configured to generate character combinations and character pronunciation rules according to a user's request;

A sending unit, configured to send the character combination and the pronunciation rules of the characters to the requesting terminal, and send the voiceprint authentication result to the requesting terminal;

a receiving unit, configured to receive the user voice input by the requesting terminal according to the character combination and character pronunciation rules;

A sound detection unit, configured to perform voiceprint authentication according to the user's voice, the character combination and the pronunciation rules of the characters;

The present invention also provides a voiceprint authentication terminal capable of preventing recording attacks,

A request unit, configured to send a user's voiceprint authentication request to the server;

The receiving unit is used to receive and display the character combination and character pronunciation rules sent by the server, and receive the voiceprint authentication result sent by the server;

The input unit is used to receive the user's voice input by the user according to the pronunciation rules of the character combination and characters;

A sending unit, configured to send the user voice to the server.

The present invention further provides a voiceprint authentication system capable of preventing recording attacks. The system includes a server and a requesting terminal, wherein the server is used to generate character combinations and character pronunciation rules according to a user's voiceprint authentication request; The character combination and the pronunciation rules of the character are sent to the requesting terminal; the user voice input by the requesting terminal according to the character combination and the pronunciation rule of the character is received; the pronunciation is performed according to the user voice, the character combination and the pronunciation rule of the character fingerprint authentication; sending the voiceprint authentication result to the requesting terminal;

The request terminal is used to send a user's voiceprint authentication request to the server; receive and display the character combination and the pronunciation rules of the characters sent by the server; receive the user voice input by the user according to the character combination and the pronunciation rules of the characters; Sending the user voice to the server; receiving the voiceprint authentication result sent by the server.

The voiceprint authentication method, server, terminal and system proposed by the present invention, which can prevent recording attacks, can effectively prevent recording attacks by verifying whether the characters and pronunciation methods in the user's voice are consistent with the character combinations and character pronunciation rules generated by the server , even if the attacker can obtain the user's voice through other channels to meet the content of the voice, it cannot meet the requirements of the pronunciation method. Further, in order to prevent the user voice repeatedly input by the user from being attacked by recording, after judging that the characters and pronunciation methods in the user voice are consistent with the character combinations and character pronunciation rules generated by the server, it is also judged that the current voice to be verified is consistent with the historical voice database Check whether the voice of the user is consistent in the user. If they are consistent, it means that there is a recording attack. The invention can effectively prevent recording attacks in voiceprint authentication.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

Fig. 1 is a flow chart of a voiceprint authentication method capable of preventing recording attacks according to an embodiment of the present invention;

Fig. 2 is a flow chart of the voiceprint authentication process capable of preventing recording attacks according to an embodiment of the present invention;

Fig. 3 is a flow chart of the voiceprint authentication process capable of preventing recording attacks according to an embodiment of the present invention;

FIG. 4 is a waveform diagram corresponding to the pronunciation of the number "0" according to an embodiment of the present invention;

5 is a flowchart of a voiceprint authentication method capable of preventing recording attacks according to an embodiment of the present invention;

Fig. 6 is a voiceprint authentication server capable of preventing recording attacks according to an embodiment of the present invention;

FIG. 7 is a voiceprint authentication terminal capable of preventing recording attacks according to an embodiment of the present invention;

Fig. 8 is a voiceprint authentication system capable of preventing recording attacks according to an embodiment of the present invention;

FIG. 9 is a flowchart of a voiceprint authentication method with the function of preventing recording attacks according to an embodiment of the present invention.

detailed description

In order to make the technical features and effects of the present invention more obvious, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings. The present invention can also be described or implemented in other different specific examples. The equivalent transformations done within all belong to the protection category of the present invention.

As shown in FIG. 1 , FIG. 1 is a flowchart of a voiceprint authentication method capable of preventing recording attacks according to an embodiment of the present invention.

This embodiment is a voiceprint authentication method described from the server side. The voiceprint authentication is performed according to the user voice fed back by the terminal, the character combination generated by the server, and the character pronunciation rules. This embodiment can prevent recording attacks to a certain extent.

Specifically, the voiceprint authentication method capable of preventing recording attacks includes the following steps:

Step 101: Generate character combinations and character pronunciation rules according to a user's voiceprint authentication request;

Character combinations include but are not limited to letters, numbers, Chinese characters, etc., and the pronunciation rules of characters include but are not limited to the tone of pronunciation, the length of pronunciation, etc. In one embodiment, each character in the character combination corresponds to a pronunciation rule, another implementation In the example, two characters in the character combination correspond to one pronunciation rule, and the present invention does not limit the specific form of the character combination and the pronunciation rules of the characters in the character combination.

In an embodiment of the present application, the character combinations and the pronunciation rules of the characters are randomly generated.

Step 102: Send the character combination and character pronunciation rules to the requesting terminal;

The terminals described in the present invention include but are not limited to mobile phones, PADs, computers and notebooks.

Step 103: receiving the user voice input by the requesting terminal according to the character combination and character pronunciation rules;

Step 104: Perform voiceprint authentication according to the user's voice, the character combination and the pronunciation rules of the characters;

Step 105: Send the voiceprint authentication result to the requesting terminal.

In this embodiment, even if the attacker can obtain phonetic character information, he cannot obtain the pronunciation rules of the characters. By adding the authentication of the pronunciation rules, recording attacks can be effectively prevented.

In detail, step 104 further includes:

Judging whether the voice of the user and the voice input by the user in history are the voice of the same person;

judging whether the characters in the user's voice are the same as the characters in the character combination;

Judging whether the pronunciation mode of the character in the user's voice matches the pronunciation rule of the character;

Only the voice of the user and the voice input by the user are the same person, the characters in the voice of the user are the same as the characters in the character combination, and the pronunciation of the characters in the voice of the user is the same as the pronunciation of the characters The voiceprint authentication will only pass when the rules match and are satisfied at the same time. In other cases, the voiceprint authentication will not pass, that is, if the user's voice and the user's historical input voice are not the same person, and/or the characters in the user's voice are not the same as If the characters in the character combination are different, and/or the pronunciation of the characters in the user's voice does not match the pronunciation rules of the characters, the voiceprint authentication fails.

The present invention does not limit the order of the above-mentioned judging process, any combination of order can realize the judgment of voiceprint authentication.

Preferably, as shown in Figure 2, step 104 further includes:

Step 201: first judge whether the user voice and the voice input by the user are from the same person; if they are not from the same person, the voiceprint authentication fails, and if they are from the same person, continue to step 202;

During specific implementation, before performing step 202, the user voice sent by the client needs to be separated according to characters, and then the characters in the user voice are extracted.

Step 202: judging whether the characters in the user's voice are the same as the characters in the character combination;

If the characters in the user's voice are different from the characters in the character combination, the voiceprint authentication fails, that is, the voiceprint authentication fails;

If the characters in the user's voice are the same as the characters in the character combination, proceed to step 203;

Step 203: judging whether the pronunciation mode of the character in the user's voice matches the pronunciation rule of the character;

If the pronunciation of the characters in the user's voice does not match the pronunciation rules of the characters, the voiceprint authentication fails;

If the pronunciation mode of the character in the user's voice matches the pronunciation rule of the character, the voiceprint authentication is passed.

Performing voiceprint authentication according to the sequence described in this embodiment can speed up authentication, prevent recording attacks and improve user experience. In the following embodiments, unless otherwise specified, voiceprint authentication is performed in accordance with the sequence described in this embodiment.

Referring again to Fig. 2, it is judged that the voice of the user and the voice input by the user history are the same person, the characters in the voice of the user are the same as the characters in the character combination, and the pronunciation of the characters in the voice of the user is the same After the method is matched with the pronunciation rules of the characters, it also includes storing the user's voice in the historical voice database, so as to facilitate the subsequent retrieval of the voice information input by the user.

As shown in Figure 3, in one embodiment of the present application, it is determined that the user voice and the voice input by the user are from the same person, the characters in the user voice are the same as the characters in the character combination, and the After the pronunciation of the character in the user's voice matches the pronunciation rule of the character, it also includes:

Step 204: judging whether the user's voice is consistent with the user's voice in the historical voice library;

If the user's voice is consistent with the user's voice in the historical voice library, the voiceprint authentication fails;

If the user's voice is inconsistent with the user's voice in the historical voice database, the voiceprint authentication is passed, and the user's voice is stored in the historical voice database.

By verifying whether the user's voice is consistent with the user's voice in the historical voice database, it is possible to prevent recording attacks on the same user's voice input in different voice authentications of the same user.

In an embodiment of the present invention, step 204 of the previous embodiment further includes:

extracting feature parameters of the user's voice;

calculating the Euclidean distance between the characteristic parameters of the user's speech and the characteristic parameters of the user's speech in the historical database, and when the Euclidean distance is less than a predetermined threshold, the user's speech and the user's historical The speech in the speech database is consistent, and when the Euclidean distance is greater than a predetermined threshold, the user's speech is inconsistent with the user's speech in the historical speech database.

The predetermined threshold described in this embodiment can be determined according to the difference of the same sound made by people.

During specific implementation, the detailed process of judging whether the user's voice is consistent with the voice of the user in the historical voice bank is:

1) Divide the user's speech into multiple speeches according to characters, and perform preprocessing on each speech, including processing such as framing, pre-emphasis, and windowing, to obtain a section of sound that can be further calculated.

2) Find the start and end points of the valid speech parts in each speech.

As shown in Figure 4, Figure 4 is a waveform diagram corresponding to the pronunciation of the number "0". It can be seen from Figure 4 that there are many silent segments or subtle noise segments before and after the sound. If these invalid sound signals are not removed, the attacker can process the invalid sound end of the recording and affect the effect of recording detection.

During specific implementation, the start and end points of the effective part of the speech can be judged by the short-term energy and the short-term zero-crossing rate.

The short-term energy refers to the sum of the strength of a frame of speech signal, and the short-term energy En of the nth frame of speech signal:

${\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

Wherein, m is the mth sampling point of the nth frame, N is the size of the frame, and x _n (m) is the normalized frequency of the mth sampling point of the nth frame.

The short-term zero-crossing rate refers to the number of times that the voice signal waveform crosses the horizontal axis in one frame, which is denoted as Z _n ,

${\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&lsqb;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&lsqb;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; |</span>$

Wherein, m is the mth sampling point of the nth frame, N is the size of the frame, and x _n (m) is the normalized frequency of the mth sampling point of the nth frame.

When the short-term energy En exceeds the threshold E or the short-term zero-crossing rate Zn exceeds the threshold Z, the speech is the beginning of a valid speech; when the short-term energy En is lower than the threshold E or the short-term zero-crossing rate Zn is lower than the threshold When the value is Z, the speech is the end of the valid speech.

3) Using Mel scale cepstral coefficients (MFCC) to extract feature parameters for effective speech. This method is a relatively common feature parameter extraction method in sound processing at present, and will not be repeated here in the present invention.

Record the user's voice of a character after the first three steps of preprocessing, segmenting the invalid part of the voice, and extracting feature parameters this time:

T has N frame vectors {T(1), T(2), ... T(n), ..., T(N)}, where T(n) is the speech feature vector of the nth frame.

Perform the same preprocessing for the user's character pronunciation in the historical database, segment the invalid part of the voice and extract the feature parameters, and denote it as R:

R has M frame vectors R={R(1), R(2),...R(m),...,R(M)}, and R(m) is the speech feature vector of the mth frame.

4) Calculating the similarity between the user's voice and the voice stored in the historical voice database, that is, calculating the similarity between T and R, which can be judged by calculating the Euclidean distance between T and R.

d(T(i _n ),R(i _m )) represents the Euclidean distance between the i _nth frame feature in T and the i _m frame feature in R, if the two waveforms completely overlap in a certain frame, then The distance d is 0. In order to compare the similarity between them, the distance D[T,R] between them can be calculated, the smaller the distance, the higher the similarity.

If N=M, that is, the length of the two speeches is the same, directly and simply calculate the Euclidean distance D[T, R]=d(1,1)+d(2,2) between the user's speech and the speech stored in the historical speech database )+…+d(N,N), if the voices at both ends are exactly the same, then D[T,R]=0, in this way only T and R can be judged whether they are exactly the same, but the recording attacker will The original recording is often stretched, shortened, or deleted in some positions, so simply calculating the distance between the two is not a good defense against such attacks.

When N and M are different, consider aligning T(n) and R(m). Alignment can use the method of linear expansion, if N<M, T can be linearly mapped to a sequence of M frames, and then calculate the relationship between it and {R(1), R(2),...,R(M)} distance. However, the attacker will not process the entire sound, but only part of the sound. If this method is adopted, it will be recognized that the similarity between the two sounds is very low.

Therefore, comparing the similarity of speech T and R needs to combine the time rule and distance measurement. By finding the function i _m = Φ(i _n ), the time axis n of T is nonlinearly mapped to the time axis m of R, and Make the distance D[T,R] between T and R satisfy:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}}\underset{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Phi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

in:

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; \&Phi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; \&Phi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\end{array}\right.$

Φ(i _n +1)≥Φ(i _n )

Φ(i _n +1)-Φ(i _n )≤1

It can be seen that the condition of dynamic programming is obviously satisfied, and the dynamic programming algorithm can be used to solve it, where the dynamic programming polynomial is:

D(in,im)=d(T(in),R(im))+min{D(in-1,im),D(in-1,im-1),D(in-1,im- 2)}

In this way, starting from point (l, 1) (let D(1, 1) = 0) search, and repeat recursion until (N, M) can get the optimal path, and D (N, M) is the best match The matching distance corresponding to the path.

Since each person's speech is affected by many factors, it is impossible for anyone to repeat the same character's voice to be completely similar in sound waves, and there must be differences. This difference is defined as the predetermined threshold for judgment. If D(N,M)=0, it means that the voices T and R at both ends are completely consistent, and it can be proved that the voices T and R are one voice, and there may be a recording attack; if D(N,M)<threshold, it means The voices T and R at both ends are very similar, and there may also be a recording attack; if D(N,M)>=threshold, it means that T and R are not the same voice, and there is no recording attack.

The voiceprint authentication method proposed by the present invention that can prevent recording attacks can effectively prevent recording attacks by verifying whether the characters and pronunciation methods in the user's voice are consistent with the character combinations and character pronunciation rules generated by the server. The user's voice obtained through other channels satisfies the content of the voice, but also fails to meet the requirements of the pronunciation method. Further, in order to prevent the user voice repeatedly input by the user from being attacked by recording, after judging that the characters and pronunciation methods in the user voice are consistent with the character combinations and character pronunciation rules generated by the server, it is also judged that the current voice to be verified is consistent with the historical voice database Check whether the voice of the user is consistent in the user. If they are consistent, it means that there is a recording attack. The invention can effectively prevent recording attacks in voiceprint authentication.

As shown in FIG. 5 , FIG. 5 is a flowchart of a voiceprint authentication method capable of preventing recording attacks according to an embodiment of the present invention. This method is described from the side of the requesting terminal. Specifically, the voiceprint authentication method includes:

Step 501: Send a user's voiceprint authentication request to the server;

Step 502: receiving and displaying the character combination and character pronunciation rules sent by the server;

Step 503: receiving the user's voice input by the user according to the character combination and character pronunciation rules;

Step 504: Send the user voice to the server;

Step 505: Receive the voiceprint authentication result sent by the server.

As shown in Figure 6, Figure 6 is a voiceprint authentication server capable of preventing recording attacks according to an embodiment of the present invention, the server 600 includes a generation unit 601, which is used to generate character combinations and pronunciation of characters according to a user's request rule;

A sending unit 602, configured to send the character combination and pronunciation rules of the characters to the requesting terminal, and send the voiceprint authentication result to the requesting terminal;

The receiving unit 603 is configured to receive the user voice input by the requesting terminal according to the character combination and character pronunciation rules;

The sound detection unit 604 is configured to perform voiceprint authentication according to the user's voice, the character combination and the pronunciation rules of the characters.

As shown in FIG. 7, FIG. 7 is a voiceprint authentication terminal capable of preventing recording attacks according to an embodiment of the present invention. Specifically, the authentication terminal 700 includes: a request unit 701, configured to send a user's voiceprint authentication request to the server;

The receiving unit 702 is configured to receive and display the character combination and character pronunciation rules sent by the server, and receive the voiceprint authentication result sent by the server;

The input unit 703 is used to receive the user's voice input by the user according to the pronunciation rules of the character combination and characters;

A sending unit 704, configured to send the user voice to the server.

As shown in FIG. 8 , FIG. 8 is a voiceprint authentication system capable of preventing recording attacks according to an embodiment of the present invention.

The voiceprint authentication system includes a server 600 and a request terminal 700, wherein the server 600 is used to generate character combinations and pronunciation rules of characters according to a user's voiceprint authentication request; send the character combination and pronunciation rules of characters to Requesting terminal; receiving the user voice input by the requesting terminal according to the character combination and the pronunciation rules of characters; performing voiceprint authentication according to the user voice, the character combination and the pronunciation rules of characters; sent to the requesting terminal;

The request terminal 700 is used to send a user's voiceprint authentication request to the server; receive and display the character combination and character pronunciation rules sent by the server; receive the user voice input by the user according to the character combination and character pronunciation rules ; Sending the user voice to the server; receiving the voiceprint authentication result sent by the server.

The voiceprint authentication method, server, terminal and system proposed by the present invention, which can prevent recording attacks, can effectively prevent recording attacks by verifying whether the characters and pronunciation methods in the user's voice are consistent with the character combinations and character pronunciation rules generated by the server , even if the attacker can obtain the user's voice through other channels to meet the content of the voice, it cannot meet the requirements of the pronunciation method. Further, in order to prevent the user voice repeatedly input by the user from being attacked by recording, after judging that the characters and pronunciation methods in the user voice are consistent with the character combinations and character pronunciation rules generated by the server, it is also judged that the current voice to be verified is consistent with the historical voice database Check whether the voice of the user is consistent in the user. If they are consistent, it means that there is a recording attack. The invention can effectively prevent recording attacks in voiceprint authentication.

In order to illustrate the technical solution of the present application more clearly, a specific embodiment is used below to illustrate, and as shown in FIG. 9 , the system workflow for preventing recording attacks is as follows:

Step 901: the client sends an identity authentication request to the server;

Step 902: the server receives the identity authentication request;

Step 903: The server randomly generates a verification character combination and the pronunciation of the characters according to the identity authentication request, and sends them to the client;

Step 904: After the client receives the character combination to be verified and the pronunciation rules of the characters issued by the server, it prompts the user to read the characters as required;

Step 905: the client receives the user voice read in by the user, and sends the user voice read in by the user to the server;

Step 906: The server performs voiceprint verification, and judges whether the received user's voice and the pre-stored user's voice are the same person, and the current conventional voiceprint verification algorithm can be used for specific implementation;

If the voiceprint verification is not the same person, it will directly return the user authentication failure to the client;

If the voiceprint is verified as the same person, continue recording detection;

Step 907: Verify whether the characters in the user voice are the same as the characters in the character combination generated by the server; if the characters in the user voice are not the same as the characters in the character combination generated by the server, then the verification of the characters in the user voice fails, and returns If the user authentication fails to the client; if the characters in the user voice are identical to the characters in the character combination generated by the server, the character verification in the user voice is passed, and continue to step 908;

Step 908: Verify that the pronunciation of the characters in the user's voice is the same as the pronunciation of the characters generated by the server, if the pronunciation of the characters in the user's voice is different from the pronunciation of the characters generated by the server, verify the pronunciation of the characters in the user's voice If not, return the user authentication failure to the client; if the pronunciation mode of the character in the user voice is identical to the pronunciation mode of the character generated by the server, the verification of the pronunciation mode of the character in the user voice is passed, and proceed to step 909;

Step 909: Verify whether the user's voice exists in the historical voice database, if it exists, it proves that there is a recording attack, the authentication fails, and the authentication failure result is sent to the client; if it does not exist, the voiceprint authentication passes, and the user's voice is stored in In the historical voice database, the voiceprint authentication pass result is sent to the client.

The process of verifying whether the user's voice exists in the historical speech database has been described in detail in the above embodiments, and will not be repeated here. After the voiceprint authentication is passed, the client continues corresponding operations, which is not limited in the present invention.

The voiceprint authentication method, server, terminal and system proposed by the present invention, which can prevent recording attacks, can effectively prevent recording attacks by verifying whether the characters and pronunciation methods in the user's voice are consistent with the character combinations and character pronunciation rules generated by the server , even if the attacker can obtain the user's voice through other channels to meet the content of the voice, it cannot meet the requirements of the pronunciation method. Further, in order to prevent the user voice repeatedly input by the user from being attacked by recording, after judging that the characters and pronunciation methods in the user voice are consistent with the character combinations and character pronunciation rules generated by the server, it is also judged that the current voice to be verified is consistent with the historical voice database Check whether the voice of the user is consistent in the user. If they are consistent, it means that there is a recording attack. The invention can effectively prevent recording attacks in voiceprint authentication.

The above description is only used to illustrate the technical solutions of the present application, and any person skilled in the art can modify and change the above embodiments without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be determined by the scope of the claims.

Claims (11)

1. A voiceprint authentication method capable of preventing recording attacks, characterized in that, comprising, Generate character combinations and character pronunciation rules according to a user's voiceprint authentication request; Send the pronunciation rules of the character combination and characters to the requesting terminal; receiving the user voice input by the requesting terminal according to the pronunciation rules of the character combination and characters; performing voiceprint authentication according to the user's voice, the character combination, and the pronunciation rules of the characters; and sending the voiceprint authentication result to the requesting terminal. 2. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 1, wherein performing voiceprint authentication according to the user voice, the character combination and the pronunciation rules of the characters further comprises, Judging whether the voice of the user and the voice input by the user in history are the voice of the same person; judging whether the characters in the user's voice are the same as the characters in the character combination; Judging whether the pronunciation mode of the character in the user's voice matches the pronunciation rule of the character; Only the voice of the user and the voice input by the user are the same person, the characters in the voice of the user are the same as the characters in the character combination, and the pronunciation of the characters in the voice of the user is the same as the pronunciation of the characters Only when the rules match, the voiceprint authentication will pass, otherwise the voiceprint authentication will not pass. 3. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 2, wherein it is determined that the user's voice and the voice input by the user in history are the same person, and the characters in the user's voice are the same as the voice of the user. After the characters in the character combination are the same and the pronunciation of the characters in the user voice matches the pronunciation rules of the characters, it also includes, The user's voice is stored in a historical voice database. 4. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 2, wherein it is determined that the user's voice and the voice input by the user are from the same person, and the characters in the user's voice are the same as the user's voice. After the characters in the character combination are the same and the pronunciation of the characters in the user voice matches the pronunciation rules of the characters, it also includes, Judging whether the voice of the user is consistent with the voice of the user in the historical voice library; If the user's voice is consistent with the user's voice in the historical voice library, the voiceprint authentication fails; If the user's voice is inconsistent with the user's voice in the historical voice database, the voiceprint authentication is passed, and the user's voice is stored in the historical voice database. 5. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 4, wherein judging whether the user's voice is consistent with the user's voice in the historical voice database further comprises, extracting feature parameters of the user's voice; calculating the Euclidean distance between the characteristic parameters of the user's speech and the characteristic parameters of the user's speech in the historical database, and when the Euclidean distance is less than a predetermined threshold, the user's speech and the user's historical The speech in the speech database is consistent, and when the Euclidean distance is greater than a predetermined threshold, the user's speech is inconsistent with the user's speech in the historical speech database. 6. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 5, wherein extracting the characteristic parameters of the user's voice further comprises, Preprocessing the user's voice, dividing the user's voice into multiple segments of voice by character; find the start and end points of valid speech parts in each speech; Extract the feature parameters of the effective speech part. 7. The voiceprint authentication method capable of preventing recording attacks as claimed in claim 1, wherein the character combination and the pronunciation rules of the characters are randomly generated. 8. A voiceprint authentication method capable of preventing recording attacks, comprising: Send a user's voiceprint authentication request to the server; receiving and displaying the character combination and the pronunciation rules of the characters sent by the server; receiving the user's voice input by the user according to the pronunciation rules of the character combination and characters; sending the user voice to the server; Receive the voiceprint authentication result sent by the server. 9. A voiceprint authentication server capable of preventing recording attacks, comprising: A generating unit, configured to generate character combinations and character pronunciation rules according to a user's request; A sending unit, configured to send the character combination and the pronunciation rules of the characters to the requesting terminal, and send the voiceprint authentication result to the requesting terminal; a receiving unit, configured to receive the user voice input by the requesting terminal according to the character combination and character pronunciation rules; The voice detection unit is configured to perform voiceprint authentication according to the user's voice, the character combination and the pronunciation rules of the characters. 10. A voiceprint authentication terminal capable of preventing recording attacks, characterized in that it comprises, A request unit, configured to send a user's voiceprint authentication request to the server; The receiving unit is used to receive and display the character combination and character pronunciation rules sent by the server, and receive the voiceprint authentication result sent by the server; The input unit is used to receive the user's voice input by the user according to the pronunciation rules of the character combination and characters; A sending unit, configured to send the user voice to the server. 11. A voiceprint authentication system capable of preventing recording attacks, characterized in that it includes a server and a requesting terminal, wherein the server is used to generate character combinations and pronunciation rules for characters according to a user's voiceprint authentication request; The character combination and the pronunciation rules of the character are sent to the requesting terminal; the user voice input by the requesting terminal according to the character combination and the pronunciation rule of the character is received; the pronunciation is performed according to the user voice, the character combination and the pronunciation rule of the character fingerprint authentication; sending the voiceprint authentication result to the requesting terminal; The request terminal is used to send a user's voiceprint authentication request to the server; receive and display the character combination and the pronunciation rules of the characters sent by the server; receive the user voice input by the user according to the character combination and the pronunciation rules of the characters; Sending the user voice to the server; receiving the voiceprint authentication result sent by the server.