CN109492385A - A kind of method for generating cipher code based on deep learning - Google Patents
A kind of method for generating cipher code based on deep learning Download PDFInfo
- Publication number
- CN109492385A CN109492385A CN201811308670.4A CN201811308670A CN109492385A CN 109492385 A CN109492385 A CN 109492385A CN 201811308670 A CN201811308670 A CN 201811308670A CN 109492385 A CN109492385 A CN 109492385A
- Authority
- CN
- China
- Prior art keywords
- password
- generator
- data
- determining device
- initial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/45—Structures or tools for the administration of authentication
- G06F21/46—Structures or tools for the administration of authentication by designing passwords or checking the strength of passwords
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Storage Device Security (AREA)
Abstract
The invention discloses a kind of method for generating cipher code based on deep learning, 1) construct initial password generator and initial password determining device;2) data prediction is carried out to user data;3) pretreated user data will be passed through, inputted in initial password generator, initial password is generated;4) password generator receives the true password of user, in the initial password input password determining device that step 3) is generated, password determining device differentiates true password and initial password, distinguishing results is fed back to password generator, password generator carries out study optimization according to feedback result;5) step 2- step 4) is repeated, until password generator study is optimal;6) password generator of optimization is generated into password, as final password;This method generates password that is complicated and meeting safety standard using the feature for the generator and arbiter for generating sex resistance network, and generating mode is flexible, and decoding difficulty is big, can provide different passwords and select for user, password is easily remembered.
Description
Technical field
The present invention relates to artificial intelligence and depth learning technology field, specifically a kind of password based on deep learning is generated
Method.
Background technique
With flourishing for internet, user increases sharply, and information produced by user surfs the Internet constantly increases, this is big number
According to research and development provide huge data, data rapidly increase with increasing for user, in many SCN subscriber connection networks,
When wishing to carry out some specific service routines on the internet, need to apply corresponding account and password, due to existing
Code breaking technology obtain continuous breakthrough, in order to protect the account number safety of user, the complexity of account application is also continuous
It improves, this allows for user, and originally known password loses effect.One kind how is generated for user and meets safety standard, decodes
Difficulty increases, and the cry for being easy the password recommended program learnt by heart by user is growing.
With flourishing for all kinds of Internet applications in the market, some traditional cryptoguard technologies are growing
Seem under the attack of hacker and be easy to crack easily, some passwords that user can remember before this is resulted in do not comply with web application
Registration requirement, under new internet operation condition, who preferably can provide better service for user, can first occupy elder generation
Machine.
Currently, no one of prior art completely generates the integrity techniques of password according to user personality, but it is random
The random string of a string of difficult memories is generated, what is be unable to fully makes good use of user data.This method is by training a password
Generator generates password, and principle generates network based on the Gan in artificial intelligence direction, there is good technical support and flexibility.
Summary of the invention
It is an object of the invention to customer service the deficiencies in the prior art, and provide a kind of password generation side based on deep learning
Method, this method generate complicated and meet the close of safety standard using the feature for the generator and arbiter for generating sex resistance network
Code, generating mode is flexible, and decoding difficulty is big, can provide different passwords and select for user, and password is easy to be remembered by user, is
A kind of preferable password generating mode.
Realizing the technical solution of the object of the invention is:
A kind of method for generating cipher code based on deep learning, includes the following steps:
1) initial password generator and initial password determining device are constructed;
2) data prediction is carried out to user data;
3) pretreated user data will be passed through, inputted in initial password generator, initial password is generated;
4) password generator receives the true password of user, and the initial password that step 3) is generated inputs password determining device
In, password determining device differentiates true password and initial password, distinguishing results are fed back into password generator, password generator according to
Feedback result carries out study optimization;
5) step 2), step 3) and step 4) are repeated, until password generator study is optimal;
6) password generator of optimization is generated into password, as final password;
By above-mentioned steps, the generation of password is completed.
The password generator generates initial password according to user data, initial password is inputted in password determining device
Repetition learning and training;The training is carried out using RNN Recognition with Recurrent Neural Network model, is added in RNN Recognition with Recurrent Neural Network model
Enter length memory lstm module, the output end of lstm module connects full articulamentum, and for providing the output of data, it is raw to complete password
At the building of structure.
The generator is specifically configured to using full articulamentum as the data receiver layer of the bottom, for receiving user
Data are laid with upper lstm module after full connection, and the effect of this section of parameter layer is to receive the semantic information of full articulamentum, study
The parameter attribute of front and back, subsequent full articulamentum is output layer, per one-dimensional output one-bit digital, represents character string ID, will
It is the password being initially generated that this string character string ID, which combines,.
The password determining device generates result as one and judges score, exercises supervision to password generator, receive first
The true password of user, distinguishes true password, then judges the password of the input of password generator, by password
The password of generator generates result and feeds back to password generator, makes the continuous regularized learning algorithm optimization of generator, and password determining device is certainly
Body also constantly study optimization, makes password generator and password determining device be optimal state, password generator generates optimal password.
In step 2), the data prediction, including cleaning data, semi-structured, non-structured data structured
And data normalization;
The cleaning data, be crawler is crawled to get off from user data it is disorderly and unsystematic, useless, need to define
The content of regular expression is removed;
Semi-structured, the non-structured data structured is carried out to the user data after over cleaning data
Standardization;
The data normalization, by the mapping of obtained data after semi-structured, non-structured data structured
To [0-1] range, preferably learnt by neural network by normalized data.
In step 3), the generation of the initial password, initial password includes the following steps:
3-1) input data entrance: the data of N-dimensional dimension, input password to generation will be corresponded to after step 2)
The entrance of device has been added to what the data of user structure had no to lose in generator;
3-1) parameter operation: after password generator receives data, the inside of internal RNN and lstm include
Parameter carries out operation to the data in input password generator, and the parameter weight of hidden layer every time can be with judgement in generator
The feedback of device reaches variation, and the parameter that inside includes and the data that input is come in carry out operation;
3-3) password generates: after the parameter operation by step 3-2), password generator is raw in last full articulamentum
At the digital vectors of N-dimensional, it is the ID of character string that the digital vectors are corresponding, and that each number corresponds to is the ID of character string,
All numbers are converted to character string, just generate initial password.
The step 4) specifically comprises the following steps:
4-1) receive truthful data password: the true code data of user is sent in password arbiter, password arbiter
The code data for identifying real user, its feature is remembered by intrinsic nerve network parameter;
4-2) receive generator and generate password: in the Password Input password determining device that password generator is generated, password is sentenced
Whether the password that the neural network Parameter analysis password generator inside disconnected device generates meets actual user data feature;
4-3) study optimization password determining device: when the password that password generator generates cheats password determining device, by password
Determining device optimizes, until password determining device is restrained;
4-4) will generate result and feed back to generator: password determining device obtains password generator generation by parameter operation
Password as a result, the result is fed back in password generator.
The step 6), specifically when password generator and password determining device restrain, password generator is generated at this time
Password it is optimal, using the password as final password.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of password generator;
Fig. 2 is RNN neural network schematic diagram;
Fig. 3 is schematic diagram of the length memory module in conjunction with RNN.
Specific embodiment
The present invention is further elaborated with reference to the accompanying drawings and examples, but is not limitation of the invention.
A kind of method for generating cipher code based on deep learning, includes the following steps:
1) initial password generator and initial password determining device are constructed;
2) data prediction is carried out to user data;
3) pretreated user data will be passed through, inputted in initial password generator, initial password is generated;
4) password generator receives the true password of user, and the initial password that step 3) is generated inputs password determining device
In, password determining device differentiates true password and initial password, distinguishing results are fed back into password generator, password generator according to
Feedback result carries out study optimization;
5) step 2), step 3) and step 4) are repeated, until password generator study is optimal;
6) password generator of optimization is generated into password, as final password;
By above-mentioned steps, the generation of password is completed.
The password generator, be building one it is initial, without the password generator of any study, this generator
Some coarse passwords are initially generated according to user data, the password with user data is not close to just according to user data generation
Initial password is inputted repetition learning and training in password determining device by beginning password;The training, using RNN as shown in Figure 2
Recognition with Recurrent Neural Network model carries out, and RNN Recognition with Recurrent Neural Network model can be very good processing sequence model, can completely learn
Front and back is semantic, and correlative connection is remained, in order to solve the problems, such as the prolonged semantic loss of RNN, RNN circulation mind
Length is added in network model and remembers lstm module, length is remembered just as man memory power, and memory is limited, and is had
The memory of limit can not remember whole semantic informations, and the output end of lstm module connects full articulamentum, as shown in figure 3, being used for
It provides the output of data, completes the building that password generates structure.
The password generator is specifically configured to using full articulamentum as the data receiver layer of the bottom, for receiving
User data is laid with upper lstm module after full connection, and the effect of this section of parameter layer is to receive the semantic information of full articulamentum,
Learn the parameter attribute of front and back, subsequent full articulamentum is output layer, per one-dimensional output one-bit digital, represents character string
ID, it is the password being initially generated that this string character string ID, which is combined, and specific structure figures are as shown in Figure 1.
The mathematical formulae expression formula of RNN Recognition with Recurrent Neural Network model are as follows:
The password determining device generates result as one and judges score, exercises supervision to password generator, receive first
The true password of user, distinguishes true password, then judges the password of the input of password generator, by password
The password of generator generates result and feeds back to password generator, makes the continuous regularized learning algorithm optimization of generator, and password determining device is certainly
Body also constantly study optimization, makes password generator and password determining device be optimal state, password generator generates optimal password.
In step 2), the data prediction, including cleaning data, semi-structured, non-structured data structured
And data normalization;
The cleaning data, be crawler is crawled to get off from user data it is disorderly and unsystematic, useless, need to define
The content of regular expression is removed;
Semi-structured, the non-structured data structured is carried out to the user data after over cleaning data
Standardization;Such as the age: expression on computers in 9 years old and 19 years old is just 1 and 2, so needing to be converted to 9 herein
09;
The data normalization, by the mapping of obtained data after semi-structured, non-structured data structured
To [0,1] range, preferably learnt by neural network by normalized data.
In the present embodiment, four kinds of information are had chosen as experimental data: height, weight, date of birth and name.
Due to the particularity of data, height, weight, the data such as date of birth can be carried out defeated in the form of digit integer
Enter, it is only necessary to by these data be normalized to [0,1] section can, the formula used herein are as follows:
Xmin indicates that the smallest sample data in height sample, and Xmax indicates the highest sample data of height, X table
Show to be currently to need normalized sample data, Xnorm is then the sample data after being normalized, the weight of user with
And the date of birth can be normalized according to this formula.
The name of user, such as the pretreatment operation of ' Li Ming ' are as follows:
Chinese character ' Li Ming ' is converted to phonetic ' LiMing '.Since each letter has corresponding ASCII character table, in this way
Chinese character can be converted to number by son.Therefore, all pretreated data can be converted to the number of [0-1] from any character
Word, these numbers are exactly to be sent into the initial data feature of generator.
In step 3), the generation of the initial password, initial password includes the following steps:
3-1) input data entrance: the data of N-dimensional dimension, input password to generation will be corresponded to after step 2)
The entrance of device has been added to what the data of user structure had no to lose in generator;
3-1) parameter operation: after password generator receives data, the inside of internal RNN and lstm include
Parameter carries out operation to the data in input password generator, and generator hides layer parameter weight every time can be with determining device
Feedback reaches variation, and the parameter that inside includes and the data that input is come in carry out operation;
3-3) password generates: after the parameter operation by step 3-2), password generator is raw in last full articulamentum
At the digital vectors of N-dimensional, it is the ID of character string that the digital vectors are corresponding, and that each number corresponds to is the ID of character string,
All numbers are converted to character string, just generate initial password.
The step 4) specifically comprises the following steps:
4-1) receive truthful data password: the true code data of user is sent in password arbiter, password arbiter
The code data for identifying real user, its feature is remembered by intrinsic nerve network parameter;
4-2) receive generator and generate password: in the Password Input password determining device that password generator is generated, password is sentenced
Whether the password that the neural network Parameter analysis password generator inside disconnected device generates meets actual user data feature;
4-3) study optimization password determining device: when the password that password generator generates cheats password determining device, by password
Determining device optimizes, until password determining device is restrained;
4-4) will generate result and feed back to generator: password determining device obtains password generator generation by parameter operation
Password as a result, the result is fed back in password generator.
The step 6), specifically when password generator and password determining device restrain, password generator is generated at this time
Password it is optimal, using the password as final password.
Claims (8)
1. a kind of method for generating cipher code based on deep learning, which comprises the steps of:
1) initial password generator and initial password determining device are constructed;
2) data prediction is carried out to user data;
3) pretreated user data will be passed through, inputted in initial password generator, initial password is generated;
4) password generator receives the true password of user, and in the initial password input password determining device that step 3) is generated,
Password determining device differentiates true password and initial password, distinguishing results is fed back to password generator, password generator is according to anti-
Feedback result carries out study optimization;
5) step 2, step 3) and step 4) are repeated, until password generator study is optimal;
6) password generator of optimization is generated into password, as final password;
By above-mentioned steps, the generation of password is completed.
2. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that the password
Generator, generates initial password according to user data, and initial password is inputted repetition learning and training in password determining device;It is described
Training, carried out using RNN Recognition with Recurrent Neural Network model, length be added in RNN Recognition with Recurrent Neural Network model and remembers lstm module,
The output end of lstm module connects full articulamentum, for providing the output of data, completes the building that password generates structure.
3. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that the generation
Device is specifically configured to using full articulamentum as the data receiver layer of the bottom, for receiving user data, after full connection
Lstm module in laying, the effect of this section of parameter layer are to receive the semantic information of full articulamentum, learn the parameter attribute of front and back, after
Continuous full articulamentum is output layer, per one-dimensional output one-bit digital, represents character string ID, this string character string ID group is closed
Being the password being initially generated.
4. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that the password
Determining device generates result as one and judges score, exercises supervision to password generator, receives the true password of user first, right
True password is distinguished, and is then judged the password of the input of password generator, and the password of password generator is generated
As a result password generator is fed back to, the continuous regularized learning algorithm optimization of generator, and password determining device itself also constantly study optimization are made,
Password generator and password determining device is set to be optimal state, password generator generates optimal password.
5. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that in step 2,
The data prediction, including cleaning data, semi-structured, non-structured data structured and data normalization;
The cleaning data, be crawler is crawled to get off from user data it is disorderly and unsystematic, useless, need to define some canonicals
The content of expression formula is removed;
Semi-structured, the non-structured data structured is to carry out standard to the user data after over cleaning data
Change;
Obtained data after semi-structured, non-structured data structured are mapped to [0- by the data normalization
1] in range, preferably learnt by neural network by normalized data.
6. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that in step 3),
The generation of the initial password, initial password includes the following steps:
3-1) input data entrance: by the data for corresponding to N-dimensional dimension after step 2, password is inputted to generator
Entrance has been added to what the data of user structure had no to lose in generator;
3-1) parameter operation: after password generator receives data, parameter that the inside of internal RNN and lstm include
Operation is carried out to the data in input password generator, the parameter weight of hidden layer every time can be with determining device in generator
Feedback reaches variation, and the parameter that inside includes and the data that input is come in carry out operation;
3-3) password generates: after the parameter operation by step 3-2), password generator generates N-dimensional in last full articulamentum
Digital vectors, it is the ID of character string that the digital vectors are corresponding, and that each number corresponds to is the ID of character string, will be owned
Number be converted to character string, just generate initial password.
7. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that the step
4), specifically comprise the following steps:
4-1) receive truthful data password: the true code data of user is sent in password arbiter, the identification of password arbiter
The code data of real user remembers its feature by intrinsic nerve network parameter;
It 4-2) receives generator and generates password: in the Password Input password determining device that password generator is generated, password determining device
Whether the password that internal neural network Parameter analysis password generator generates meets actual user data feature;
4-3) study optimization password determining device: when the password that password generator generates cheats password determining device, password is judged
Device optimizes, until password determining device is restrained;
4-4) will generate result and feed back to generator: password determining device obtains password generator and generates password by parameter operation
As a result, the result is fed back in password generator.
8. a kind of method for generating cipher code based on deep learning according to claim 1, which is characterized in that the step
6), specifically when password generator and password determining device restrain, the password that password generator generates at this time is optimal, by the password
As final password.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811308670.4A CN109492385A (en) | 2018-11-05 | 2018-11-05 | A kind of method for generating cipher code based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811308670.4A CN109492385A (en) | 2018-11-05 | 2018-11-05 | A kind of method for generating cipher code based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109492385A true CN109492385A (en) | 2019-03-19 |
Family
ID=65693800
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811308670.4A Pending CN109492385A (en) | 2018-11-05 | 2018-11-05 | A kind of method for generating cipher code based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109492385A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112019354B (en) * | 2020-11-02 | 2021-02-09 | 之江实验室 | Password traversal device and method based on generating type countermeasure network |
CN112613027A (en) * | 2020-12-16 | 2021-04-06 | 广州岸边网络科技有限公司 | Multi-password management method, equipment and storage medium based on machine learning |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107122479A (en) * | 2017-05-03 | 2017-09-01 | 西安交通大学 | A kind of user cipher conjecture system based on deep learning |
CN107579821A (en) * | 2017-09-19 | 2018-01-12 | 厦门市美亚柏科信息股份有限公司 | Password dictionary generation method and computer-readable recording medium |
CN107947921A (en) * | 2017-11-22 | 2018-04-20 | 上海交通大学 | Based on recurrent neural network and the password of probability context-free grammar generation system |
CN108229138A (en) * | 2018-01-11 | 2018-06-29 | 浙江九州量子信息技术股份有限公司 | A kind of password dictionary generation method based on deep learning algorithm |
CN108509790A (en) * | 2018-03-14 | 2018-09-07 | 华东师范大学 | A kind of password strength assessment method based on group |
CN108734276A (en) * | 2018-04-28 | 2018-11-02 | 同济大学 | A kind of learning by imitation dialogue generation method generating network based on confrontation |
-
2018
- 2018-11-05 CN CN201811308670.4A patent/CN109492385A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107122479A (en) * | 2017-05-03 | 2017-09-01 | 西安交通大学 | A kind of user cipher conjecture system based on deep learning |
CN107579821A (en) * | 2017-09-19 | 2018-01-12 | 厦门市美亚柏科信息股份有限公司 | Password dictionary generation method and computer-readable recording medium |
CN107947921A (en) * | 2017-11-22 | 2018-04-20 | 上海交通大学 | Based on recurrent neural network and the password of probability context-free grammar generation system |
CN108229138A (en) * | 2018-01-11 | 2018-06-29 | 浙江九州量子信息技术股份有限公司 | A kind of password dictionary generation method based on deep learning algorithm |
CN108509790A (en) * | 2018-03-14 | 2018-09-07 | 华东师范大学 | A kind of password strength assessment method based on group |
CN108734276A (en) * | 2018-04-28 | 2018-11-02 | 同济大学 | A kind of learning by imitation dialogue generation method generating network based on confrontation |
Non-Patent Citations (1)
Title |
---|
BRILAND HITAJ.ETC: "PassGAN: A Deep Learning Approach for Password Guessing", 《HTTPS://ARXIV.ORG/ABS/1709.00440V》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112019354B (en) * | 2020-11-02 | 2021-02-09 | 之江实验室 | Password traversal device and method based on generating type countermeasure network |
CN112613027A (en) * | 2020-12-16 | 2021-04-06 | 广州岸边网络科技有限公司 | Multi-password management method, equipment and storage medium based on machine learning |
CN112613027B (en) * | 2020-12-16 | 2022-06-17 | 广州岸边网络科技有限公司 | Multi-password management method, device and storage medium based on machine learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108875807B (en) | Image description method based on multiple attention and multiple scales | |
US20190220514A1 (en) | Keyword extraction method, computer equipment and storage medium | |
CN108763284A (en) | A kind of question answering system implementation method based on deep learning and topic model | |
CN109902665A (en) | Similar face retrieval method, apparatus and storage medium | |
CN104598611B (en) | The method and system being ranked up to search entry | |
CN107330444A (en) | A kind of image autotext mask method based on generation confrontation network | |
CN109299342A (en) | A kind of cross-module state search method based on circulation production confrontation network | |
CN106502985A (en) | A kind of neural network modeling approach and device for generating title | |
CN107577737A (en) | Method and apparatus for pushed information | |
CN114565104A (en) | Language model pre-training method, result recommendation method and related device | |
CN109919078A (en) | A kind of method, the method and device of model training of video sequence selection | |
CN109949929A (en) | A kind of assistant diagnosis system based on the extensive case history of deep learning | |
CN112288075A (en) | Data processing method and related equipment | |
CN109492385A (en) | A kind of method for generating cipher code based on deep learning | |
CN109299211A (en) | A kind of text automatic generation method based on Char-RNN model | |
CN108304364A (en) | keyword extracting method and device | |
CN112131368B (en) | Dialogue generation method and device, electronic equipment and storage medium | |
CN107679225A (en) | A kind of reply generation method based on keyword | |
Shao et al. | Collaborative learning for answer selection in question answering | |
CN104915399A (en) | Recommended data processing method based on news headline and recommended data processing method system based on news headline | |
CN114462018A (en) | Password guessing system and method based on Transformer model and deep reinforcement learning | |
CN109543176A (en) | A kind of abundant short text semantic method and device based on figure vector characterization | |
Zheng et al. | A deep learning approach for expert identification in question answering communities | |
CN101963988A (en) | Intelligent engine for normalizing discretion and implementation method thereof | |
CN115994224A (en) | Phishing URL detection method and system based on pre-training language model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190319 |