CN108304877A - A kind of physical layer channel authentication method based on machine learning - Google Patents
A kind of physical layer channel authentication method based on machine learning Download PDFInfo
- Publication number
- CN108304877A CN108304877A CN201810104718.3A CN201810104718A CN108304877A CN 108304877 A CN108304877 A CN 108304877A CN 201810104718 A CN201810104718 A CN 201810104718A CN 108304877 A CN108304877 A CN 108304877A
- Authority
- CN
- China
- Prior art keywords
- receiver
- data
- channel
- legal
- information
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/12—Applying verification of the received information
- H04L63/126—Applying verification of the received information the source of the received data
Abstract
The physical layer channel authentication method based on machine learning that the invention discloses a kind of, includes the following steps:S1. receiver B carries out legal information transmitter A and simulation invalid information sender E channel information (CSI) acquisition of packet;S2. receiver B pre-processes the packet channel information of legal information transmitter A and simulation invalid information sender E;S3. receiver B generates the training dataset of two classification;S4. receiver B is trained according to the training dataset of two classification using the sorting algorithm in machine learning algorithm, is generated grader;S5. receiver B carries out the packet received using grader the judgement of legal person or illegal person, to realize the certification to packet.The physical layer channel authentication method based on machine learning that the present invention provides a kind of being suitable for resource-constrained authenticating device and scene, has the advantage that computation complexity is low, certification accuracy rate is high.
Description
Technical field
The present invention relates to physical layer channel certifications, are based especially on the physical layer channel certification of machine learning.
Background technology
In future broadband wireless communication systems, micro terminal equipment pours in the quantity in wireless network will exponentially multiplication
It is long.Micro terminal needs to establish the safety measures such as secure accessing certification in accessing wireless network process.But it is currently, traditional
Access authentication is mainly based upon cryptographic technique, and micro terminal equipment is diversified, and some is very single, for example wearable sets
Standby, easy internet-of-things terminal etc. can not provide complicated computing resource, edge calculations node have certain data processing and
Storage capacity carries out access authentication between asymmetric resource and is difficult to certificate scheme of the large-scale use tradition based on password.
Uniqueness when physical layer channel certification utilizes the sky of radio channel information is believed by comparing the channel between successive frame
Similitude is ceased to judge subscriber identity information.Physical layer channel certification directly utilizes channel information, without complicated upper layer encryption
Operation has fast and efficiently advantage, is very suitable for resource-constrained micro terminal.But physics in practical radio communication environment
The thresholding of layer certification is difficult to obtain and determine, to affect certification accuracy rate, even if thresholding can be traversed manually, but efficiency
Low and certification accuracy rate is also low.
For example, the scheme that dualism hypothesis method is authenticated physical layer is as follows:If receiver B has had authenticated k-th of number
Legal information sender A is derived from according to frame, the channel information of extraction isThe sender of+1 data frame of kth is still not
Know, channel information isBinary hypothesis test is:
Null hypothesisIt indicates that+1 data frame of kth is identical as k-th of data frame source, is legal information sender A;
Alternative hypothesisIt indicates that k-th of data frame source of+1 data frame of kth is different, is sent for simulation invalid information
Person E.
Ensure that the maximum time difference between continuous data frame is the premise that certification carries out within coherence time.
It is influenced and the evaluated error of channel estimation method itself by noise and ICI, the channel information actually obtained
Estimated valueWithIt can be expressed as respectively:
Wherein, NkAnd Nk+1It is all independent identically distributed multiple Gauss noise, and obeys N (0, σ2) distribution.So directly transporting
Hypothesis testing is carried out with channel information, needs to consider that noise variance influences, increases authentication complexity.Due to NkAnd Nk+1With same
The statistical property of sample, " difference " of channel information can eliminate the influence of noise variance.Physical layer certification is converted into channel letter
The comparison between " difference " and the threshold value of setting is ceased, above formula can be expressed as:
Null hypothesisWhen the channel information " difference " of front and back continuous data frame is less than threshold value, identity of the sender is legal;
Alternative hypothesisWhen the channel information " difference " of front and back continuous data frame is more than threshold value, identity of the sender is non-
Method.
In above formula as can be seen that physical layer certification it is practical be exactly comparison between channel information " difference " and certification thresholding,
So channel information " difference " and certification thresholding are the key points of physical layer certification.
Test statistics T can calculate channel information " difference ", and most common test statistics is the inspection based on amplitude
Statistic ΤA, in an ofdm system, had differences between subcarrier amplitude, this species diversity can be used for carrying out correspondent's identity
Certification.Continuous two data frame channel informations areWithInclude NsA frequency domain channel matrix is the OFDM of N-dimensional square formation
Symbol, wherein the phase offset of m (1≤m≤N) row n (1≤n≤N) column element is:
WhereinIndicate the conjugate complex number of plural number A.
Test statistics Τ A based on amplitude can be expressed as:
Wherein, σ2For noise power, ηAFor ΤACorresponding threshold value.But ΤAPresentation Non-random distribution, threshold value ηA
It is difficult to determine, to affect certification accuracy rate.
In recent years, some scholar's research application machine learning solve the above problems.It is adaptive to increase in machine learning field
(Adaptive Boosting) algorithm is to be proposed by Freund and Schapire nineteen ninety-five by force, most representational promotion
One of method is intended in classification problem, by changing the weight of training sample, learns multiple graders, and by these weak point
Class device combines to form strong classifier, and the algorithm Generalization error rate is low, easily encodes, and can be not necessarily to ginseng on most of grader
Number adjustment, has very high precision, is frequently used in two classification and more classification scenes.
Invention content
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of physical layer channels based on machine learning to recognize
Card method is acquired by the channel information to known legitimate information transmitter A and simulation invalid information sender E first,
And collected channel information is pre-processed, the training set of two classification is formed, is established based on sorting algorithm in machine learning
Model training set is trained, obtain grader, receiver B carries out being legal person using grader to the packet received
Or the judgement of illegal person obtains verification and measurement ratio to realize the certification to packet, under the conditions of verification and measurement ratio reaches prescribed requirement,
Available categorical device is generated, then carries out physical layer channel certification, there is the advantage that computation complexity is low, certification accuracy rate is high.
The purpose of the present invention is achieved through the following technical solutions:A kind of physical layer channel based on machine learning is recognized
Card method, includes the following steps:
S1. receiver B carries out legal information transmitter A and simulation invalid information sender E the channel information of packet
(CSI) it acquires:
Receiver B is acquired the channel information of the packet of legal information transmitter A, obtains adopting comprising continuous N frames
Collect the data set of resultPacket channels of the receiver B to simulation invalid information sender E
Information is acquired, and obtains the data set for including continuous N frames collection result
S2. receiver B is to data set Carry out pretreatment and data extraction:
Calculate data setTest statisticsObtain data set
In difference per two continuous frames channel informations, by TABMiddle partial frame is divided into training setRemaining frame number is as test setCalculate data setTest statisticsObtain data setIn
Per the difference of two continuous frames channel information, by TEBMiddle partial frame is divided into training setRemaining frame number is as test setWherein it is divided into training setWithPartial frame frame number be t;
S3. receiver B generates the training dataset of two classification:
By the training set of legal information sender A and illegal person E It is combined into the training dataset T of two classification
={ (x1,y1),(x2,y2),...,(x2t,y2t), in training dataset T:
That is yiWhen=+ 1, indicate that the data come from legal information sender A;Then yiWhen=- 1, indicate the data from simulation
Illegal person E;
S4. receiver B is instructed using the sorting algorithm in machine learning algorithm according to the training dataset T of two classification
Practice, generates grader;
S5. receiver B utilizes grader to test setWithThe two-category data collection T formed2(N-1-t)It carries out
Judge, obtains the grader that verification and measurement ratio reaches requirement;
S6. receiver B carries out physical layer channel certification using the grader for reaching requirement to the packet newly received.
Wherein, the step S1 includes legal information sender A dataset acquisitions step and simulation invalid information sender E
Dataset acquisition step;
The legal information sender A dataset acquisition steps include:
Legal information sender A sends continuous N number of data frame to receiver B;Receiver B is received from legal information
After first data frame of sender A, the channel information of legal information sender A to receiver B is extractedReceiver B connects
After receiving second data frame from legal information sender A, the channel letter of extraction legal information sender A to receiver B
BreathSimilarly, it obtains including continuous N after the continuous N frames channel information of extraction legal information sender A to receiver B always
The data set of frame collection result
The simulation invalid information sender E dataset acquisition steps:
It simulates invalid information sender E and sends continuous N number of data frame to receiver B;Receiver B is received from simulation
After first data frame of invalid information sender E, the channel information of extraction simulation invalid information sender E to receiver BAfter receiver B receives second data frame from simulation invalid information sender E, extraction simulation invalid information is sent
Channel informations of the person E to receiver BSimilarly, it after the continuous N frames channel information for extracting legal person E to receiver B always, obtains
To the data set for including continuous N frames collection result
Further, in the step S1, the channel information of extraction was required within coherence time, otherwise it is assumed that channel is believed
Breath does not have correlation.
Further, in the step S2, test statistics is the method for weighing channel information similarity, including but
It is not limited to the test statistics based on amplitude, the test statistics based on amplitude and phase combining, and inclined based on correction phase
The test statistics etc. of shifting.
Further, in the step S3, it is divided into training setWithPartial frame frame number t meet:
N/2 < t < N-1;
And from TABIn be divided into training setIn partial frame must assure that there is between two continuous frames correlation, and
From TEBIn be divided into training setIn partial frame also must assure that the correlation between two continuous frames.
Further, in the step S4, the sorting algorithm in the machine learning algorithm includes but not limited to k- neighbours
Algorithm, NB Algorithm, SVM algorithm and decision Tree algorithms;
Further, in the step S4, the grader described in step S4 or step S5 includes but not limited to
AdaBoost graders, bagging graders and boosting graders etc..
Further, the step S5 includes:
By test setWithForm two-category data collection T2(N-1-t):
T2(N-1-t)={ (x1,y1),…,(xk,yk),…,(x2(N-1-t),y2(N-1-t), k=1 ..., L ..., 2 (N-1-t)
By T2(N-1-t)In data xkIt is sequentially inputted in grader, obtains the y of grader outputkValue, if the y of outputkValue
It is judged as simulating invalid information sender E for -1, if the y of outputkValue is judged as legal information sender A for+1;
Statistical data collection T2(N-1-t)In total amount of data 2 (N-1-t) and the judicious data volume K1 of grader, count accordingly
Verification and measurement ratio α=the K1/2 (N-1-t) for calculating grader, as the foundation for judging classifier performance:
If prescribed requirement (given threshold) is not achieved in verification and measurement ratio α, again since S1, legal person and simulation are further collected
The channel information data of illegal person, repeats the training of S2-S5, until verification and measurement ratio α reaches prescribed requirement;
Further, the step S6 includes:
Receiver B receives the packet of unknown senderAnd pre-processed, it calculates
Data setTest statisticsObtain data setIn every two frame letter
The difference of road information is that legal person or illegal person judge to the packet received using verification and measurement ratio grader up to standard,
To realize the certification to packet.
The beneficial effects of the invention are as follows:Channel of the present invention to legal information transmitter A and simulation invalid information sender E
Information is acquired, and is pre-processed to collected channel information, and the data set of two classification is formed, by the data of two classification
Collection is divided into training set and test set, and the model established based on sorting algorithm in machine learning is trained training set, is divided
Class device recycles test set to test the performance of the grader, and final output judgement is as a result, cycle obtains inspection up to standard always
After survey rate, physical layer channel certification is carried out to unknown connector's packet using grader up to standard.Suitable for resource-constrained
Authenticating device and scene have the advantage that computation complexity is low, certification accuracy rate is high.
Description of the drawings
Fig. 1 is the physical layer authentication method flow chart based on machine learning algorithm;
Fig. 2 is the scatterplot of legal information sender A and the normalization test statistics Τ A for simulating invalid information sender E
Figure;
Fig. 3 is the verification and measurement ratio curve graph based on artificial traversal thresholding.
Fig. 4 is the verification and measurement ratio comparison diagram based on machine learning and manually traversed.
Specific implementation mode
Technical scheme of the present invention is described in further detail below in conjunction with the accompanying drawings, but protection scope of the present invention is not limited to
It is as described below.
The present invention is implemented in the actual environment with multiple illegal nodes and legitimate node, and receiver B first is to legal
Information transmitter A and simulation invalid information sender E carries out channel information (CSI) acquisition of packet, ensures continuous data frame
Between maximum time difference within coherence time, the grader of generation is by taking AdaBoost graders as an example.
As shown in Figure 1, a kind of physical layer channel authentication method based on machine learning, includes the following steps:
S1. receiver B carries out legal information transmitter A and simulation invalid information sender E the channel information of packet
(CSI) it acquires:
Receiver B is acquired the channel information of the packet of legal information transmitter A, obtains adopting comprising continuous N frames
Collect the data set of resultPacket channels of the receiver B to simulation invalid information sender E
Information is acquired, and obtains the data set for including continuous N frames collection result
S2. receiver B is to data set Carry out pretreatment and data extraction:
Calculate data setTest statisticsObtain data set
In difference per two continuous frames channel informations, by TABMiddle partial frame is divided into training setRemaining frame number is as test setCalculate data setTest statisticsObtain data setIn
Per the difference of two continuous frames channel information, by TEBMiddle partial frame is divided into training setRemaining frame number is as test setWherein, it is divided into training setWithPartial frame frame number be t;
S3. receiver B generates the training dataset of two classification:
By the training set of legal information sender A and illegal person E It is combined into the training dataset T of two classification
={ (x1,y1),(x2,y2),...,(x2t,y2t), in training dataset T:
That is yiWhen=+ 1, indicate that the data come from legal information sender A;Then yiWhen=- 1, indicate the data from simulation
Illegal person E;
S4. receiver B is instructed using the sorting algorithm in machine learning algorithm according to the training dataset T of two classification
Practice, generates grader;
S5. receiver B utilizes grader to test setWithThe two-category data collection T formed2(N-1-t)It carries out
Judge, obtains the grader that verification and measurement ratio reaches requirement;
S6. receiver B carries out physical layer channel certification using the grader for reaching requirement to the packet newly received.
The step S1 includes legal information sender A dataset acquisitions step and simulation invalid information sender's E data
Collect acquisition step;
The legal information sender A dataset acquisition steps include:
Legal information sender A sends continuous N number of data frame to receiver B;Receiver B is received from legal information
After first data frame of sender A, the channel information of legal information sender A to receiver B is extractedReceiver B is received
To after second data frame of legal information sender A, the channel information of legal information sender A to receiver B is extractedSimilarly, it obtains including continuous N frames after the continuous N frames channel information of extraction legal information sender A to receiver B always
The data set of collection result
The simulation invalid information sender E dataset acquisition steps:
It simulates invalid information sender E and sends continuous N number of data frame to receiver B, receiver B is received from simulation
After first data frame of invalid information sender E, the channel information of extraction simulation invalid information sender E to receiver BAfter receiver B receives second data frame from simulation invalid information sender E, extraction simulation invalid information hair
Channel informations of the person of the sending E to receiver BSimilarly, the continuous N frames channel information of legal person E to receiver B is extracted always
Afterwards, obtain including the data set of continuous N frames collection result
Sorting algorithm in the machine learning algorithm includes but not limited to k- nearest neighbor algorithms, NB Algorithm, SVM
Algorithm and decision Tree algorithms.
In the embodiment of this patent, the process that step S4 generates grader is as follows:Receiver B is calculated using Adaboost
Method generates Adaboost graders (i.e. strong classifier), specifically in conjunction with Weak Classifier:
The first step, input training dataset T={ (x1,y1),(x2,y2),...,(x2t,y2t)};
Second step initializes the weights distribution D of training data1=(w11,…,w1i,…,w1,t),
Third walks, to m=1,2 ..., M
(1) it is distributed D using with weightsmTraining dataset study, obtain basic classification device, i.e. Weak Classifier;
Gm(x):xi→{-1,+1}
(2) G is calculatedm(x) the error in classification rate on training dataset:
(3) G is calculatedm(x) coefficient:
Here logarithm is natural logrithm;
(4) the weights distribution of update training dataset
Dm+1=(wm+1,1,…,wm+1,i,…,wm+1,2t),
Here, ZmIt is standardizing factor, it makes Dm+1As a probability distribution:
4th step builds the linear combination of basic classification device:
Obtain final classification device:
G (x)=sign (f (x)).
In above-mentioned steps, using Adaboost algorithm repetition learning basic classification device, is sequentially performed in each round
The coefficient calculation method of three steps, the 4th step, including but not limited to index, loss function etc., you can obtain final Adaboost
Grader G (x) (strong classifier).
In embodiments herein, the step S5 includes:
By test setWithThe two-category data collection T formed2(N-1-t):
T2(N-1-t)={ (x1,y1),…,(xk,yk),…,(x2(N-1-t),y2(N-1-t)), k=1 ..., L ..., 2 (N-1-t)
By T2(N-1-t)In data xkIt is sequentially inputted in grader G (x), obtains the y of grader G (x) outputskValue, if
The y of outputkValue is judged as simulating invalid information sender E for -1, if the y of outputkValue is judged as legal information sender A for+1;
Statistical data collection T2(N-1-t)In total amount of data 2 (N-1-t) and the judicious data volume K1 of grader G (x), according to
This calculates verification and measurement ratio α=K1/2 (N-1-t) of grader, as the foundation for judging classifier performance:
If prescribed requirement is not achieved in verification and measurement ratio α, again since S1, further collects legal person and simulate the letter of illegal person
Road information data repeats the training of S2-S5, until verification and measurement ratio α reaches prescribed requirement;
In embodiments herein, the step S6 includes:Receiver B receives the packet of unknown senderAnd pre-processed, calculate data setTest statisticsObtain data setIn every two frames channel information difference, utilize verification and measurement ratio
Grader up to standard is that legal person or illegal person judge to the packet received, to realize the certification to packet.
Such as:When receiver B receives the access request of identity unknown object, the grader for reaching prescribed requirement can be utilized
G (x) judges the legitimacy of the object:Specifically, receiver B receives the two continuous frames data (M=2 at this time) from the object,
And the channel information for including in this two frame data is extracted, it is poor further according to the channel information of channel information two frame data of calculating of extraction
The difference is input to grader G (x), the y values that grader is inputted, if output judges the identity unknown object for -1 by value
For illegal person, receiver B refuses object access, if output be+1 to judge that identity unknown object is legal person, receiver B permissions
The object accesses.
When carrying out Channel authentication using conventional method, since actual channel information is unknown, thresholding can not be determined, to influence
Certification accuracy rate, as shown in Fig. 2, for the normalization test statistics of legal information sender A and simulation invalid information sender E
ΤAScatter plot, from figure we have seen that:Simulate the normalization Τ of invalid information sender E and legal information sender AAValue is deposited
Staggeredly, it is meant that can not use a threshold value that they are accurate separately;Even if by manually traversing thresholding, there is also efficiency
Problem low, certification accuracy rate is low obtains the verification and measurement ratio of Accurate classification most as shown in figure 3, thresholding is traversed between [0,1]
It is high also there was only 79.8%.And the application trains two-category data collection T using machine learning algorithm, the rate of can be detected is
89.7%, the verification and measurement ratio comparison diagram based on machine learning and manually traversed is as shown in Figure 4, it is seen that relative to artificial traversal thresholding,
There is higher certification accuracy rate using the authentication model obtained based on machine learning method;Machine learning is practical to be widely used in
In two classification or more classification scenes, when equipment accesses, can quickly it be sentenced using the physical layer authentication model based on machine learning
It is legal to determine identity, if allow to access network, be entirely capable of realizing low time delay, real-time access authentication.
To sum up, the present invention is acquired the channel information of legal information transmitter A and simulation invalid information sender E,
And collected channel information is pre-processed, the data set of two classification is formed, the data set of two classification is divided into training set
And test set, the model established based on sorting algorithm in machine learning are trained training set, obtain grader, recycle and survey
The performance of the examination set pair grader is tested, after final output judgement is as a result, cycle obtains verification and measurement ratio up to standard always, using up to
Target grader carries out physical layer channel certification to unknown connector's packet.Suitable for resource-constrained authenticating device and field
Scape has the advantage that computation complexity is low, certification accuracy rate is high.
The above is the preferred embodiment of the present invention, it should be understood that the present invention is not limited to shape described herein
Formula should not be viewed as excluding other embodiments, and can be used for other combinations, modification and environment, and can be in this paper institutes
It states in contemplated scope, modifications can be made through the above teachings or related fields of technology or knowledge.And what those skilled in the art were carried out
Modifications and changes do not depart from the spirit and scope of the present invention, then all should be in the protection domain of appended claims of the present invention.
Claims (9)
1. a kind of physical layer channel authentication method based on machine learning, it is characterised in that:Include the following steps:
S1. receiver B carries out legal information transmitter A and simulation invalid information sender E the channel information acquisition of packet:
Receiver B is acquired the channel information of the packet of legal information transmitter A, obtains comprising continuous N frames acquisition knot
The data set of fruitPacket channel informations of the receiver B to simulation invalid information sender E
It is acquired, obtains the data set for including continuous N frames collection result
S2. receiver B is to data set Carry out pretreatment and data extraction:
Calculate data setTest statisticsObtain data setIn it is every
The difference of two continuous frames channel information, by TABMiddle partial frame is divided into training setRemaining frame number is as test set
Calculate data setTest statisticsObtain data setIn per continuous
The difference of two frame channel informations, by TEBMiddle partial frame is divided into training set Tt EB, remaining frame number is as test setIts
In, it is divided into training setAnd Tt EBPartial frame frame number be t;
S3. receiver B generates the training dataset of two classification:
By the training set of legal information sender A and illegal person ETt EBIt is combined into the training dataset T=of two classification
{(x1,y1),(x2,y2),...,(x2t,y2t), in training dataset T:
That is yiWhen=+ 1, indicate that the data come from legal information sender A;Then yiWhen=- 1, indicate that the data are illegal from simulation
Person E;
S4. receiver B is trained using the sorting algorithm in machine learning algorithm according to the training dataset T of two classification, raw
Constituent class device;
S5. receiver B utilizes grader to test setWithThe two-category data collection T formed2(N-1-t)Sentenced
It is disconnected, obtain the grader that verification and measurement ratio reaches requirement;
S6. receiver B carries out physical layer channel certification using the grader for reaching requirement to the packet newly received.
2. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
Step S1 includes legal information sender A dataset acquisitions step and simulation invalid information sender's E dataset acquisition steps;
The legal information sender A dataset acquisition steps include:
Legal information sender A sends continuous N number of data frame to receiver B;Receiver B is received to be sent from legal information
After first data frame of person A, the channel information of legal information sender A to receiver B is extractedReceiver B is received
After second data frame from legal information sender A, the channel information of legal information sender A to receiver B is extractedSimilarly, it obtains including continuous N frames after the continuous N frames channel information of extraction legal information sender A to receiver B always
The data set of collection result
The simulation invalid information sender E dataset acquisition steps:
It simulates invalid information sender E and sends continuous N number of data frame to receiver B, receiver B receives illegal from simulation
After first data frame of information transmitter E, the channel information of extraction simulation invalid information sender E to receiver BIt connects
After receipts machine B receives second data frame from simulation invalid information sender E, extraction simulation invalid information sender E is arrived
The channel information of receiver BSimilarly, it after the continuous N frames channel information for extracting legal person E to receiver B always, is wrapped
Data set containing continuous N frames collection result
3. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
In step S1, the channel information of extraction was required within coherence time, otherwise it is assumed that channel information does not have correlation.
4. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
In step S2, test statistics is the method for weighing channel information similarity, including but not limited to the inspection based on amplitude
Statistic, the test statistics based on amplitude and phase combining, and the test statistics etc. based on correction phase offset.
5. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
In step S3, it is divided into training setAnd Tt EBPartial frame frame number t meet:
N/2 < t < N-1;
And from TABIn be divided into training setIn partial frame must assure that there is between two continuous frames correlation, and from TEB
In be divided into training setIn partial frame also must assure that the correlation between two continuous frames.
6. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
In step S4, the sorting algorithm in the machine learning algorithm include but not limited to k- nearest neighbor algorithms, NB Algorithm,
SVM algorithm and decision Tree algorithms.
7. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:Step
Grader described in S4 or step S5 includes but not limited to AdaBoost graders, bagging graders and boosting points
Class device.
8. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
Step S5 includes:
By test setWithForm two-category data collection T2(N-1-t):
T2(N-1-t)={ (x1,y1),…,(xk,yk),…,(x2(N-1-t),y2(N-1-t)), k=1 ..., L ..., 2 (N-1-t)
By T2(N-1-t)In data xkIt is sequentially inputted in grader, obtains the y of grader outputkValue, if the y of outputkValue is -1
It is judged as simulating invalid information sender E, if the y of outputkValue is judged as legal information sender A for+1;
Statistical data collection T2(N-1-t)In total amount of data 2 (N-1-t) and the judicious data volume K1 of grader, accordingly calculate point
Verification and measurement ratio α=K1/2 (N-1-t) of class device, as the foundation for judging classifier performance:
If prescribed requirement is not achieved in verification and measurement ratio α, again since S1, further collects legal person and simulate the channel letter of illegal person
Data are ceased, the training of S2-S5 are repeated, until verification and measurement ratio α reaches prescribed requirement.
9. a kind of physical layer channel authentication method based on machine learning according to claim 1, it is characterised in that:It is described
Step S6 includes:
Receiver B receives the packet of unknown senderAnd pre-processed, calculate data
CollectionTest statisticsObtain data setIn every two frames channel letter
The difference of breath is that legal person or illegal person judge to the packet received using verification and measurement ratio grader up to standard, with reality
Now to the certification of packet.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810104718.3A CN108304877B (en) | 2018-02-02 | 2018-02-02 | Physical layer channel authentication method based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810104718.3A CN108304877B (en) | 2018-02-02 | 2018-02-02 | Physical layer channel authentication method based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108304877A true CN108304877A (en) | 2018-07-20 |
CN108304877B CN108304877B (en) | 2021-10-08 |
Family
ID=62864178
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810104718.3A Active CN108304877B (en) | 2018-02-02 | 2018-02-02 | Physical layer channel authentication method based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108304877B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109343951A (en) * | 2018-08-15 | 2019-02-15 | 南京邮电大学 | Mobile computing resource allocation methods, computer readable storage medium and terminal |
CN109358962A (en) * | 2018-08-15 | 2019-02-19 | 南京邮电大学 | The autonomous distributor of mobile computing resource |
CN109829477A (en) * | 2018-12-29 | 2019-05-31 | 北京邮电大学深圳研究院 | More attribute physical layer authentication methods, device and server based on heuristic cluster |
CN109862563A (en) * | 2019-01-22 | 2019-06-07 | 杭州电子科技大学 | A kind of physical layer authentication method and system suitable for mobile wireless network environment |
CN110149342A (en) * | 2019-05-30 | 2019-08-20 | 四川少泽电子设备有限公司 | A kind of edge device physical layer authentication method for combining thresholding and machine learning |
CN110944002A (en) * | 2019-12-06 | 2020-03-31 | 深圳供电局有限公司 | Physical layer authentication method based on exponential average data enhancement |
CN111541632A (en) * | 2020-04-20 | 2020-08-14 | 四川农业大学 | Physical layer authentication method based on principal component analysis and residual error network |
CN111563033A (en) * | 2020-04-02 | 2020-08-21 | 北京明略软件系统有限公司 | Method and device for generating simulation data |
CN112396132A (en) * | 2021-01-19 | 2021-02-23 | 国网江苏省电力有限公司南京供电分公司 | SVM-based wireless terminal secure access method |
CN112887371A (en) * | 2021-01-12 | 2021-06-01 | 深圳市中博科创信息技术有限公司 | Edge calculation method and device, computer equipment and storage medium |
CN114501446A (en) * | 2022-01-17 | 2022-05-13 | 河北大学 | Physical layer authentication method based on PU (polyurethane) bagging strategy in dynamic industrial scene |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030158960A1 (en) * | 2000-05-22 | 2003-08-21 | Engberg Stephan J. | System and method for establishing a privacy communication path |
CN1933463A (en) * | 2006-10-10 | 2007-03-21 | 东南大学 | Block mode channel estimating method in electric line communication system |
CN102991164A (en) * | 2011-07-28 | 2013-03-27 | 卡姆特有限公司 | System and method for inspecting solder mask |
US20130163761A1 (en) * | 2010-03-05 | 2013-06-27 | University Of Maryland | Wireless communication method and system for transmission authentication at the physical layer |
CN103413078A (en) * | 2013-07-25 | 2013-11-27 | 武汉汉德瑞庭科技有限公司 | Double-layer online identification system and identification method based on user's mark and handwriting |
CN105510348A (en) * | 2015-12-31 | 2016-04-20 | 南京协辰电子科技有限公司 | Flaw detection method and device of printed circuit board and detection equipment |
CN107046468A (en) * | 2017-06-14 | 2017-08-15 | 电子科技大学 | A kind of physical layer certification thresholding determines method and system |
CN107615358A (en) * | 2015-03-31 | 2018-01-19 | 深圳市大疆创新科技有限公司 | For identifying the Verification System and method of authorized participant |
-
2018
- 2018-02-02 CN CN201810104718.3A patent/CN108304877B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030158960A1 (en) * | 2000-05-22 | 2003-08-21 | Engberg Stephan J. | System and method for establishing a privacy communication path |
CN1933463A (en) * | 2006-10-10 | 2007-03-21 | 东南大学 | Block mode channel estimating method in electric line communication system |
US20130163761A1 (en) * | 2010-03-05 | 2013-06-27 | University Of Maryland | Wireless communication method and system for transmission authentication at the physical layer |
CN102991164A (en) * | 2011-07-28 | 2013-03-27 | 卡姆特有限公司 | System and method for inspecting solder mask |
CN103413078A (en) * | 2013-07-25 | 2013-11-27 | 武汉汉德瑞庭科技有限公司 | Double-layer online identification system and identification method based on user's mark and handwriting |
CN107615358A (en) * | 2015-03-31 | 2018-01-19 | 深圳市大疆创新科技有限公司 | For identifying the Verification System and method of authorized participant |
CN105510348A (en) * | 2015-12-31 | 2016-04-20 | 南京协辰电子科技有限公司 | Flaw detection method and device of printed circuit board and detection equipment |
CN107046468A (en) * | 2017-06-14 | 2017-08-15 | 电子科技大学 | A kind of physical layer certification thresholding determines method and system |
Non-Patent Citations (6)
Title |
---|
CHENGCHENG PEI 等: "Channel-based Physical Layer Authentication", 《2014 IEEE GLOBAL COMMUNICATIONS CONFERENCE》 * |
NING WANG 等: "Physical-Layer Authentication Based on Extreme Learning Machine", 《IEEE COMMUNICATIONS LETTERS》 * |
张金玲 等: "时变OFDM系统中基于基扩展模型的物理层认证", 《电子技术应用》 * |
杨建喜 等: "基于支持向量机的4G室内物理层认证算法", 《计算机应用》 * |
范靓 等: "无线Ad hoc网络中的联合信源信道有效用户安全识别方法", 《电子与信息学报》 * |
陈洁 等: "物联网中移动终端的协同身份检测机制研究", 《通信技术》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109343951A (en) * | 2018-08-15 | 2019-02-15 | 南京邮电大学 | Mobile computing resource allocation methods, computer readable storage medium and terminal |
CN109358962A (en) * | 2018-08-15 | 2019-02-19 | 南京邮电大学 | The autonomous distributor of mobile computing resource |
CN109343951B (en) * | 2018-08-15 | 2022-02-11 | 南京邮电大学 | Mobile computing resource allocation method, computer-readable storage medium and terminal |
CN109358962B (en) * | 2018-08-15 | 2022-02-11 | 南京邮电大学 | Mobile computing resource autonomous allocation device |
CN109829477A (en) * | 2018-12-29 | 2019-05-31 | 北京邮电大学深圳研究院 | More attribute physical layer authentication methods, device and server based on heuristic cluster |
CN109862563A (en) * | 2019-01-22 | 2019-06-07 | 杭州电子科技大学 | A kind of physical layer authentication method and system suitable for mobile wireless network environment |
CN109862563B (en) * | 2019-01-22 | 2021-08-10 | 杭州电子科技大学 | Physical layer authentication method and system suitable for mobile wireless network environment |
CN110149342A (en) * | 2019-05-30 | 2019-08-20 | 四川少泽电子设备有限公司 | A kind of edge device physical layer authentication method for combining thresholding and machine learning |
CN110149342B (en) * | 2019-05-30 | 2020-01-10 | 四川少泽电子设备有限公司 | Edge device physical layer authentication method combining threshold and machine learning |
CN110944002B (en) * | 2019-12-06 | 2020-08-21 | 深圳供电局有限公司 | Physical layer authentication method based on exponential average data enhancement |
CN110944002A (en) * | 2019-12-06 | 2020-03-31 | 深圳供电局有限公司 | Physical layer authentication method based on exponential average data enhancement |
CN111563033A (en) * | 2020-04-02 | 2020-08-21 | 北京明略软件系统有限公司 | Method and device for generating simulation data |
CN111563033B (en) * | 2020-04-02 | 2023-06-02 | 北京明略软件系统有限公司 | Simulation data generation method and device |
CN111541632A (en) * | 2020-04-20 | 2020-08-14 | 四川农业大学 | Physical layer authentication method based on principal component analysis and residual error network |
CN112887371A (en) * | 2021-01-12 | 2021-06-01 | 深圳市中博科创信息技术有限公司 | Edge calculation method and device, computer equipment and storage medium |
CN112887371B (en) * | 2021-01-12 | 2022-05-13 | 深圳市中博科创信息技术有限公司 | Edge calculation method and device, computer equipment and storage medium |
CN112396132A (en) * | 2021-01-19 | 2021-02-23 | 国网江苏省电力有限公司南京供电分公司 | SVM-based wireless terminal secure access method |
CN112396132B (en) * | 2021-01-19 | 2022-04-08 | 国网江苏省电力有限公司南京供电分公司 | SVM-based wireless terminal secure access method |
US11678189B2 (en) | 2021-01-19 | 2023-06-13 | State Grid Jiangsu Electric Power Co., Ltd | SVM-based secure access method for wireless terminals |
CN114501446A (en) * | 2022-01-17 | 2022-05-13 | 河北大学 | Physical layer authentication method based on PU (polyurethane) bagging strategy in dynamic industrial scene |
CN114501446B (en) * | 2022-01-17 | 2023-07-25 | 河北大学 | Physical layer authentication method based on PU (polyurethane) bagging strategy in dynamic industrial scene |
Also Published As
Publication number | Publication date |
---|---|
CN108304877B (en) | 2021-10-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108304877A (en) | A kind of physical layer channel authentication method based on machine learning | |
Chen et al. | ConFi: Convolutional neural networks based indoor Wi-Fi localization using channel state information | |
CN108932535B (en) | A kind of edge calculations clone's node recognition methods based on machine learning | |
Xie et al. | A generalizable model-and-data driven approach for open-set RFF authentication | |
CN108924836A (en) | A kind of edge side physical layer channel authentication method based on deep neural network | |
Huang et al. | WiDet: Wi-Fi based device-free passive person detection with deep convolutional neural networks | |
Qiu et al. | Physical layer authentication enhancement using a Gaussian mixture model | |
Chen et al. | Physical-layer channel authentication for 5G via machine learning algorithm | |
CN104683051B (en) | Wireless multipath fading channel simulating method and channel simulator | |
CN107046468A (en) | A kind of physical layer certification thresholding determines method and system | |
CN109117747A (en) | Radar signal classification method based on cyclo-stationary Yu depth convolutional neural networks | |
Cai et al. | PILC: Passive indoor localization based on convolutional neural networks | |
CN110149342A (en) | A kind of edge device physical layer authentication method for combining thresholding and machine learning | |
CN104660356B (en) | A kind of collaborative frequency spectrum sensing method of half-blindness with reliable false-alarm performance | |
CN112399413A (en) | Physical layer identity authentication method based on deep support vector description method | |
CN112749633B (en) | Separate and reconstructed individual radiation source identification method | |
CN110138459A (en) | Sparse underwater sound orthogonal frequency division multiplexing channel estimation methods and device based on base tracking denoising | |
CN110944002B (en) | Physical layer authentication method based on exponential average data enhancement | |
Cheng et al. | Deep learning network based spectrum sensing methods for OFDM systems | |
CN115294615A (en) | Radio frequency fingerprint identification method based on machine learning | |
CN104112062A (en) | Method for obtaining wind resource distribution based on interpolation method | |
Mohammadpourfard et al. | Generation of false data injection attacks using conditional generative adversarial networks | |
CN114867018A (en) | Device identification authentication method, system and storage medium using carrier frequency offset | |
CN108491719A (en) | A kind of Android malware detection methods improving NB Algorithm | |
Zhang et al. | Adaptive RF fingerprints fusion via dual attention convolutions |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |