CN110118926A - PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method - Google Patents
PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method Download PDFInfo
- Publication number
- CN110118926A CN110118926A CN201910446026.1A CN201910446026A CN110118926A CN 110118926 A CN110118926 A CN 110118926A CN 201910446026 A CN201910446026 A CN 201910446026A CN 110118926 A CN110118926 A CN 110118926A
- Authority
- CN
- China
- Prior art keywords
- magnetic distribution
- layer
- training
- distribution matrix
- output
- 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
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/28—Testing of electronic circuits, e.g. by signal tracer
- G01R31/2801—Testing of printed circuits, backplanes, motherboards, hybrid circuits or carriers for multichip packages [MCP]
- G01R31/2803—Testing of printed circuits, backplanes, motherboards, hybrid circuits or carriers for multichip packages [MCP] by means of functional tests, e.g. logic-circuit-simulation or algorithms therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a kind of PCB based on Electromagnetic Environmental Effect to distort intelligent detecting method comprising it has been known that there is without the pcb board composing training collection distorted for selection；The magnetic distribution of all pcb boards obtains magnetic distribution matrix M × N in measurement training set；After carrying out out-of-order and normalized to magnetic distribution matrix, mathematical feature is extracted；The sample for choosing preset ratio in training set has supervision neural network as training sample training BP；It is inputted magnetic distribution matrix remaining in training set as verifying sample in the neural network of initial training, and exports verifying accuracy；Judge to verify whether accuracy reaches preset requirement, if so, otherwise, enabling M=M+1, N=N+1 into next step, and returns to magnetic distribution matrix acquisition step；The test magnetic distribution matrix of pcb board to be tested is obtained, and is successively normalized and mathematics feature extraction；It inputs trained neural network to be identified, and exports recognition result.
Description
Technical field
The present invention relates to the detection techniques of PCB hardware, and in particular to a kind of PCB based on Electromagnetic Environmental Effect distorts intelligent inspection
Survey method.
Background technique
The physics of PCB, which is distorted, to be referred to through the component of malice distorted on PCB and size, trace width of via hole etc.
Deng, thus cause the change of circuit function and performance, reveal circuit information or cause fault etc., this is to hardware structure
At great security risk.However, this makes it be very difficult to examine since this physics is distorted small in size and easy to operate
It surveys.Currently, there is a method of the hardware Trojan horse of many detection chip grades in academia, including wing passage analysis, logic testing and inverse
To engineering etc..
Hardware Trojan horse for chip-scale is mainly that the malice carried out in logic is distorted, and detection is easier to, but for PCB
The hardware Trojan horse of plate grade mainly carries out some malice physically and distorts, and detection is got up extremely difficult.
Serious consequence caused by largely resting on wooden horse for the detection method that some malice of pcb board grade are distorted at present
And in defensive measure.Such as: safety analysis having been carried out to the commonly used smart machine of people day, to the hardware Trojan horse on PCB
Detailed classification has been carried out, preliminary analysis etc. has been carried out to the hardware Trojan horse on PCB, but these detection methods are all difficult to reality
The detection that the malice of hardware on existing plank is distorted.
Summary of the invention
For above-mentioned deficiency in the prior art, the PCB provided by the invention based on Electromagnetic Environmental Effect distorts intelligent measurement
Method can be gone out with higher Probability Detection plank power on hold and cover copper face product malice distort.
In order to achieve the above object of the invention, the technical solution adopted by the present invention are as follows:
A kind of PCB based on Electromagnetic Environmental Effect is provided and distorts intelligent detecting method comprising:
S1, multiple pcb board composing training collection for having the pcb board distorted and nothing to distort are chosen；
S2, the magnetic distribution that all pcb boards in training set are measured using probe, obtaining line number is M, and columns is the electricity of N
Distribution of Magnetic Field matrix；
S3, label is added to magnetic distribution matrix each in training set, and successively carries out out-of-order and normalized, later
Extract the mathematical feature of each magnetic distribution matrix；
S4, the magnetic distribution matrix of preset ratio in training set extracted by mathematical feature is chosen as training sample
There is supervision neural network to be trained BP, obtains the neural network of initial training；
S5, it is inputted just using the magnetic distribution matrix extracted by mathematical feature remaining in training set as verifying sample
In the neural network for walking training, and export the verifying accuracy distorted；
S6, judge to verify whether accuracy reaches preset requirement, if reaching, complete the training of neural network, and enter
Otherwise step S7 enables M=M+1, N=N+1, and return step S2；
S7, the test magnetic distribution matrix M × N for obtaining pcb board to be tested, and successively to test magnetic distribution matrix
It is normalized and mathematics feature extraction；And
S8, the test magnetic distribution matrix extracted by mathematical feature is known using trained neural network
Not, and the recognition result whether distorted is exported, completes PCB tampering detection.
Further, the mathematical feature for extracting each magnetic distribution matrix further comprises:
S31, according to the magnetic distribution Matrix Solving covariance matrix G in training set:
Wherein, Di(i ∈ [1,2 ..., K]) it is magnetic distribution matrix in training set；K is the electromagnetic field point in training set
The quantity of cloth matrix；For all magnetic distribution matrix DsiAverage value；T is the transposition of matrix；
S32, the corresponding feature vector of maximum P characteristic value in covariance matrix G is chosen:
Wherein,For the characteristic value of covariance matrix G a column, a ∈ (1,2 ..., N), N are magnetic distribution matrix Di
Columns, λ is constant, be equal to 0.9 or 0.95；
S33, descending sort is carried out to feature vector according to the corresponding characteristic value size of P feature vector selected, it is raw
At best projection axis XO:
Wherein,For k-th of feature vector after sequence, k ∈ (1,2 ..., P)；
S34, according to best projection XOCalculate the eigenmatrix F of each pcb board in training seti:
Fi=DiXO；
S35, eigenmatrix F is calculatediIn each column vector two norms, obtain two norm matrix Ls of each trained pcb boardi:
Wherein,It is characterized matrix LiJth column,Fi jFor sample characteristics matrix FiJth column,For
Fi jTwo norms.
Further, the magnetic distribution matrix extracted by mathematical feature for choosing preset ratio in training set is made
Have to supervise neural network and be trained to BP for training sample and further comprise:
S41, initialization BP have weight, biasing, learning rate and frequency of training in supervision neural network；
S42, operation is carried out to input node using excitation function, obtains the output that BP has supervision neural network hidden layer
Hj:
Wherein, i=1 ... n, j=1 ... l, n are the node number of input layer, and l is the node number of hidden layer；wijFor input
Layer arrives the weight of hidden layer, ajFor the biasing of input layer to hidden layer, xiFor the input data of i-th of node；G () is excitation letter
Number；
S43, it is calculated according to the biasing of the output of hidden layer, the weight of hidden layer to output layer and hidden layer to output layer
The output result of output layer；
S44, error amount is calculated according to the output result of output layer and the desired output of training sample；
S45, using gradient descent method update hidden layer to output layer weight and input layer to hidden layer weight, wherein
Hidden layer to output layer weight more new formula are as follows:
wjk=wjk+ηHjek
Input layer to hidden layer weight more new formula are as follows:
Wherein, η is learning rate；ek=Yk-Ok；wjkFor the weight of hidden layer to output layer；
S46, hidden layer is updated to the biasing of output layer, update the biasing of input layer to hidden layer again later:
S47, judge that BP has whether supervision neural network is completed training:
When the difference between adjacent error twice is less than given threshold or the number of iterations is equal to preset times, obtain preliminary
Trained neural network；
When between adjacent error twice difference and the number of iterations condition is not satisfied when, return step S42.
Further, the output result O of the output layerkCalculation formula are as follows:
Wherein, OkTo export result；K=1 ... m, m are the node number of output layer；bkFor hidden layer to the inclined of output layer
It sets.
Further, the calculation formula of the error amount are as follows:
Wherein, E is error amount；YkFor desired output.
The invention has the benefit that this programme has supervision neural network to the various objects on PCB by the BP trained
Reason distorts carry out classification and Detection, can be realized and covers distorting for copper face product and distorting for component resistance and capacitor on PCB, is carrying out
In identification process, to covering, copper face is long-pending and component capacitor distorts verification and measurement ratio with higher.
Detailed description of the invention
Fig. 1 is the flow chart that the PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method.
Fig. 2 is principle of lowpass filter figure in embodiment.
Fig. 3 is the pcb board figure of low-pass filter in Fig. 2.
Specific embodiment
A specific embodiment of the invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific embodiment, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the column of protection.
The flow chart that the PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method is shown with reference to Fig. 1, Fig. 1, such as Fig. 1 institute
Show, this method S includes step S1 to step S8.
In step sl, multiple pcb board composing training collection for having the pcb board distorted and nothing to distort are chosen；
In step s 2, using the magnetic distribution of all pcb boards in probe measurement training set, obtaining line number is M, columns
For the magnetic distribution matrix of N；
In step s3, label is added to magnetic distribution matrix each in training set, and successively carries out out-of-order and normalization
Processing, extracts the mathematical feature of each magnetic distribution matrix later, obtains two norm matrixes of electromagnetic field value；
Since this programme is that BP has a supervision neural network, therefore need to known that whether there is or not the magnetic distributions for the plank distorted
Matrix adds label, wherein distorting is label 1, and not distorting is label 2；Later again to the electromagnetic field point plus label
Cloth matrix carries out out-of-order processing, to upset recombination convenient for subsequent training sample and the selection for verifying sample.Place is normalized
The purpose of reason is that the data of different range are placed under a referential to be compared, and as subsequent sorting algorithm is made
Prepare.
In one embodiment of the invention, the mathematical feature for extracting each magnetic distribution matrix includes:
S31, according to the magnetic distribution Matrix Solving covariance matrix G in training set:
Wherein, Di(i ∈ [1,2 ..., K]) it is magnetic distribution matrix in training set；K is the electromagnetic field point in training set
The quantity of cloth matrix；For all magnetic distribution matrix DsiAverage value；T is the transposition of matrix；
S32, the corresponding feature vector of maximum P characteristic value in covariance matrix G is chosen:
Wherein,For the characteristic value of covariance matrix G a column, a ∈ (1,2 ..., N), N are magnetic distribution matrix Di
Columns, λ is constant, be equal to 0.9 or 0.95；
S33, descending sort is carried out to feature vector according to the corresponding characteristic value size of P feature vector selected, it is raw
At best projection axis XO:
Wherein,For k-th of feature vector after sequence, k ∈ (1,2 ..., P)；
S34, according to best projection XOCalculate the eigenmatrix F of each pcb board in training seti:
Fi=DiXO；
S35, eigenmatrix F is calculatediIn each column vector two norms, obtain two norm matrix Ls of each trained pcb boardi:
Wherein,It is characterized matrix LiJth column,Fi jFor sample characteristics matrix FiJth column,
ForTwo norms.
In step s 4, pass through mathematical feature (the two norm matrixes of electromagnetic field value) for choosing preset ratio in training set mention
The magnetic distribution matrix taken has supervision neural network to be trained BP as training sample, obtains the nerve net of initial training
Network；In the training process according to known label classification, the error between desired output and reality output is obtained, not according to error
Disconnected adjustment BP has weight and the biasing of supervision neural network, and finally obtaining trained has supervision neural network.
In one embodiment of the invention, the electromagnetic field of preset ratio in training set extracted by mathematical feature is chosen
Distribution matrix has supervision neural network to be trained BP as training sample:
S41, initialization BP have weight, biasing, learning rate and frequency of training in supervision neural network；
S42, operation is carried out to input node using excitation function, obtains the output that BP has supervision neural network hidden layer
Hj:
Wherein, i=1 ... n, j=1 ... l, n are the node number of input layer, and l is the node number of hidden layer；wijFor input
Layer arrives the weight of hidden layer, ajFor the biasing of input layer to hidden layer, xiFor the input data of i-th of node；G () is excitation letter
Number；
S43, it is calculated according to the biasing of the output of hidden layer, the weight of hidden layer to output layer and hidden layer to output layer
The output result O of output layerk:
Wherein, OkTo export result；K=1 ... m, m are the node number of output layer；bkFor hidden layer to the inclined of output layer
It sets.
S44, error amount is calculated according to the output result of output layer and the desired output of training sample:
Wherein, E is error amount；YkFor desired output.
S45, using gradient descent method update hidden layer to output layer weight and input layer to hidden layer weight, wherein
Hidden layer to output layer weight more new formula are as follows:
wjk=wjk+ηHjek
Input layer to hidden layer weight more new formula are as follows:
Wherein, η is learning rate；ek=Yk-Ok；wjkFor the weight of hidden layer to output layer；
S46, hidden layer is updated to the biasing of output layer, update the biasing of input layer to hidden layer again later:
S47, judge that BP has whether supervision neural network is completed training:
When the difference between adjacent error twice is less than given threshold or the number of iterations is equal to preset times, obtain preliminary
Trained neural network；
When between adjacent error twice difference and the number of iterations condition is not satisfied when, return step S42.
In step s 5, using the magnetic distribution matrix extracted by mathematical feature remaining in training set as verifying sample
In the neural network of this input initial training, and export the verifying accuracy distorted；In particular:
The magnetic distribution matrix that remainder has already passed through mathematical feature extraction is conveyed to the neural network tentatively obtained, mind
It can be analyzed through network whether belong to and distort plate, compare further according to known label, be made whether to verify correct judgement, obtain
To the verifying accuracy distorted.
In step s 6, judge to verify whether accuracy reaches preset requirement, if reaching, complete the instruction of neural network
Practice, and enter step S7, otherwise, enables M=M+1, N=N+1, and return step S2；
In the step s 7, test magnetic distribution matrix M × N of pcb board to be tested is obtained, and to test magnetic distribution
Matrix is successively normalized and mathematics feature extraction；The line number M and column of test magnetic distribution matrix in this step
Number N is updated last value.
In step s 8, using trained neural network to the test magnetic distribution matrix extracted by mathematical feature
It is identified, and the recognition result whether distorted is exported, complete PCB tampering detection.
The detection effect of this programme is illustrated with a specific example below:
There is supervision neural network to the plank that do not distorted by BP using MATLAB in this example and has carried out
The plank distorted is detected.
The circuit diagram used in this example as shown in Fig. 2, plate figure as shown in figure 3, the cutoff frequency of the circuit is
2kHZ, the low-pass filter that voltage gain is 2, has carried out 50 groups of instructions to not distorted and being carried out various types of distort
Practice sample, 30 groups of test samples.
Random manufacturing error, mode such as 1 institute of table of error addition are added to the various parameters in pcb board in this example
Show.Mode is as shown in table 2, and testing result is as shown in table 3 for the distorting of distorting of different physics carried out to pcb board.
The mode of 1 dual platen of table addition foozle
What 2 difference of table distorted type distorts mode
BP neural network testing result of the table 3 based on PCB Electromagnetic Environmental Effect
From table 3 it can be seen that this programme provide method can to covered on circuit board copper face product distort and component electricity
Resistance and capacitor the identification distorted, especially to it is therein cover copper face product distort and the detection with higher of electrical part capacitor
Rate.
Claims (5)
1. the PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method characterized by comprising
S1, multiple pcb board composing training collection for having the pcb board distorted and nothing to distort are chosen；
S2, the magnetic distribution that all pcb boards in training set are measured using probe, obtaining line number is M, and columns is the electromagnetic field of N
Distribution matrix；
S3, label is added to magnetic distribution matrix each in training set, and successively carries out out-of-order and normalized, extracted later
The mathematical feature of each magnetic distribution matrix；
The magnetic distribution matrix of preset ratio extracted by mathematical feature is as training sample to BP in S4, selection training set
There is supervision neural network to be trained, obtains the neural network of initial training；
S5, it is tentatively instructed the magnetic distribution matrix extracted by mathematical feature remaining in training set as the input of verifying sample
In experienced neural network, and export the verifying accuracy distorted；
S6, judge to verify whether accuracy reaches preset requirement, if reaching, complete the training of neural network, and enter step
Otherwise S7 enables M=M+1, N=N+1, and return step S2；
S7, the test magnetic distribution matrix M × N for obtaining pcb board to be tested, and test magnetic distribution matrix is successively carried out
Normalized and mathematics feature extraction；And
S8, the test magnetic distribution matrix extracted by mathematical feature is identified using trained neural network, and
The recognition result whether output distorts completes PCB tampering detection.
2. the PCB according to claim 1 based on Electromagnetic Environmental Effect distorts intelligent detecting method, which is characterized in that described
The mathematical feature for extracting each magnetic distribution matrix further comprises:
S31, according to the magnetic distribution Matrix Solving covariance matrix G in training set:
Wherein, Di(i ∈ [1,2 ..., K]) it is magnetic distribution matrix in training set；K is the magnetic distribution square in training set
The quantity of battle array；For all magnetic distribution matrix DsiAverage value；T is the transposition of matrix；
S32, the corresponding feature vector of maximum P characteristic value in covariance matrix G is chosen:
Wherein,For the characteristic value of covariance matrix G a column, a ∈ (1,2 ..., N), N are magnetic distribution matrix DiColumn
Number, λ is constant, is equal to 0.9 or 0.95；
S33, descending sort is carried out to feature vector according to the corresponding characteristic value size of P feature vector selected, generated most
Good axis of projection XO:
Wherein,For k-th of feature vector after sequence, k ∈ (1,2 ..., P)；
S34, according to best projection XOCalculate the eigenmatrix F of each pcb board in training seti:
Fi=DiXO；
S35, eigenmatrix F is calculatediIn each column vector two norms, obtain two norm matrix Ls of each trained pcb boardi:
Wherein,It is characterized matrix LiJth column,Fi jFor sample characteristics matrix FiJth column,For Fi j
Two norms.
3. the PCB according to claim 1 based on Electromagnetic Environmental Effect distorts intelligent detecting method, which is characterized in that described
The magnetic distribution matrix extracted by mathematical feature for choosing preset ratio in training set has supervision to BP as training sample
Neural network, which is trained, further comprises:
S41, initialization BP have weight, biasing, learning rate and frequency of training in supervision neural network；
S42, operation is carried out to input node using excitation function, obtains the output H that BP has supervision neural network hidden layerj:
Wherein, i=1 ... n, j=1 ... l, n are the node number of input layer, and l is the node number of hidden layer；wijIt is arrived for input layer
The weight of hidden layer, ajFor the biasing of input layer to hidden layer, xiFor the input data of i-th of node；G () is excitation function；
S43, output is calculated according to the biasing of the output of hidden layer, the weight of hidden layer to output layer and hidden layer to output layer
The output result of layer；
S44, error amount is calculated according to the output result of output layer and the desired output of training sample；
S45, using gradient descent method update hidden layer to output layer weight and input layer to hidden layer weight, wherein implying
Layer arrives the more new formula of the weight of output layer are as follows:
wjk=wjk+ηHjek
Input layer to hidden layer weight more new formula are as follows:
Wherein, η is learning rate；ek=Yk-Ok；wjkFor the weight of hidden layer to output layer；
S46, hidden layer is updated to the biasing of output layer, update the biasing of input layer to hidden layer again later:
S47, judge that BP has whether supervision neural network is completed training:
When the difference between adjacent error twice is less than given threshold or the number of iterations is equal to preset times, initial training is obtained
Neural network；
When between adjacent error twice difference and the number of iterations condition is not satisfied when, return step S42.
4. the PCB according to claim 3 based on Electromagnetic Environmental Effect distorts intelligent detecting method, which is characterized in that described
The output result O of output layerkCalculation formula are as follows:
Wherein, OkTo export result；K=1 ... m, m are the node number of output layer；bkFor the biasing of hidden layer to output layer.
5. the PCB according to claim 4 based on Electromagnetic Environmental Effect distorts intelligent detecting method, which is characterized in that described
The calculation formula of error amount are as follows:
Wherein, E is error amount；YkFor desired output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910446026.1A CN110118926A (en) | 2019-05-27 | 2019-05-27 | PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910446026.1A CN110118926A (en) | 2019-05-27 | 2019-05-27 | PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110118926A true CN110118926A (en) | 2019-08-13 |
Family
ID=67523370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910446026.1A Pending CN110118926A (en) | 2019-05-27 | 2019-05-27 | PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110118926A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831349A (en) * | 2012-08-23 | 2012-12-19 | 武汉大学 | Characteristic value processing method for hardware Trojan detection |
| CN105378751A (en) * | 2013-07-11 | 2016-03-02 | 科瑞坡特拉股份公司 | Tamper responsive sensor |
| CN106845286A (en) * | 2016-12-15 | 2017-06-13 | 北京邮电大学 | Hardware Trojan horse detection method, apparatus and system |
| CN107194625A (en) * | 2017-07-25 | 2017-09-22 | 国家电网公司 | Wind power plant based on neutral net abandons wind-powered electricity generation amount appraisal procedure |
| CN108107349A (en) * | 2017-12-26 | 2018-06-01 | 上海华岭集成电路技术股份有限公司 | A kind of wafer and the general loading plate of finished product test |
| CN108107248A (en) * | 2017-12-12 | 2018-06-01 | 宁波三星医疗电气股份有限公司 | A kind of stealing recognition methods based on neutral net |
| CN108718297A (en) * | 2018-04-27 | 2018-10-30 | 广州西麦科技股份有限公司 | Ddos attack detection method, device, controller and medium based on BP neural network |
| US20180313877A1 (en) * | 2017-04-26 | 2018-11-01 | Nokomis, Inc. | Electronics equipment testing apparatus and method utilizing unintended rf emission features |
| CN108828325A (en) * | 2018-04-23 | 2018-11-16 | 电子科技大学 | Hardware Trojan horse detection method based on FPGA Clock Tree electromagnetic radiation field |
| CN108847244A (en) * | 2018-08-22 | 2018-11-20 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Voiceprint recognition method and system based on MFCC and improved BP neural network |
| CN109272502A (en) * | 2018-09-28 | 2019-01-25 | 电子科技大学 | PCB hardware security detection method based on temperature field-effect |
-
2019
- 2019-05-27 CN CN201910446026.1A patent/CN110118926A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831349A (en) * | 2012-08-23 | 2012-12-19 | 武汉大学 | Characteristic value processing method for hardware Trojan detection |
| CN105378751A (en) * | 2013-07-11 | 2016-03-02 | 科瑞坡特拉股份公司 | Tamper responsive sensor |
| CN106845286A (en) * | 2016-12-15 | 2017-06-13 | 北京邮电大学 | Hardware Trojan horse detection method, apparatus and system |
| US20180313877A1 (en) * | 2017-04-26 | 2018-11-01 | Nokomis, Inc. | Electronics equipment testing apparatus and method utilizing unintended rf emission features |
| CN107194625A (en) * | 2017-07-25 | 2017-09-22 | 国家电网公司 | Wind power plant based on neutral net abandons wind-powered electricity generation amount appraisal procedure |
| CN108107248A (en) * | 2017-12-12 | 2018-06-01 | 宁波三星医疗电气股份有限公司 | A kind of stealing recognition methods based on neutral net |
| CN108107349A (en) * | 2017-12-26 | 2018-06-01 | 上海华岭集成电路技术股份有限公司 | A kind of wafer and the general loading plate of finished product test |
| CN108828325A (en) * | 2018-04-23 | 2018-11-16 | 电子科技大学 | Hardware Trojan horse detection method based on FPGA Clock Tree electromagnetic radiation field |
| CN108718297A (en) * | 2018-04-27 | 2018-10-30 | 广州西麦科技股份有限公司 | Ddos attack detection method, device, controller and medium based on BP neural network |
| CN108847244A (en) * | 2018-08-22 | 2018-11-20 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Voiceprint recognition method and system based on MFCC and improved BP neural network |
| CN109272502A (en) * | 2018-09-28 | 2019-01-25 | 电子科技大学 | PCB hardware security detection method based on temperature field-effect |
Non-Patent Citations (2)
| Title |
|---|
| HEJASE JA ET.AL: "A Neural Network Based Method for Predicting PCB Glass Weave Induced Skew", 《IEEE》 * |
| 赵毅强 等: "基于自组织竞争神经网络的硬件木马检测方法", 《华中科技大学学报（自然科学版）》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105184778B (en) | A kind of detection method and device | |
| CN106443379B (en) | The recognition methods of partial discharge of transformer fault type and device | |
| CN102663391B (en) | Image multifeature extraction and fusion method and system | |
| US4028674A (en) | Automated signature verification system | |
| CN109116203A (en) | Power equipment partial discharges fault diagnostic method based on convolutional neural networks | |
| CN109272502A (en) | PCB hardware security detection method based on temperature field-effect | |
| CN109165510A (en) | Android malicious application detection method based on binary channels convolutional neural networks | |
| CN109858509A (en) | Based on multilayer stochastic neural net single classifier method for detecting abnormality | |
| CN109100627A (en) | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode | |
| CN109002859B (en) | Sensor array feature selection and array optimization method based on principal component analysis | |
| CN110458195A (en) | A kind of abnormal data discriminating method based on the fusion of more criterions | |
| CN109344835A (en) | Altering detecting method based on vehicle VIN code character position | |
| Miyajima et al. | Machine learning detection of Berezinskii-Kosterlitz-Thouless transitions in q-state clock models | |
| CN108564569B (en) | A kind of distress in concrete detection method and device based on multicore classification learning | |
| CN110118926A (en) | PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method | |
| CN103824083B (en) | Web online species recognition method based on blade complete and partial two-value characteristics | |
| CN109376738A (en) | Altering detecting method based on vehicle VIN code start-stop symbol type | |
| CN108776801A (en) | It is a kind of based on owing to determine the analog circuit fault features extracting method of blind source separating | |
| CN116466408A (en) | Artificial neural network superbedrock identification method based on aeromagnetic data | |
| CN108680595B (en) | Low-field magnetic resonance edible oil type detection method | |
| CN113960152B (en) | Trace explosive identification method and system based on ion mobility spectrometry | |
| Pushpalatha et al. | Offline signature verification based on contourlet transform and textural features using HMM | |
| CN105760872A (en) | Identification method and system based on robust image feature extraction | |
| CN109993071A (en) | The method and system of discoloration forest automatic identification and investigation based on remote sensing image | |
| CN108900538A (en) | A kind of industry control signal detecting method and device |
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 |
Application publication date: 20190813 |
|
| RJ01 | Rejection of invention patent application after publication |