CN109272502A - PCB hardware safety detection method based on temperature field effect - Google Patents

PCB hardware safety detection method based on temperature field effect Download PDF

Info

Publication number
CN109272502A
CN109272502A CN201811138392.2A CN201811138392A CN109272502A CN 109272502 A CN109272502 A CN 109272502A CN 201811138392 A CN201811138392 A CN 201811138392A CN 109272502 A CN109272502 A CN 109272502A
Authority
CN
China
Prior art keywords
matrix
pcb
pcb board
neural network
node
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
Application number
CN201811138392.2A
Other languages
Chinese (zh)
Other versions
CN109272502B (en
Inventor
王坚
杨文秀
杨鍊
陈哲
李桓
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201811138392.2A priority Critical patent/CN109272502B/en
Publication of CN109272502A publication Critical patent/CN109272502A/en
Application granted granted Critical
Publication of CN109272502B publication Critical patent/CN109272502B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Storage Device Security (AREA)
  • Investigating Or Analyzing Materials Using Thermal Means (AREA)

Abstract

本发明公开了一种基于温度场效应的PCB硬件安全检测方法，首先获取训练集中PCB板的温度热图矩阵，并对其进行2DPCA分析，利用分析结果训练SOM神经网络，得到一个训练好的SOM神经网络；其次获取待测PCB板的温度热图矩阵，并对其进行2DPCA分析，然后交由训练好的SOM神经网络进行分类检测，最后分辨出待测PCB板是否被篡改。本发明为PCB物理上的恶意篡改提供了一种有效的检测方法，适用于PCB上的各种类型的物理性篡改，并且SOM神经网络对于元器件以及覆铜面积篡改具有较高的检测率，因此本发明具有检测率高、易于实现的特点，能够以较高概率检测到PCB板遭受到的恶意篡改。

The invention discloses a PCB hardware safety detection method based on the temperature field effect. First, the temperature heat map matrix of the PCB board in the training set is obtained, and 2DPCA analysis is performed on it, and the SOM neural network is trained by using the analysis results to obtain a trained SOM Neural network; secondly, obtain the temperature heat map matrix of the PCB board to be tested, and perform 2DPCA analysis on it, and then send it to the trained SOM neural network for classification and detection, and finally distinguish whether the PCB board to be tested has been tampered with. The present invention provides an effective detection method for physically malicious tampering of PCBs, and is applicable to various types of physical tampering on PCBs, and the SOM neural network has a high detection rate for tampering of components and copper clad areas, Therefore, the present invention has the characteristics of high detection rate and easy implementation, and can detect malicious tampering suffered by the PCB board with a high probability.

Description

PCB hardware security detection method based on temperature field-effect
Technical field
The invention belongs to PCB hardware security technical fields, and in particular to a kind of PCB hardware security based on temperature field-effect The design of detection method.
Background technique
It is that artificial malice distorts component, line width, the via hole etc. on PCB that the malice of PCB, which is distorted, it can change Become circuit function, reveal circuit information or make circuit refusal service etc., this constitutes hardware security and greatly threatens.So And due to this malice distort it is easy to operate, small in size, so detection get up it is extremely difficult.Currently, academia have much about The detection of hardware Trojan horse in chip, detection method are also to emerge one after another.Common methods have logic testing, test design, side letter Trace analysis, reverse-engineering etc..
Hardware Trojan horse for chip-scale is mainly distorting in logic, and detection is relatively easy to, and for higher level PCB Some hardware Trojan horses of plate grade are mainly that some malice physically are distorted, and detection is got up extremely difficult.At present for pcb board grade The detection method distorted of some malice also not it has been proposed that, most of research about PCB hardware Trojan horse rests on wooden horse and makes At serious consequence on.Such as by tentatively finding to the wooden horse classification and the analysis and introduction of some challenge models on PCB Some possible detection countermeasures；Or by being manufactured to some existing intelligent electronic devices common in people's lives The analysis of process, finds that some there may be the loopholes of safety issue；Or some features by equipment, it carries out equipment and recognizes Card, but these analyze relevant detection technique not yet for the hardware security of PCB physically.
Summary of the invention
The purpose of the present invention is existing PCB physically hardware security analysis lack correlation detection technology aiming at the problem that, Propose a kind of PCB hardware security detection method based on temperature field-effect, can be detected with high probability pcb board by Malice distort.
The technical solution of the present invention is as follows: the PCB hardware security detection method based on temperature field-effect, comprising the following steps:
S1, multiple pcb boards for having the pcb board distorted and nothing to distort are chosen as training pcb board, composing training collection.
S2, temperature value acquisition is carried out to each pcb board in training set, obtains the temperature thermal map square of each pcb board in training set Battle array.
S3,2DPCA two-dimensional principal component analysis is carried out to the temperature thermal map matrix of pcb board each in training set, obtains training set In each pcb board temperature value two norm matrixes.
S4, SOM neural network is trained according to two norm matrixes of pcb board temperature value each in training set, is instructed The SOM neural network perfected.
S5, multiple pcb boards to be measured are chosen, constitutes test set.
S6, temperature value acquisition is carried out to the pcb board to be measured in test set, obtains the temperature thermal map matrix of each pcb board to be measured.
S7,2DPCA two-dimensional principal component analysis is carried out to the temperature thermal map matrix of all pcb boards to be measured, it is each obtains test set Two norm matrixes of pcb board temperature value.
S8, it is identified using two norm matrixes of the trained SOM neural network to each pcb board temperature value of test set, Judge whether each pcb board to be measured is distorted, completes the safety detection to PCB hardware.
The beneficial effects of the present invention are: the present invention distorts for the malice of PCB physically provides a kind of effective detection side Method, suitable on PCB it is various types of it is physical distort, and SOM neural network for component and cover copper face product usurp Change verification and measurement ratio with higher, has the characteristics that verification and measurement ratio is high, is easily achieved, therefore the present invention can be detected with high probability The malice that pcb board is subjected to is distorted.
Detailed description of the invention
Fig. 1 show the PCB hardware security detection method flow chart provided in an embodiment of the present invention based on temperature field-effect.
Fig. 2 show low-pass filter circuit schematic diagram provided in an embodiment of the present invention.
Fig. 3 show low-pass filter pcb board figure provided in an embodiment of the present invention.
Fig. 4 show the pcb board figure that GND cabling provided in an embodiment of the present invention is nearby added to via hole.
Specific embodiment
Carry out detailed description of the present invention illustrative embodiments with reference to the drawings.It should be appreciated that shown in attached drawing and The embodiment of description is only exemplary, it is intended that is illustrated the principle and spirit of the invention, and is not limited model of the invention It encloses.
The PCB hardware security detection method based on temperature field-effect that the embodiment of the invention provides a kind of, as shown in Figure 1, Include the following steps S1-S8:
S1, multiple pcb boards for having the pcb board distorted and nothing to distort are chosen as training pcb board, composing training collection.
S2, temperature value acquisition is carried out to each pcb board of training set, obtains the temperature thermal map matrix of each pcb board of training set.
S3,2DPCA two-dimensional principal component analysis is carried out to the temperature thermal map matrix of each pcb board in training set, is trained Concentrate two norm matrixes of each pcb board temperature value.To the temperature thermal map matrixes of all trained pcb boards carry out 2DPCA two dimension it is main at Analysis, its purpose is to make feature more obvious, it is easier to be distinguished out.
Step S3 includes following S31-S35 step by step:
S31, the temperature thermal map matrix D according to pcb board each in training setiCalculate covariance matrix G, calculation formula are as follows:
Wherein K is the quantity of pcb board in training set,For the temperature thermal map matrix D of pcb board each in training setiBe averaged Value, the transposition of subscript T representing matrix.
S32, the corresponding feature vector of maximum P characteristic value in covariance matrix G is chosen；The determination formula of P value are as follows:
WhereinIndicate the characteristic value of covariance matrix G a column, a ∈ (1,2 ..., N), N are temperature thermal map matrix Di's Columns.In the embodiment of the present invention, each plank has the temperature value for measuring and acquiring on 1296 points altogether, then N=36, i.e. matrix Di For one 36 × 36 temperature value matrix.
S33, descending sort is carried out to feature vector according to the corresponding characteristic value size of P feature vector selected, obtained To best projection axis XO:
WhereinIndicate k-th of feature vector after sorting, k ∈ (1,2 ..., P).
S34, according to best projection axis XOCalculate the eigenmatrix F of each pcb board in training seti, calculation formula are as follows:
Fi=DiXO (4)
Wherein i ∈ (1,2 ..., K).
S35, eigenmatrix F is calculatediIn each column vector two norms, obtain two norm matrix Ls of each trained pcb boardi:
WhereinIndicate eigenmatrix FiIn j-th of column vector two norms, calculation formula are as follows:
Wherein Fi jIndicate eigenmatrix FiJth column, j ∈ (1,2 ..., Q), Q is characterized matrix FiColumns.
Due to original temperature thermal map matrix DiFor a three-dimensional matrice, and obtained after 2DPCA two-dimensional principal component analysis Two norm matrix LsiFor a two-dimensional matrix, therefore the matrix after 2DPCA two-dimensional principal component analysis easily facilitates differentiation.
S4, SOM neural network is trained according to two norm matrixes of pcb board temperature value each in training set, is instructed The SOM neural network perfected.In the embodiment of the present invention, rule is found to input data using SOM unsupervised neural network and is returned Class realizes in the class of input data separation property between similitude and class.
Step S4 includes following S41-S46 step by step:
S41, the two norm matrix Ls to pcb board each in training setiIt is normalized and random ordering is handled, obtain input number According to X={ xn, n=1 ..., D }, D is the dimension of input data.In the embodiment of the present invention, D=3.
The node weights w of S42, random initializtion SOM neural network topology structurebnAnd its weighted error Δ wbn, wherein wbn Indicate the n-th dimensional weight vector of b-th of node of SOM neural network.
The most matched node of S43, selection and input data X is as activation node I (X).
In the embodiment of the present invention, due to being unsupervised mode, do not have for the vector of one group of input desired defeated Outgoing label can only carry out cluster classification by similitude, and similitude is usually exactly to be distinguished by the measurement of distance , commonly used range measurement is Euclidean distance, therefore using Euclidean distance as discriminant function, calculates each The Euclidean distance d of node b and input data Xb(X):
Select Euclidean distance db(X) the smallest node of value is made as with the most matched node of input data X To activate node I (X).
S44, the weight for updating activation node I (X), more new formula are as follows:
Wherein Tm,I(X)Indicate the weight of activation node I (X), Sm,I(X)Indicate activation node I (X) adjacent node m it Between distance, σ be activate node I (X) neighborhood size.
S45, according to the weight T of the currently active node I (X)m,I(X), update weighted error, more new formula are as follows:
Δwmn=η (t) Tm,I(X)·(xn-wmn) (8)
Wherein wmnIndicate the n-th dimensional weight vector of node m, Δ wmnFor its weighted error, η (t) is SOM neural network Learning rate.
S46, judge whether iteration restrains, if then obtaining updated SOM neural network, and enter step S47, otherwise Return step S43 carries out next iteration training.
In the embodiment of the present invention, the condition of iteration convergence are as follows:
Weighted error Δ wmnReach pre-set error threshold or the number of iterations reaches pre-set the number of iterations Threshold value.In the embodiment of the present invention, the number of iterations threshold value needs to be arranged enough to big.
S47, after the temperature thermal map matrix of standard gold pcb board is carried out 2DPCA two-dimensional principal component analysis, it is input to update SOM neural network afterwards distinguishes no tampering class pcb board and has tampering class pcb board, obtains trained SOM neural network.
2DPCA and SOM nerve is carried out by temperature thermal map matrix of the step S3 and S41~S46 to each pcb board of training set After network analysis, each pcb board temperature thermal map matrix of training set can be divided into two classes, but by then passing through Unsupervised clustering analysis Obtained classification, cannot be distinguished specific classification, (i.e. which kind of is no tampering class pcb board, which kind of is that have tampering class PCB Plate), therefore also need the temperature thermal map matrix of standard gold pcb board carrying out 2DPCA two dimension principal component in the embodiment of the present invention After analysis, it is input to updated SOM neural network, distinguish no tampering class pcb board and has tampering class pcb board, is trained Good SOM neural network.
S5, multiple pcb boards to be measured are chosen, constitutes test set.
S6, temperature value acquisition is carried out to the pcb board to be measured in test set, obtains the temperature thermal map matrix of each pcb board to be measured.
S7,2DPCA two-dimensional principal component analysis is carried out to the temperature thermal map matrix of all pcb boards to be measured, it is each obtains test set Two norm matrixes of pcb board temperature value.2DPCA two-dimensional principal component analysis is carried out to the temperature thermal map matrix of all pcb boards to be measured, Its purpose is to make feature more obvious, it is easier to be distinguished out.
In the embodiment of the present invention, step S7 uses method identical with step S3, specifically, step S7 includes following point Step S71-S75:
S71, the temperature thermal map matrix A according to each pcb board to be measuredlCalculate covariance matrix G ', calculation formula are as follows:
Wherein H is the quantity of pcb board to be measured,For the temperature thermal map matrix A of each pcb board to be measuredlAverage value, subscript T The transposition of representing matrix.
S72, the corresponding feature vector of maximum R characteristic value in covariance matrix G ' is chosen；The determination formula of R value are as follows:
WhereinIndicate the characteristic value of c column in covariance matrix G ', c ∈ (1,2 ..., M), M are temperature thermal map matrix AlColumns.
S73, descending sort is carried out to feature vector according to the corresponding characteristic value size of R feature vector selected, obtained To best projection axis X 'O:
WhereinIndicate h-th of feature vector after sorting, h ∈ (1,2 ..., R).
S74, according to best projection axis X 'OCalculate the eigenmatrix F of each pcb board to be measuredl, calculation formula are as follows:
Fl=AlX′O (13)
Wherein l ∈ (1,2 ..., H).
S75, eigenmatrix F is calculatedlIn each column vector two norms, obtain two norm matrix Ls of each pcb board to be measuredl:
WhereinIndicate eigenmatrix FlIn d-th of column vector two norms, calculation formula are as follows:
Wherein Fl dIndicate eigenmatrix FlD column, d ∈ (1,2 ..., V), V is characterized matrix FlColumns.
S8, it is identified using two norm matrixes of the trained SOM neural network to each pcb board temperature value of test set, Judge whether each pcb board to be measured is distorted, completes the safety detection to PCB hardware.
By step S47, no tampering class pcb board has been distinguished in trained SOM neural network and has had tampering class Pcb board can be detected out then by after the trained SOM neural network of two norm Input matrixes of each pcb board temperature value of test set Whether each pcb board to be measured is distorted.
In the embodiment of the present invention, it is necessary first to two norm matrix Ls of each pcb board to be measuredlIt is normalized and out-of-order Processing, the Input matrix then obtained trained SOM neural network, goes out temperature thermal map by neural network recognization In include distort information, finally determine whether each pcb board to be measured is distorted.
Detection effect of the invention is described in detail with a specific embodiment below:
The PCB hardware security detection method based on temperature field-effect provided according to the present invention usurps nothing using MATLAB Change PCB and carry out the PCB that different type is distorted and is detected.
In the embodiment of the present invention, the pcb board detected is that a cutoff frequency is 2KHZ, the low pass that voltage gain is 2 Filter plate, as shown in Figure 2.The plate figure of the pcb board as shown in figure 3, pcb board standard size are as follows: 15.24cm × 10.16cm ×0.392mm.Due to not using via hole on normal pcb board, also need to add the detection that via hole carries out via hole size, The plate figure of via hole is added to as shown in figure 4, the point of addition of via hole is near GND network.
Need to add the various parameters in pcb board random manufacturing error, the mode of error addition in the embodiment of the present invention As shown in table 1.
Table 1
The error that table 1 will appear primarily with respect to each parameter of the plate mill in practice when manufacturing dual platen on plank Always there are some foozles in the fabrication process in tolerance values, during the test take into account these random noises It goes, can more meet practical situation, obtained testing result can be closer to actual result.
In the embodiment of the present invention, what difference distorted type distorts that mode is as shown in table 2, and testing result is as shown in table 3.
Due to mainly detecting to distorting for pcb board in invention, and table 2 is to low-pass filter shown in Fig. 2 The actual conditions of PCB modification, the normal pcb board situation that the data that do not distort are not distorted mainly, each data after distorting are main It is the data situation specifically distorted, the type to be distorted only is distorted when distorting every time, component is selected Value be actual common component values, and with the data do not distorted very close to value, can be become apparent from by table 2 Find out which the present invention mainly distort type and detect to.
Table 2
Table 3
Table 3 is the corresponding testing result respectively distorting type and obtaining of table 2, by testing result it can be found that for component with And it is relatively high for covering the accuracy for the detection of copper face product distorted.Testing result in the embodiment of the present invention is based on 560 groups What above sample obtained, without distorted and carried out component, cover copper face product and when distorting of trace width, There is plate thickness error and has carried out 30 groups of training samples, 18 groups of test samples without plate thickness error.The presence or absence of via hole size is usurped Change 20 groups of training samples, 12 groups of test samples.Last testing result is shown, distorts and cover copper face product for component values Distort can accurate detection come out.It is by the data in table 3 it is found that provided in an embodiment of the present invention based on temperature field-effect The verification and measurement ratio that PCB hardware security detection method obtains is higher, and be suitable on PCB various types of physical distorts.
Those of ordinary skill in the art will understand that the embodiments described herein, which is to help reader, understands this hair Bright principle, it should be understood that protection scope of the present invention is not limited to such specific embodiments and embodiments.This field Those of ordinary skill disclosed the technical disclosures can make according to the present invention and various not depart from the other each of essence of the invention The specific variations and combinations of kind, these variations and combinations are still within the scope of the present invention.

Claims (6)

1.基于温度场效应的PCB硬件安全检测方法，其特征在于，包括以下步骤：A PCB hardware security detection method based on temperature field effect, characterized in that it comprises the following steps: S1、选取多个有篡改的PCB板和无篡改的PCB板作为训练PCB板，构成训练集；S1, selecting a plurality of tamper-evident PCB boards and tamper-free PCB boards as training PCB boards to form a training set; S2、对训练集中的各PCB板进行温度值采集，得到训练集中各PCB板的温度热图矩阵；S2, collecting temperature values of each PCB board in the training set, and obtaining a temperature heat map matrix of each PCB board in the training set; S3、对训练集中各PCB板的温度热图矩阵进行2DPCA二维主成分分析，得到训练集中各PCB板温度值的二范数矩阵；S3, performing 2DPCA two-dimensional principal component analysis on the temperature heat map matrix of each PCB board in the training set, and obtaining a two-norm matrix of temperature values of each PCB board in the training set; S4、根据训练集中各PCB板温度值的二范数矩阵对SOM神经网络进行训练，得到训练好的SOM神经网络；S4. Training the SOM neural network according to the two norm matrix of the temperature values of the PCB boards in the training set, and obtaining the trained SOM neural network; S5、选取多个待测PCB板，构成测试集；S5, selecting a plurality of PCB boards to be tested to form a test set; S6、对测试集中的待测PCB板进行温度值采集，得到各待测PCB板的温度热图矩阵；S6. Perform temperature value collection on the PCB to be tested in the test set, and obtain a temperature heat map matrix of each PCB board to be tested; S7、对所有待测PCB板的温度热图矩阵进行2DPCA二维主成分分析，得到测试集各PCB板温度值的二范数矩阵；S7, performing 2DPCA two-dimensional principal component analysis on the temperature heat map matrix of all the PCB boards to be tested, and obtaining a two-norm matrix of temperature values of each PCB board in the test set; S8、采用训练好的SOM神经网络对测试集各PCB板温度值的二范数矩阵进行识别，判断各待测PCB板是否发生了篡改，完成对PCB硬件的安全检测。S8. Using the trained SOM neural network to identify the two norm matrix of the temperature values of the PCBs in the test set, determine whether the PCB boards to be tested have been tampered, and complete the security detection of the PCB hardware. 2.根据权利要求1所述的PCB硬件安全检测方法，其特征在于，所述步骤S3包括以下分步骤：The PCB hardware security detecting method according to claim 1, wherein the step S3 comprises the following sub-steps: S31、根据训练集中各PCB板的温度热图矩阵Di计算协方差矩阵G，计算公式为：S31. Calculate a covariance matrix G according to a temperature heat map matrix D i of each PCB board in the training set, and the calculation formula is: 其中K为训练集中PCB板的数量，为训练集中各PCB板的温度热图矩阵Di的平均值，上标T表示矩阵的转置；Where K is the number of PCB boards in the training set, To train the average of the temperature heat map matrix D i of each PCB board, the superscript T indicates the transposition of the matrix; S32、选取协方差矩阵G中最大的P个特征值对应的特征向量；P值的确定公式为：S32. Select a feature vector corresponding to the largest P eigenvalues in the covariance matrix G; the formula for determining the P value is: 其中表示协方差矩阵G第a列的特征值，a∈(1,2,...,N)，N为温度热图矩阵Di的列数；among them Representing the eigenvalue of the a- th column of the covariance matrix G, a ∈ (1, 2, ..., N), where N is the number of columns of the temperature heat map matrix D i ; S33、根据选取出的P个特征向量对应的特征值大小对特征向量进行降序排序，得到最佳投影轴XO：S33. Sorting the feature vectors in descending order according to the selected feature value corresponding to the P feature vectors to obtain an optimal projection axis X O : 其中表示排序后第k个特征向量，k∈(1,2,...,P)；among them Represents the kth eigenvector after sorting, k∈(1,2,...,P); S34、根据最佳投影轴XO计算训练集中每个PCB板的特征矩阵Fi，计算公式为：S34. Calculate a characteristic matrix F i of each PCB board in the training set according to an optimal projection axis X O , and the calculation formula is: Fi＝DiXO (4)F i =D i X O (4) 其中i∈(1,2,...,K)；Where i∈(1,2,...,K); S35、计算特征矩阵Fi中各列向量的二范数，得到每个训练PCB板的二范数矩阵Li：S35. Calculate a second norm of each column vector in the feature matrix F i to obtain a two norm matrix L i of each training PCB board: 其中表示特征矩阵Fi中第j个列向量的二范数，计算公式为：among them The two norm representing the jth column vector in the feature matrix F i is calculated as: 其中Fi j表示特征矩阵Fi的第j列，j∈(1,2,...,Q)，Q为特征矩阵Fi的列数。Where F i j represents the jth column of the feature matrix F i , j ∈ (1, 2, . . . , Q), and Q is the number of columns of the feature matrix F i . 3.根据权利要求2所述的PCB硬件安全检测方法，其特征在于，所述步骤S4包括以下分步骤：The PCB hardware security detecting method according to claim 2, wherein the step S4 comprises the following sub-steps: S41、对训练集中各个PCB板的二范数矩阵Li进行归一化和乱序处理，得到输入数据X＝{xn,n＝1,...,D}，D为输入数据的维数；S41. Perform normalization and out-of-order processing on the two norm matrix L i of each PCB board in the training set to obtain input data X={x n , n=1, . . . , D}, where D is the dimension of the input data. number; S42、随机初始化SOM神经网络拓扑结构的节点权重wbn及其权重误差Δwbn，其中wbn表示SOM神经网络第b个节点的第n维权重向量；S42. Randomly initialize a node weight w bn of the SOM neural network topology and a weight error Δw bn , where w bn represents an nth-dimensional weight vector of the b-th node of the SOM neural network; S43、选取与输入数据X最匹配的节点作为激活节点I(X)；S43, selecting a node that best matches the input data X as the activation node I(X); S44、更新激活节点I(X)的权重，更新公式为：S44. Update the weight of the active node I(X), and update the formula as: 其中Tm,I(X)表示激活节点I(X)的权重，Sm,I(X)表示激活节点I(X)与其临近节点m之间的距离，σ为激活节点I(X)的邻域的大小；Where T m, I(X) represents the weight of the active node I(X), S m, I(X) represents the distance between the active node I(X) and its neighboring node m, and σ is the active node I(X) The size of the neighborhood; S45、根据当前激活节点I(X)的权重Tm,I(X)，更新权重误差，更新公式为：S45. Update the weight error according to the weights T m, I(X) of the currently activated node I(X), and update the formula as: Δwmn＝η(t)·Tm,I(X)·(xn-wmn) (8)Δw mn =η(t)·T m,I(X) ·(x n -w mn ) (8) 其中wmn表示节点m的第n维权重向量，Δwmn为其权重误差，η(t)为SOM神经网络的学习率；Where w mn represents the n-th weight vector of node m, Δw mn is its weight error, and η(t) is the learning rate of the SOM neural network; S46、判断迭代是否收敛，若是则得到更新后的SOM神经网络，并进入步骤S47，否则返回步骤S43进行下一次迭代训练；S46: determining whether the iteration converges, if yes, obtaining the updated SOM neural network, and proceeding to step S47, otherwise returning to step S43 for the next iteration training; S47、将标准黄金PCB板的温度热图矩阵进行2DPCA二维主成分分析后，输入到更新后的SOM神经网络，区分出无篡改类PCB板和有篡改类PCB板，得到训练好的SOM神经网络。S47. Perform 2DPCA two-dimensional principal component analysis on the temperature heat map matrix of the standard gold PCB board, input it into the updated SOM neural network, distinguish the tamper-free PCB board and the tamper-type PCB board, and obtain the trained SOM nerve. The internet. 4.根据权利要求3所述的PCB硬件安全检测方法，其特征在于，所述步骤S43具体为：The hardware security detection method of the PCB according to claim 3, wherein the step S43 is specifically: 计算每一个节点b与输入数据X的欧几里得距离db(X)：Calculate the Euclidean distance d b (X) of each node b from the input data X: 选择欧几里得距离db(X)值最小的节点作为激活节点I(X)。The node with the smallest Euclide distance d b (X) is selected as the active node I(X). 5.根据权利要求3所述的PCB硬件安全检测方法，其特征在于，所述步骤S46中迭代收敛的条件为：The hardware security detection method of the PCB according to claim 3, wherein the condition of the iterative convergence in the step S46 is: 权重误差Δwmn达到预先设置的误差阈值，或者迭代次数达到预先设置的迭代次数阈值。The weight error Δw mn reaches a preset error threshold, or the number of iterations reaches a preset number of iterations threshold. 6.根据权利要求1所述的PCB硬件安全检测方法，其特征在于，所述步骤S7包括以下分步骤：The PCB hardware security detecting method according to claim 1, wherein the step S7 comprises the following sub-steps: S71、根据各待测PCB板的温度热图矩阵Al计算协方差矩阵G′，计算公式为：S71. Calculate a covariance matrix G′ according to a temperature heat map matrix A l of each PCB board to be tested, and the calculation formula is: 其中H为待测PCB板的数量，为各待测PCB板的温度热图矩阵Al的平均值，上标T表示矩阵的转置；Where H is the number of PCB boards to be tested, For the average value of the temperature heat map matrix A l of each PCB board to be tested, the superscript T indicates the transposition of the matrix; S72、选取协方差矩阵G′中最大的R个特征值对应的特征向量；R值的确定公式为：S72: Select a feature vector corresponding to the largest R eigenvalues in the covariance matrix G′; the formula for determining the R value is: 其中表示协方差矩阵G′中第c列的特征值，c∈(1,2,...,M)，M为温度热图矩阵Al的列数；among them Representing the eigenvalue of the cth column in the covariance matrix G', c ∈ (1, 2, ..., M), M is the number of columns of the temperature heat map matrix A l ; S73、根据选取出的R个特征向量对应的特征值大小对特征向量进行降序排序，得到最佳投影轴X′O：S73. Sorting the feature vectors in descending order according to the selected feature values corresponding to the R feature vectors, to obtain an optimal projection axis X′ O : 其中表示排序后第h个特征向量，h∈(1,2,...,R)；among them Indicates the hth eigenvector after sorting, h∈(1,2,...,R); S74、根据最佳投影轴X′O计算每个待测PCB板的特征矩阵Fl，计算公式为：S74. Calculate a characteristic matrix F l of each PCB board to be tested according to an optimal projection axis X′ O , and the calculation formula is: Fl＝AlX′O (13)F l =A l X' O (13) 其中l∈(1,2,...,H)；Where l∈(1,2,...,H); S75、计算特征矩阵Fl中各列向量的二范数，得到每个待测PCB板的二范数矩阵Ll：S75. Calculate a norm of each column vector in the feature matrix F l to obtain a two norm matrix L l of each PCB to be tested: 其中表示特征矩阵Fl中第d个列向量的二范数，计算公式为：among them The two norm representing the d-th column vector in the feature matrix F l is calculated as: 其中Fl d表示特征矩阵Fl的第d列，d∈(1,2,...,V)，V为特征矩阵Fl的列数。Where F l d represents the d-th column of the feature matrix F l , d ∈ (1, 2, . . . , V), and V is the number of columns of the feature matrix F l .
CN201811138392.2A 2018-09-28 2018-09-28 PCB hardware safety detection method based on temperature field effect Active CN109272502B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811138392.2A CN109272502B (en) 2018-09-28 2018-09-28 PCB hardware safety detection method based on temperature field effect

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811138392.2A CN109272502B (en) 2018-09-28 2018-09-28 PCB hardware safety detection method based on temperature field effect

Publications (2)

Publication Number Publication Date
CN109272502A true CN109272502A (en) 2019-01-25
CN109272502B CN109272502B (en) 2022-05-20

Family

ID=65198728

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811138392.2A Active CN109272502B (en) 2018-09-28 2018-09-28 PCB hardware safety detection method based on temperature field effect

Country Status (1)

Country Link
CN (1) CN109272502B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110556A (en) * 2019-05-27 2019-08-09 电子科技大学 Board vulnerability analysis method based on multiple physical field effect
CN110118926A (en) * 2019-05-27 2019-08-13 电子科技大学 PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method
CN110287737A (en) * 2019-07-05 2019-09-27 电子科技大学 A Hardware Trojan Horse Detection Method Based on Chip Transient Temperature Characteristic Analysis
CN110298204A (en) * 2019-07-05 2019-10-01 电子科技大学 Asic chip hardware Trojan horse diagnostic method based on temperature field tau-effect
CN110298202A (en) * 2019-07-05 2019-10-01 电子科技大学 Hardware back door intelligent diagnosing method based on chip temperature field Space Time feature
CN114579768A (en) * 2022-05-06 2022-06-03 深圳市信润富联数字科技有限公司 Maintenance method for realizing intelligent operation and maintenance knowledge base of equipment

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103198251A (en) * 2013-03-28 2013-07-10 哈尔滨工业大学（威海） Hardware Trojan horse recognition method based on neural network
CN106546628A (en) * 2016-09-22 2017-03-29 西安理工大学 A kind of lossless detection method based on temperature field tomography
CN106815532A (en) * 2015-11-27 2017-06-09 天津科技大学 A kind of hardware Trojan horse based on distance measure distribution detects method of discrimination
CN107703186A (en) * 2017-09-26 2018-02-16 电子科技大学 Hardware Trojan horse detection method based on chip temperature field-effect
WO2018111601A1 (en) * 2016-12-16 2018-06-21 Square, Inc. Tamper detection system
CN108256357A (en) * 2018-01-10 2018-07-06 中国人民解放军国防科技大学 Hardware Trojan Horse Detection Method Combining Infrared Image and Normal Distribution Analysis

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103198251A (en) * 2013-03-28 2013-07-10 哈尔滨工业大学（威海） Hardware Trojan horse recognition method based on neural network
CN106815532A (en) * 2015-11-27 2017-06-09 天津科技大学 A kind of hardware Trojan horse based on distance measure distribution detects method of discrimination
CN106546628A (en) * 2016-09-22 2017-03-29 西安理工大学 A kind of lossless detection method based on temperature field tomography
WO2018111601A1 (en) * 2016-12-16 2018-06-21 Square, Inc. Tamper detection system
CN107703186A (en) * 2017-09-26 2018-02-16 电子科技大学 Hardware Trojan horse detection method based on chip temperature field-effect
CN108256357A (en) * 2018-01-10 2018-07-06 中国人民解放军国防科技大学 Hardware Trojan Horse Detection Method Combining Infrared Image and Normal Distribution Analysis

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
KANGQIAO HU等: "High-sensitivity hardware Trojan detection using multimodal characterization", 《2013 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE)》 *
NING WEN等: "Hardware Trojan Detection Technique Based on SOM Neural Network", 《2018 EIGHTH INTERNATIONAL CONFERENCE ON INSTRUMENTATION AND MEASUREMENT, COMPUTER, COMMUNICATION AND CONTROL》 *
牟江: "暗硅约束下的片上多处理器性能优化策略", 《中国优秀硕士学位论文全文数据库信息科技辑》 *
胡阳春: "基于改进k均值聚类算法的电力负荷模式识别方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110556A (en) * 2019-05-27 2019-08-09 电子科技大学 Board vulnerability analysis method based on multiple physical field effect
CN110118926A (en) * 2019-05-27 2019-08-13 电子科技大学 PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method
CN110110556B (en) * 2019-05-27 2021-03-30 电子科技大学 Board card vulnerability analysis method based on multi-physical field effect
CN110287737A (en) * 2019-07-05 2019-09-27 电子科技大学 A Hardware Trojan Horse Detection Method Based on Chip Transient Temperature Characteristic Analysis
CN110298204A (en) * 2019-07-05 2019-10-01 电子科技大学 Asic chip hardware Trojan horse diagnostic method based on temperature field tau-effect
CN110298202A (en) * 2019-07-05 2019-10-01 电子科技大学 Hardware back door intelligent diagnosing method based on chip temperature field Space Time feature
CN110298202B (en) * 2019-07-05 2021-01-26 电子科技大学 Intelligent diagnosis method for hardware backdoor based on time-space characteristics of chip temperature field
CN114579768A (en) * 2022-05-06 2022-06-03 深圳市信润富联数字科技有限公司 Maintenance method for realizing intelligent operation and maintenance knowledge base of equipment

Also Published As

Publication number Publication date
CN109272502B (en) 2022-05-20

Similar Documents

Publication Publication Date Title
CN109272502A (en) PCB hardware safety detection method based on temperature field effect
US10761127B2 (en) Electronic component classification
Bayar et al. Towards open set camera model identification using a deep learning framework
CN109165671A (en) Confrontation sample testing method based on sample to decision boundary distance
CN109729091A (en) An LDoS attack detection method based on multi-feature fusion and CNN algorithm
CN107577945A (en) URL attack detection methods, device and electronic equipment
CN112668557B (en) Method for defending image noise attack in pedestrian re-identification system
CN109740606A (en) A kind of image-recognizing method and device
Lin PCA/SVM‐based method for pattern detection in a multisensor system
CN110324178A (en) A kind of network inbreak detection method based on mostly through checking study
Temel et al. Object recognition under multifarious conditions: A reliability analysis and a feature similarity-based performance estimation
CN118521858A (en) Aviation deep learning model robustness assessment method for challenge attack
Ahmad et al. Artificial neural network approaches to intrusion detection: a review
CN107330429A (en) A kind of localization method and device of certificate entry
CN112085170A (en) A Differential Detection Method for Image Class Adversarial Samples
CN106407942A (en) Image processing method and device thereof
CN110110556B (en) Board card vulnerability analysis method based on multi-physical field effect
KR101133225B1 (en) System and method for recognizing face using pose estimation
Yang et al. Learning vector quantization neural network method for network intrusion detection
CN109271902A (en) Method for detecting infrared puniness target based on time domain empirical mode decomposition under complex background
CN113537381A (en) Human Rehabilitation Motion Data Enhancement Method Based on Adversarial Samples
CN110118926A (en) PCB based on Electromagnetic Environmental Effect distorts intelligent detecting method
CN110390309B (en) A method for identifying illegal users of finger veins based on residual distribution
Zheng et al. A user behavior-based random distribution scheme for adversarial example generated CAPTCHA
Bolle et al. Performance evaluation in 1: 1 biometric engines

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