CN105893876A - Chip hardware Trojan horse detection method and system - Google Patents
Chip hardware Trojan horse detection method and system Download PDFInfo
- Publication number
- CN105893876A CN105893876A CN201610188381.XA CN201610188381A CN105893876A CN 105893876 A CN105893876 A CN 105893876A CN 201610188381 A CN201610188381 A CN 201610188381A CN 105893876 A CN105893876 A CN 105893876A
- Authority
- CN
- China
- Prior art keywords
- chip
- individuality
- measured
- individual
- class
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/70—Protecting specific internal or peripheral components, in which the protection of a component leads to protection of the entire computer
- G06F21/71—Protecting specific internal or peripheral components, in which the protection of a component leads to protection of the entire computer to assure secure computing or processing of information
- G06F21/76—Protecting specific internal or peripheral components, in which the protection of a component leads to protection of the entire computer to assure secure computing or processing of information in application-specific integrated circuits [ASIC] or field-programmable devices, e.g. field-programmable gate arrays [FPGA] or programmable logic devices [PLD]
Abstract
The invention relates to a chip hardware Trojan horse detection method and system. The method comprises the steps that parameter bypass testing is conducted on all chips to be detected, and bypass data of the chips to be detected in a working state is collected; clustering analysis is conducted on the chips to be detected by means of a genetic algorithm fused with an K-means algorithm according to the bypass data, and a clustering chip set is obtained; the chips to be detected in the clustering chip set are extracted for backward analysis, and the clustering chip set is classified as a Trojan horse chip or a non-Trojan horse chip according to a result of backward analysis. The clustering analysis capacity is improved by means of the high local searching capability of the K-means algorithm; by means of the high global searching capability of the genetic algorithm and unceasing iterative optimization, an optimal classification result is automatically screened out, artificial intervention is greatly reduced, the automatic identification capability for characters is improved, the hardware Trojan horse detection capability is high, and the efficiency is high.
Description
Technical field
The present invention relates to electronic technology field, particularly relate to a kind of chip hardware Trojan detecting method and system.
Background technology
Hardware Trojan horse refers to certain circuit structure embedded in integrated circuits.Along with sending out of electronic science and technology
Exhibition, the use of chip is more and more extensive, and its safety used the most increasingly comes into one's own.But, chip
In the course of processing not exclusively the most controlled, assailant can embed hardware Trojan horse to chip easily, when
When hardware Trojan horse activates, it is likely to result in the temporary paralysis of Circuits System, chip failure, dysfunction or information
The problems such as leakage.
In order to strengthen the safety of chip, need chip is carried out hardware Trojan horse detection.Traditional detection method
Mainly have: failure analysis, logic function detection and by-passing signal analysis.Failure analysis is mainly by going decoring
Sheet encapsulates, and grinding chip layer is repeatedly scanned with each layer of circuit by precision instrument and equipment, in order to new to chip
Road function is analyzed comprehensively, it may be judged whether there is wooden horse;This method needs chip carries out destructive inspection
Surveying, process is loaded down with trivial details and cost is high, and is difficult to for highly integrated chip.Logic function detection be
Input applies test vector, observes the difference between output signal and the intended output of circuit, thus sentences
Break and whether there is hardware Trojan horse;The method needs to activate hardware Trojan horse, and the longest;By-passing signal analysis master
The physical characteristic information that acquisition chip to be passed through operationally is revealed, utilizes signal processing technology to carry out space change
Change and compression realizes feature extraction, the physical features of fiducial chip and chip to be measured is portrayed and sentences with difference
Not, whether there is hardware Trojan horse according to diversity judgement;The method is vulnerable to the shadow of accuracy of instrument and process noise
Ring, it is difficult to detect the hardware Trojan horse of little area.Therefore, traditional hardware Trojan horse detection method power of test is poor,
Efficiency is low.
Summary of the invention
Based on this, it is necessary to for the problems referred to above, it is provided that the chip hardware that a kind of power of test is strong, efficiency is high
Trojan detecting method and system.
A kind of chip hardware Trojan detecting method, comprises the steps:
All of chip to be measured is carried out parameter bypass test, gather described chip to be measured in working order under
Bypass data;
According to described bypass data, use the genetic algorithm merging K mean algorithm that described chip to be measured is carried out
Cluster analysis, obtains clustering chipset;
The chip described to be measured extracted in described cluster chipset is reversely analyzed, and according to reversely analysis
Described cluster chipset is classified as wooden horse class chip or non-wooden horse class chip by result.
A kind of chip hardware Trojan horse detection system, including:
Parameter bypass test module, for all of chip to be measured carries out parameter bypass test, gathers described
Chip to be measured in working order under bypass data;
Cluster Analysis module, for merging the genetic algorithm of K mean algorithm according to described bypass data, employing
Described chip to be measured is carried out cluster analysis, obtains clustering chipset;
Trojan horse detection module, reversely analyzes for the chip described to be measured extracted in described cluster chipset,
And according to the result reversely analyzed, described cluster chipset is classified as wooden horse class chip or non-wooden horse class chip.
Said chip hardware Trojan horse detection method and system, carry out parameter bypass test to all of chip to be measured,
Gather chip to be measured in working order under bypass data after, according to bypass data, use merge K average calculate
The genetic algorithm of method carries out cluster analysis to chip to be measured, obtains clustering chipset, then extracts cluster chip
The chip to be measured concentrated reversely is analyzed, and cluster chipset is classified as wooden horse class by the result according to reversely analyzing
Chip or non-wooden horse class chip.By utilizing the local search ability that K mean algorithm is stronger to improve cluster point
Analysis ability, utilizes the ability of searching optimum that genetic algorithm is stronger, and by continuous iteration optimization, Automatic sieve is selected
Optimum classification results, greatly reduces artificial intervention, improves the automatic identification ability of feature, hardware Trojan horse
Power of test is strong, efficiency is high.
Accompanying drawing explanation
Fig. 1 is the flow chart of chip hardware Trojan detecting method of the present invention in an embodiment;
Fig. 2 is the three-dimensional result figure of the bypass data of chip to be measured in a specific embodiment;
Fig. 3 is to merge the genetic algorithm of K mean algorithm to be measured according to bypass data, employing in an embodiment
Chip carries out cluster analysis, obtains clustering the particular flow sheet of chipset;
Fig. 4 is the result figure using chip hardware Trojan detecting method of the present invention in a specific embodiment;
Fig. 5 is the module map of chip hardware Trojan horse detection system of the present invention in an embodiment;
Fig. 6 is the unit figure of Cluster Analysis module in an embodiment.
Detailed description of the invention
With reference to Fig. 1, the chip hardware Trojan detecting method in one embodiment of the invention, comprise the steps.
S110: all of chip to be measured is carried out parameter bypass test, gather chip to be measured in working order under
Bypass data.
Parameter bypass test refers to chip to be measured is carried out by-passing signal analysis, gathers chip to be measured and operationally lets out
The bypass data corresponding to physical characteristic information of dew.
In the present embodiment, the bypass data of collection includes global clock average voltage, dynamic power consumption average voltage
Output frequency with ring oscillator.Each bypass data i.e. is the three-dimensional data including three kinds of information.?
Wherein in a specific embodiment, chip to be measured has 160, and Fig. 2 is the global clock of 160 chips to be measured
Average voltage (Clk average), dynamic power consumption average voltage (Power average) and the output of ring oscillator
The three-dimensional result figure of frequency (fmax), it is seen that simple time domain data analysis can not efficiently differentiate open wood
Horse and non-wooden horse chip.
S130: according to bypass data, use the genetic algorithm merging K mean algorithm that chip to be measured is gathered
Alanysis, obtains clustering chipset.
K mean algorithm is a kind of Dynamic Clustering Algorithm in cluster analysis, and basic thought is with certain distance degree
Measure as similarity measure between pattern, the method using iteration to update, multidimensional characteristic vectors is divided into some
Individual set, the similarity measure making characteristic vector between identity set is maximum.
Genetic algorithm (Genetic Algorithms) is a kind of global search optimal solution method that new development is got up.It
Simulation life concern mechanism, including the breeding, copulation and the sudden change that occur in natural selection and genetic evolution
Phenomenon, from any one initial population, by randomly choosing, intersect and mutation operation, produces a group
The new individuality more adapting to environment, uses iteration update method, make Swarm Evolution in search volume increasingly
Good region, tries to achieve the optimal solution of problem.Genetic algorithm need not model for complicated optimization problem and complicated
Computing, as long as utilizing three operators (selection opertor, crossover operator and mutation operator) of genetic algorithm with regard to energy
Obtain optimal solution, there is stronger ability of searching optimum.
Wherein in an embodiment, with reference to Fig. 3, step S130 includes that step S131 is to step S138.
S132: according to default clusters number all of chip to be measured encoded obtain that serial number represents
Body, repeatedly circulation is until individual quantity is equal to presetting number of individuals, and using group of individuals as parent colony.
Wherein, default clusters number refers to the class number classifying sample to be tested.Preset clusters number and
Default number of individuals can be arranged according to practical situation.In the present embodiment, the problem solved as required, arrange
Default clusters number is 2, and sample to be tested can be divided into two classes.In the present embodiment, default number of individuals is 200,
I.e. parent colony includes 200 individualities.
Coding of the present invention refers to sample to be tested is numbered classification.
Wherein in an embodiment, all of chip to be measured is carried out by step S132 according to default clusters number
Coding obtains the step of the individuality that serial number represents and includes: obtain each one class-mark of chip random assortment to be measured
The individuality conspired to create to multidigit class-mark.
Wherein, the species number of class-mark is equal to presetting clusters number, and the corresponding gene position of each class-mark.
In the present embodiment, class-mark includes " 1 " and " 2 ".Being appreciated that in other examples, class-mark is also
Other numerals can be used.
In one specific embodiment, by step S132 to 160 samples to be tested random assortment class-mark " 1 " or
" 2 ", obtain 121121 ... the numeric string of 21, and this numeric string forms body one by one, a length of 160 (etc.
In chip count to be measured), to there being 160 gene position.Obtain, one by one after body, repeating core to be measured
Sheet encodes, until obtaining 200 individualities as a parent colony.
In genetic algorithm, the coded system of chip to be measured has binary coding, symbolic coding and floating-point encoding
Deng.According to actual clustering problem in the present embodiment, symbolization encodes, i.e. arranges class-mark " 1 " and " 2 "
Chip to be measured is encoded, it is simple to follow-up to including representing that the individuality of chip to be measured carries out cluster point with class-mark
Analysis, it is possible to be greatly improved the efficiency of genetic algorithm.
S133: obtain individual fitness and selection opertor according to bypass data.
Fitness is a part extremely crucial in whole genetic algorithm, directly influences the speed of algorithmic statement
With the efficiency finding optimum individual.The individuality that fitness is high retains during evolution, and fitness low
Body is the most gradually eliminated.
Selection opertor, according to the most how selecting which individuality to replicate from parent colony, is genetic to
Of future generation.
Wherein in an embodiment, step S133 includes that step 21 is to step 23.
Step 21: individuality is classified according to class-mark, and the cluster of each class is obtained according to bypass data
Center, obtains individual assessed value according to cluster centre.
In the individuality of 160 figure place word strings " 121121 ... 21 ", the chip to be measured that class-mark " 1 " is represented as
The first kind, the chip to be measured that class-mark " 2 " represents, as Equations of The Second Kind, obtains the according to bypass data the most respectively
One class and the cluster centre of Equations of The Second Kind, and obtain individual assessed value according to cluster centre.
Step 22: individual assessed value is carried out size sequence, and according to sequence sequence number and default factor of influence
Obtain corresponding individual fitness.
Step 23: obtain corresponding individual selection opertor according to individual fitness.
In the present embodiment, step 21 obtains the cluster centre of each class according to bypass data, according to cluster
Center obtain individual assessed value step particularly as follows:
Step 22 particularly as follows:
M_pop (k) .fitness=(1-α)index-1;
Step 23 particularly as follows:
Wherein, Xj kiRepresent the bypass data of individual the i-th apoplexy due to endogenous wind jth chip to be measured of kth corresponding to
Amount, Ni (k) represents the number of the i-th apoplexy due to endogenous wind chip to be measured of kth individuality,Represent that kth is individual
The cluster centre of middle i-th class;CenterNum represents that default clusters number, m_pop (k) .value represent
K individual assessed value;α represents default factor of influence, and index represents the sequence sequence number of assessed value,
M_pop (k) .fitness represents the fitness that kth is individual;M_pop (m) .fitness represents that m-th is individual
Fitness, m_pop (n) .fitness represents the fitness of the n-th individuality, popSize represent preset individuality
Number, cFitness (k) represents the selection opertor that kth is individual.
Wherein, the span presetting factor of influence α is (0,1), and in the present embodiment, α takes 0.6.
In the present embodiment, calculate in the formula of assessed valueRepresent the i-th apoplexy due to endogenous wind that kth is individual
Bypass data corresponding to jth chip to be measured is to the Euclidean distance of such cluster centre.It is different from traditional commenting
Valuation calculates, and in the present embodiment, the calculating of assessed value has used the model of K mean algorithm as genetic algorithm
Clustering Model, the valuation functions of structure genetic algorithm.For each individuality, similar K mean algorithm is used to obtain
Take assessed value, owing to K mean algorithm has stronger local search ability, therefore merge K mean algorithm and obtain
The assessed value taken can characterize the good and bad degree of each individuality, improves cluster analysis ability.Assessed value is the least,
Then the error of presentation class is the least, and it is the highest that individuality is chosen to follow-on probability.
Step 22 is by obtaining fitness after being ranked up individual assessed value, according to assessed value sort method,
Introduce new fitness computational methods, solve local Fast Convergent problem.No matter the size of assessed value is how,
Individual selected probability is only relevant with the sequence number of assessed value sequence, this avoid in generation colony the suitableeest
Or should excessively be not suitable with individual interference, improve the accuracy that optimized choice is individual.
S134: select corresponding individual from parent colony according to individual selection opertor, repeatedly circulation is until choosing
The individual quantity selected is equal to presetting number of individuals, and the individuality of Replica Selection is as transitional population.
By selecting individuality to replicate, in can being remained into by individuality higher for fitness from parent colony
Between colony, it is to avoid gene loss, improve global convergence and computational efficiency.
In the present embodiment, step S134 selects corresponding individuality according to individual selection opertor from parent colony
Selection parameter, particularly as follows: one Selection parameter of stochastic generation, is less than selection opertor from parent colony by step
Corresponding individuality is selected, if the individuality selected has multiple, then selects fitness closest from the individuality selected
The individuality of Selection parameter.That is, by the Selection parameter randomly generated and the comparison of selection opertor, once from father
Replicate for colony selects body one by one.
In one specific embodiment, select one by one after body, one Selection parameter of stochastic generation again, repeats
Individuality is selected to carry out the step replicated from parent colony, until the individual amount of transitional population is equal to default
Body number, i.e. completes once to select to replicate, and the group size obtained after selecting each time to replicate is constant.
S135: carry out the individuality in transitional population intersecting according to default crossover probability and default mutation probability and
Variation, obtains colony of future generation, and obtains fitness and the selection opertor of individual in population of future generation.
Obtain colony of future generation after intersection, variation, i.e. complete an iteration.By individuality is intersected,
Individual primitive character can be preserved, original excellent genes is entailed of future generation individual, and generation comprises more
The new individuality of labyrinth;By the individuality after intersecting is made a variation, the office of crossover operator can be overcome
Portion's search capability, to be quickly found out optimum individual.
Wherein in an embodiment, according to presetting crossover probability and default mutation probability to next in step S135
Carrying out intersecting and making a variation for the individuality in colony, the step obtaining colony of future generation includes step 31 and step 32.
Step 31: generate individual intersection gene position according to default crossover probability, the most not repeatedly from next
For colony selects two individualities, according to intersection gene position, two individualities selected are intersected, follow
Ring is until all individualities in colony of future generation all complete to intersect.
Default crossover probability specifically can be arranged according to practical situation.In the present embodiment, in order to make intersection more abundant
Ground is carried out, and crossover probability is set to 0.6.
In the present embodiment, according to intersection gene position, two individualities selected are carried out cross one another step concrete
Intersect duplication for: the class-mark after intersection gene position individual to two chosen.Intersection operation one
As be divided into: single-point intersection, multiple-spot detection, unanimously intersect.Intersection in the present embodiment uses single-point to intersect,
I.e. make a living conclusion of the business vent with default crossover probability, amount of calculation can be reduced.
Step 32: the individual class-mark after traversal intersection, enters individual each class-mark according to default mutation probability
Row variation, obtains colony of future generation.
Mutation probability can be the numerical value between 0.001 and 0.1.In the present embodiment, mutation probability is 0.05.If
Put less mutation probability, it is to avoid destroy a lot of excellent sample and optimum individual cannot be obtained.
In the present embodiment, the step that individual each class-mark made a variation according to default mutation probability particularly as follows:
Corresponding one Mutation parameter of each class-mark stochastic generation, when Mutation parameter is less than mutation probability, gives birth at random
Become a class-mark as the new class-mark of corresponding gene position.
In one specific embodiment, body is to there being 160 class-marks one by one, corresponding 160 chips to be measured.Hand over
Fork probability is 0.6, and the intersection gene position obtained is the 96th, therefore, the most not repeatedly from group of future generation
After body selects two individualities, the class-mark after choose two individualities the 96th is mutually replicated, is
Complete the intersection that the two is individual.For completing the individuality after intersecting, travel through each class-mark, to each
Position one Mutation parameter of class-mark stochastic generation, Mutation parameter less than mutation probability time, generate class-mark " 1 " or
" 2 " are assigned to this class-mark, then to next one class-mark of class-mark stochastic generation, so circulation until time
Go through all class-marks of this individuality, be the variation of this individuality.
In the present embodiment, after obtaining colony of future generation, obtain the individual fitness in colony of future generation and
The step of selection opertor is identical with step S133, does not repeats at this.
S136: default iterations is added one and obtains new iterations, it is judged that whether new iterations
Equal to presetting maximum iteration time;If it is not, represent and also need to proceed iteration, perform step S137;If
It is to represent and be complete iterative operation, performs step S138.
In the present embodiment, the iterations preset is zero, and maximum iteration time is the number of times needing iteration.
S137: using colony of future generation as new parent colony, and return step S134.
By repeating step S134 to step S136, in order to find optimum individual.
S138: the individuality that selection fitness is maximum from colony of future generation is as optimum individual, according to optimum
Chip to be measured is classified by the coding of body, obtains clustering chipset.
Such as, after selecting optimum individual, according to class-mark " 1 " and " 2 ", corresponding chip to be measured is divided into the
One class and Equations of The Second Kind, obtain one or two cluster chipset.Chip to be measured in same cluster chipset
Bypass data characteristic of correspondence the most similar, belong to wooden horse chip or belong to non-wooden horse chip.
The information of the bypass data that the hardware Trojan horse of small area produces compares the letter of the bypass data of ifq circuit
Cease very little, it is easy to covered by process noise and measurement noise;And wooden horse chip and non-wooden horse core
The feature difference of sheet is the least, identifies that the feature difference of wooden horse chip and non-wooden horse chip is crucial the most efficiently
Problem.The present invention, by step S131 to step S138, uses the genetic algorithm merging K mean algorithm,
Utilize the local search ability that K mean algorithm is stronger to improve cluster analysis ability, it is achieved multidimensional data micro-
Little feature extraction;Utilize the ability of searching optimum that genetic algorithm is stronger, by continuous iteration optimization, Automatic sieve
Select the classification results of optimum, greatly reduce artificial intervention, improve the automatic identification ability of feature.
S150: the chip to be measured extracted in cluster chipset is reversely analyzed, and according to the knot reversely analyzed
Cluster chipset is classified as wooden horse class chip or non-wooden horse class chip by fruit.
Reversely analyze and refer to chip to be measured is carried out by existing technology a kind of method of hardware Trojan horse detection.Pass through
Chip to be measured is reversely analyzed, is appreciated that whether this chip to be measured exists hardware Trojan horse, thus learn
Whether the chip to be measured belonging to same cluster chipset with this chip to be measured is wooden horse chip.
Wherein in an embodiment, the quantity of cluster chipset is 2, and step S150 includes: from any one
Cluster chipset randomly chooses a chip to be measured reversely analyze;Judge according to the result reversely analyzed
Whether the chip to be measured of extraction exists hardware Trojan horse.If so, by the cluster chip at the chip place to be measured of extraction
Collection is classified as wooden horse class chip, and another cluster chipset is classified as non-wooden horse class chip.If it is not, treating extraction
The cluster chipset surveying chip place is classified as non-wooden horse class chip, and another cluster chipset is classified as wooden horse class core
Sheet.
By after chip to be measured is classified, extract one of them chip to be measured and reversely analyze,
Result according to reversely analyzing i.e. can determine whether whether all chips to be measured exist hardware Trojan horse, it is not necessary to all
Chip to be measured detects, simple to operate efficiently.
In one specific embodiment, objective circuit uses the Cordic numeral IP kernel circuit of 18, hardware Trojan horse
Being the enumerator of 1, wooden horse circuit is about 5*10 with the area ratio of objective circuit-4.With reference to Fig. 4, for adopting
The result 160 chips to be measured detected with above-mentioned chip hardware Trojan detecting method, 1~No. 16 chip
For non-wooden horse chip, 17~No. 32 chips are wooden horse chip, and 33~No. 96 chips are non-wooden horse chip, 97~
No. 160 chips are wooden horse chip, and as can be seen from the figure wooden horse chip gathers for " 1 " class, and non-wooden horse chip gathers for " 2 "
Class, it is seen that wooden horse chip and can be polymerized to two classes exactly without wooden horse chip, the most also demonstrating the method can
To realize 10-4Hardware Trojan horse detection resolution.
Said chip hardware Trojan horse detection method, carries out parameter bypass test, gathers all of chip to be measured
Chip to be measured in working order under bypass data after, according to bypass data, use and merge K mean algorithm
Genetic algorithm carries out cluster analysis to chip to be measured, obtains clustering chipset, then extracts in cluster chipset
Chip to be measured reversely analyze, and according to the result reversely analyzed, cluster chipset is classified as wooden horse class core
Sheet or non-wooden horse class chip.By utilizing the local search ability that K mean algorithm is stronger to improve cluster analysis
Ability, utilizes the ability of searching optimum that genetic algorithm is stronger, and by continuous iteration optimization, Automatic sieve is selected
Excellent classification results, greatly reduces artificial intervention, improves the automatic identification ability of feature, and hardware Trojan horse is examined
Survey ability is strong, efficiency is high.
With reference to Fig. 5, a kind of chip hardware Trojan horse detection system in one embodiment of the invention, bypass including parameter
Test module 110, Cluster Analysis module 130 and trojan horse detection module 150.
Parameter bypass test module 110 for carrying out parameter bypass test to all of chip to be measured, and collection is treated
Survey chip in working order under bypass data.
Parameter bypass test refers to chip to be measured is carried out by-passing signal analysis, gathers chip to be measured and operationally lets out
The bypass data corresponding to physical characteristic information of dew.
In the present embodiment, the bypass data of collection includes global clock average voltage, dynamic power consumption average voltage
Output frequency with ring oscillator.Each bypass data i.e. is the three-dimensional data including three kinds of information.?
Wherein in a specific embodiment, chip to be measured has 160, and Fig. 2 is the global clock of 160 chips to be measured
The three-dimensional result figure of the output frequency of average voltage, dynamic power consumption average voltage and ring oscillator, it is seen that letter
Single time domain data analysis can not efficiently differentiate open wooden horse and non-wooden horse chip.
Cluster Analysis module 130 for merging the genetic algorithm pair of K mean algorithm according to bypass data, employing
Chip to be measured carries out cluster analysis, obtains clustering chipset.
K mean algorithm is a kind of Dynamic Clustering Algorithm in cluster analysis, and basic thought is with certain distance degree
Measure as similarity measure between pattern, the method using iteration to update, multidimensional characteristic vectors is divided into some
Individual set, the similarity measure making characteristic vector between identity set is maximum.
Genetic algorithm (Genetic Algorithms) is a kind of global search optimal solution method that new development is got up.It
Simulation life concern mechanism, including the breeding, copulation and the sudden change that occur in natural selection and genetic evolution
Phenomenon, from any one initial population, by randomly choosing, intersect and mutation operation, produces a group
The new individuality more adapting to environment, uses iteration update method, make Swarm Evolution in search volume increasingly
Good region, tries to achieve the optimal solution of problem.Genetic algorithm need not model for complicated optimization problem and complicated
Computing, as long as utilizing three operators (selection opertor, crossover operator and mutation operator) of genetic algorithm with regard to energy
Obtain optimal solution, there is stronger ability of searching optimum.
Wherein in an embodiment, with reference to Fig. 6, Cluster Analysis module 130 includes parent colony signal generating unit
132, individual data items computing unit 133, selection copied cells 134, cross and variation unit 135 and optimization point
Class unit 136.
Parent colony signal generating unit 132 is for encoding all of chip to be measured according to default clusters number
Obtaining the individuality that serial number represents, repeatedly circulation is until individual quantity is equal to presetting number of individuals, and by individuality
Set as parent colony.
Wherein, default clusters number refers to the class number classifying sample to be tested.Preset clusters number and
Default number of individuals can be arranged according to practical situation.In the present embodiment, the problem solved as required, arrange
Default clusters number is 2, and sample to be tested can be divided into two classes.In the present embodiment, default number of individuals is 200,
I.e. parent colony includes 200 individualities.
Coding of the present invention refers to sample to be tested is numbered classification.
Wherein in an embodiment, parent colony signal generating unit 132 is treated all of according to default clusters number
Survey chip to carry out encoding and obtain the individuality that serial number represents, particularly as follows: to each chip random assortment one to be measured
Individual class-mark obtains the individuality that multidigit class-mark conspires to create.
Wherein, the species number of class-mark is equal to clusters number, and the corresponding gene position of each class-mark.This reality
Executing in example, class-mark includes " 1 " and " 2 ".It is appreciated that in other examples, class-mark can also
Use other numerals.
In one specific embodiment, by parent colony signal generating unit 132 to 160 sample to be tested random assortment
Class-mark " 1 " or " 2 ", obtain 121121 ... the numeric string of 21, and this numeric string forms body one by one, length
It is 160 (equal to chip count to be measured), to there being 160 gene position.Obtain one by one after body, then
Repeat chip to be measured is encoded, until obtaining 200 individualities as a parent colony.
In genetic algorithm, the coded system of chip to be measured has binary coding, symbolic coding and floating-point encoding
Deng.The present embodiment is compiled according to actual clustering problem, symbolization by parent colony signal generating unit 132
Code, i.e. arranges class-mark " 1 " and chip to be measured is encoded by " 2 ", it is simple to be follow-up to including using class-mark table
Show that the individuality of chip to be measured carries out cluster analysis, it is possible to be greatly improved the efficiency of genetic algorithm.
Individual data items computing unit 133 is for obtaining individual fitness and selection opertor according to bypass data.
Fitness is a part extremely crucial in whole genetic algorithm, directly influences the speed of algorithmic statement
With the efficiency finding optimum individual.The individuality that fitness is high retains during evolution, and fitness low
Body is the most gradually eliminated.
Selection opertor, according to the most how selecting which individuality to replicate from parent colony, is genetic to
Of future generation.
Wherein in an embodiment, individual data items computing unit 133 specifically for: according to class-mark, individuality is entered
Row classification, and the cluster centre of each class is obtained according to bypass data, obtain individuality according to cluster centre
Assessed value;Individual assessed value is carried out size sequence, and obtains according to sequence sequence number and default factor of influence
Corresponding individual fitness;Corresponding individual selection opertor is obtained according to individual fitness.
Such as, in the individuality of 160 figure place word strings " 121121 ... 21 ", the core to be measured that class-mark " 1 " is represented
Sheet is as the first kind, and the chip to be measured that class-mark " 2 " represents is as Equations of The Second Kind, the most respectively according to bypass data
Obtain the first kind and the cluster centre of Equations of The Second Kind, and obtain individual assessed value according to cluster centre.
In the present embodiment, individual data items computing unit 133 with specific reference to:
M_pop (k) .fitness=(1-α)index-1;
Obtain assessed value, fitness and selection opertor.Wherein, Xj kiRepresent individual the i-th apoplexy due to endogenous wind of kth the
The vector that the bypass data of j chip to be measured is corresponding, Ni (k) represents the i-th apoplexy due to endogenous wind core to be measured that kth is individual
The number of sheet,Represent the cluster centre of i-th class in kth individuality;CenterNum represents to preset and gathers
Class number, m_pop (k) .value represents the assessed value that kth is individual;α represents default factor of influence, index
Representing the sequence sequence number of assessed value, m_pop (k) .fitness represents the fitness that kth is individual;
M_pop (m) .fitness represents the fitness that m-th is individual, and m_pop (n) .fitness represents that n-th is individual
Fitness, popSize represents default number of individuals, and cFitness (k) represents the selection opertor that kth is individual.
Wherein, the span of factor of influence α is (0,1), and in the present embodiment, α takes 0.6.
In the present embodiment, calculate in the formula of assessed valueRepresent the i-th apoplexy due to endogenous wind that kth is individual
Bypass data corresponding to jth chip to be measured is to the Euclidean distance of such cluster centre.It is different from traditional commenting
Valuation calculates, and in the present embodiment, the calculating of assessed value has used the model of K mean algorithm as genetic algorithm
Clustering Model, the valuation functions of structure genetic algorithm.For each individuality, similar K mean algorithm is used to obtain
Take assessed value, owing to K mean algorithm has stronger local search ability, therefore merge K mean algorithm and obtain
The assessed value taken can characterize the good and bad degree of each individuality, improves cluster analysis ability.Assessed value is the least,
Then the error of presentation class is the least, and it is the highest that individuality is chosen to follow-on probability.
By obtaining fitness after individual assessed value is ranked up, according to assessed value sort method, introduce
New fitness computational methods, solve local Fast Convergent problem.No matter the size of assessed value how, individual
Selected probability is only relevant with the sequence number of assessed value sequence, this avoid generation colony excessively adapts to or
Excessively inadaptable individual interference, improves the accuracy that optimized choice is individual.
Select copied cells 134 corresponding individual for selecting from parent colony according to individual selection opertor,
Repeatedly circulation is until the individual quantity selected is equal to presetting number of individuals, and the individuality of Replica Selection is as centre
Colony.
By selecting individuality to replicate, in can being remained into by individuality higher for fitness from parent colony
Between colony, it is to avoid gene loss, improve global convergence and computational efficiency.
In the present embodiment, copied cells 134 is selected to select from parent colony according to individual selection opertor right
Should be individual, particularly as follows: one Selection parameter of stochastic generation, by Selection parameter less than selecting from parent colony
The individuality that operator is corresponding is selected, if the individuality selected has multiple, then selects fitness from the individuality selected
Individuality close to Selection parameter.That is, by the Selection parameter randomly generated and the comparison of selection opertor, once
Body one by one is selected to replicate from parent colony.
In one specific embodiment, copied cells 134 is selected to select one by one after body, stochastic generation one again
Individual Selection parameter, repeats to select individuality to carry out the step replicated, until the individual amount of transitional population is equal to pre-
If number of individuals, i.e. completing once to select to replicate, the group size obtained after selecting each time to replicate is constant.
Cross and variation unit 135 is for according to presetting crossover probability and default mutation probability in transitional population
Individuality carries out intersecting and making a variation, and obtains colony of future generation, and obtain individual in population of future generation fitness and
Selection opertor.
Obtain colony of future generation after intersection, variation, i.e. complete an iteration.By individuality is intersected,
Individual primitive character can be preserved, original excellent genes is entailed of future generation individual, and generation comprises more
The new individuality of labyrinth;By the individuality after intersecting is made a variation, the office of intersection operation can be overcome
Portion's search capability, to be quickly found out optimum individual.
Wherein in an embodiment, cross and variation unit 135 includes intersection subelement (not shown) and variation
Subelement (not shown).
Intersect subelement for generating individual intersection gene position according to default crossover probability, the most not repeatedly
From colony of future generation, select two individualities, according to intersection gene position, two individualities selected mutually are handed over
Fork, circulation is until all individualities in colony of future generation all complete to intersect.
Default crossover probability specifically can be arranged according to practical situation.In the present embodiment, in order to make intersection more abundant
Ground is carried out, and crossover probability is set to 0.6.
In the present embodiment, two individualities selected are intersected by the subelement that intersects according to intersection gene position,
The duplication particularly as follows: the class-mark after the intersection gene position individual to two chosen intersects.Intersect and grasp
It is generally divided into: single-point intersection, multiple-spot detection, unanimously intersect.Intersection in the present embodiment uses single-point to hand over
Fork, i.e. generates intersection position with default crossover probability, can reduce amount of calculation.
Variation subelement is for traveling through the individual class-mark after intersection, according to default mutation probability to individual each
Class-mark makes a variation, and obtains colony of future generation.
Mutation probability can be the numerical value between 0.001 and 0.1.In the present embodiment, mutation probability is 0.05.If
Put less mutation probability, it is to avoid destroy a lot of excellent sample and optimum individual cannot be obtained.
In the present embodiment, individual each class-mark is made a variation by variation subelement according to presetting mutation probability, tool
Body is: corresponding one Mutation parameter of each class-mark stochastic generation, when Mutation parameter is less than mutation probability,
One class-mark of stochastic generation is as the new class-mark of corresponding gene position.
In one specific embodiment, body is to there being 160 class-marks one by one, corresponding 160 chips to be measured.Hand over
Fork probability is 0.6, and the intersection gene position obtained is the 96th, therefore, the most not repeatedly from group of future generation
After body selects two individualities, the class-mark after choose two individualities the 96th is mutually replicated, is
Complete the intersection that the two is individual.For completing the individuality after intersecting, travel through each class-mark, to each
Position one Mutation parameter of class-mark stochastic generation, Mutation parameter less than mutation probability time, generate class-mark " 1 " or
" 2 " are assigned to this class-mark, then to next one class-mark of class-mark stochastic generation, so circulation until time
Go through all class-marks of this individuality, be the variation of this individuality.
In the present embodiment, after obtaining colony of future generation, cross and variation unit 135 obtains in colony of future generation
Individual fitness identical, at this not with individual data items computing unit 133 with the function that selection opertor performs
Repeat.
Optimum Classification unit 136 obtains new iterations for adding one by default iterations, it is judged that new
Iterations whether equal to presetting maximum iteration time;It is not equal to maximum iteration time at new iterations
Time, using colony of future generation as new parent colony, and control to select copied cells 134 according to individual choosing
Selecting operator and select corresponding individual from parent colony, repeatedly circulation is until the individual quantity selected is equal to presetting
Number of individuals, and the individuality of Replica Selection is as transitional population;At new iterations equal to presetting greatest iteration
During number of times, the individuality that selection fitness is maximum from colony of future generation is as optimum individual, according to optimum individual
Coding chip to be measured is classified, obtain cluster chipset.
Such as, when iterations is equal to maximum iteration time, after selecting optimum individual, according to class-mark " 1 "
Corresponding chip to be measured is divided into the first kind and Equations of The Second Kind by " 2 ", obtains one or two cluster chipset.
The bypass data characteristic of correspondence of the chip to be measured in same cluster chipset is the most similar, belongs to wooden horse
Chip or belong to non-wooden horse chip.
The information of the bypass data that the hardware Trojan horse of small area produces compares the letter of the bypass data of ifq circuit
Cease very little, it is easy to covered by process noise and measurement noise;And wooden horse chip and non-wooden horse core
The feature difference of sheet is the least, identifies that the feature difference of wooden horse chip and non-wooden horse chip is crucial the most efficiently
Problem.The present invention passes through Cluster Analysis module 130, uses the genetic algorithm merging K mean algorithm, utilizes K
The local search ability that mean algorithm is stronger improves cluster analysis ability, it is achieved the tiny characteristic of multidimensional data
Extract;Utilizing the ability of searching optimum that genetic algorithm is stronger, by continuous iteration optimization, Automatic sieve is selected
Excellent classification results, greatly reduces artificial intervention, improves the automatic identification ability of feature.
Trojan horse detection module 150 is reversely analyzed for extracting the chip to be measured in cluster chipset, and root
According to the result reversely analyzed, cluster chipset is classified as wooden horse class chip or non-wooden horse class chip.
Reversely analyze and refer to chip to be measured is carried out by existing technology a kind of method of hardware Trojan horse detection.Pass through
Chip to be measured is reversely analyzed, is appreciated that whether this chip to be measured exists hardware Trojan horse, thus learn
Whether the chip to be measured belonging to same cluster chipset with this chip to be measured is wooden horse chip.
Wherein in an embodiment, the quantity of cluster chipset is 2, and trojan horse detection module 150 is specifically used
In: from any one cluster chipset, randomly choose a chip to be measured reversely analyze;According to reversely dividing
The result of analysis judges whether the chip to be measured of extraction exists hardware Trojan horse, when there is hardware Trojan horse, and will extraction
The cluster chipset at chip place to be measured be classified as wooden horse class chip, another cluster chipset is classified as non-wooden horse
Class chip, when there is not hardware Trojan horse, is classified as non-wood by the cluster chipset at the chip place to be measured of extraction
Horse class chip, another cluster chipset is classified as wooden horse class chip.
By after chip to be measured is classified, extract one of them chip to be measured and reversely analyze,
Result according to reversely analyzing i.e. can determine whether whether all chips to be measured exist hardware Trojan horse, it is not necessary to all
Chip to be measured detects, simple to operate efficiently.
In one specific embodiment, objective circuit uses the Cordic numeral IP kernel circuit of 18, hardware Trojan horse
Being the enumerator of 1, wooden horse circuit is about 5*10 with the area ratio of objective circuit-4.With reference to Fig. 4, for adopting
The result 160 chips to be measured detected with above-mentioned chip hardware Trojan detecting method, 1~No. 16 chip
For non-wooden horse chip, 17~No. 32 chips are wooden horse chip, and 33~No. 96 chips are non-wooden horse chip, 97~
No. 160 chips are wooden horse chip, and as can be seen from the figure wooden horse chip gathers for " 1 " class, and non-wooden horse chip gathers for " 2 "
Class, it is seen that wooden horse chip and can be polymerized to two classes exactly without wooden horse chip, the most also demonstrating the method can
To realize 10-4Hardware Trojan horse detection resolution.
Said chip hardware Trojan horse detection system, all of chip to be measured is entered by parameter bypass test module 110
Line parameter bypass test, gather chip to be measured in working order under bypass data after, Cluster Analysis module 130
According to bypass data, use the genetic algorithm merging K mean algorithm that chip to be measured is carried out cluster analysis,
To cluster chipset, the chip to be measured that trojan horse detection module 150 is extracted in cluster chipset is reversely analyzed,
And according to the result reversely analyzed, cluster chipset is classified as wooden horse class chip or non-wooden horse class chip.By profit
Improve cluster analysis ability by the local search ability that K mean algorithm is stronger, utilize genetic algorithm stronger
Ability of searching optimum, by continuous iteration optimization, Automatic sieve selects the classification results of optimum, greatly reduces people
The intervention of work, improves the automatic identification ability of feature, and hardware Trojan horse power of test is strong, efficiency is high.
Each technical characteristic of embodiment described above can combine arbitrarily, for making description succinct, the most right
The all possible combination of each technical characteristic in above-described embodiment is all described, but, if these skills
There is not contradiction in the combination of art feature, is all considered to be the scope that this specification is recorded.
Embodiment described above only have expressed the several embodiments of the present invention, and it describes more concrete and detailed,
But can not therefore be construed as limiting the scope of the patent.It should be pointed out that, for this area
For those of ordinary skill, without departing from the inventive concept of the premise, it is also possible to make some deformation and change
Entering, these broadly fall into protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be with appended power
Profit requires to be as the criterion.
Claims (10)
1. a chip hardware Trojan detecting method, it is characterised in that comprise the steps:
All of chip to be measured is carried out parameter bypass test, gather described chip to be measured in working order under
Bypass data;
According to described bypass data, use the genetic algorithm merging K mean algorithm that described chip to be measured is carried out
Cluster analysis, obtains clustering chipset;
The chip described to be measured extracted in described cluster chipset is reversely analyzed, and according to reversely analysis
Described cluster chipset is classified as wooden horse class chip or non-wooden horse class chip by result.
Chip hardware Trojan detecting method the most according to claim 1, it is characterised in that described basis
Described bypass data, employing merge the genetic algorithm of K mean algorithm and described chip to be measured are carried out cluster analysis,
The step obtaining clustering chipset includes:
According to default clusters number all of described chip to be measured encoded obtain that serial number represents
Body, repeatedly circulation until described individuality quantity equal to preset number of individuals, and using described group of individuals as
Parent colony;
Fitness and the selection opertor of described individuality is obtained according to described bypass data;
Selection opertor according to described individuality selects corresponding individual from described parent colony, repeatedly circulation until
The individual quantity selected is equal to described default number of individuals, and the individuality of Replica Selection is as transitional population;
According to default crossover probability and default mutation probability the individuality in described transitional population intersected and become
Different, obtain colony of future generation, and obtain fitness and the selection opertor of described individual in population of future generation;
Default iterations is added one and obtains new iterations, it is judged that whether described new iterations is etc.
In default maximum iteration time;
If it is not, using described colony of future generation as new parent colony, and return described according to described individuality
Selection opertor selects corresponding individual from described parent colony, and repeatedly circulation is until the individual quantity etc. selected
In described default number of individuals, and the individuality of Replica Selection is as the step of transitional population;
The most then select from described colony of future generation the maximum individuality of fitness as optimum individual, according to
Described chip to be measured is classified by the coding of described optimum individual, obtains described cluster chipset.
Chip hardware Trojan detecting method the most according to claim 2, it is characterised in that described basis
Preset clusters number and all of described chip to be measured is encoded the step bag obtaining the individuality that serial number represents
Include:
Each described one class-mark of chip random assortment to be measured is obtained the individuality that multidigit class-mark conspires to create, wherein,
The species number of described class-mark is equal to described default clusters number, and the corresponding gene position of each described class-mark.
Chip hardware Trojan detecting method the most according to claim 3, it is characterised in that described basis
The step of fitness and selection opertor that described bypass data obtains described individuality includes:
According to described class-mark, described individuality is classified, and obtain each class according to described bypass data
Cluster centre, obtains the assessed value of described individuality according to described cluster centre;
The assessed value of described individuality is carried out size sequence, and the sequence number and default factor of influence according to sequence obtains
Take corresponding individual fitness;
Fitness according to described individuality obtains corresponding individual selection opertor.
Chip hardware Trojan detecting method the most according to claim 4, it is characterised in that described basis
Described bypass data obtains the cluster centre of each class, obtains commenting of described individuality according to described cluster centre
The step of valuation, particularly as follows:
The described assessed value to described individuality carries out size sequence, and according to the sequence number of sequence and presetting affect because of
Son obtain corresponding individual fitness step particularly as follows:
M_pop (k) .fitness=(1-α)index-1;
Fitness according to described individuality obtain corresponding individual selection opertor step particularly as follows:
Wherein, Xj kiRepresent that the described bypass data of the i-th apoplexy due to endogenous wind jth chip to be measured of kth individuality is corresponding
Vector, Ni (k) represents the number of chip to be measured described in the i-th class that kth is individual,Represent the
The cluster centre of i-th class in k individuality;CenterNum represents described default clusters number,
M_pop (k) .value represents the assessed value that kth is individual;α represents described default factor of influence, index table
Showing the sequence sequence number of described assessed value, m_pop (k) .fitness represents the fitness that kth is individual;
M_pop (m) .fitness represents the fitness that m-th is individual, and m_pop (n) .fitness represents that n-th is individual
Fitness, popSize represents described default number of individuals, and cFitness (k) represents that the individual selection of kth is calculated
Son.
Chip hardware Trojan detecting method the most according to claim 2, it is characterised in that described basis
Preset crossover probability and the described individuality in described transitional population intersected and makes a variation by default mutation probability,
The step obtaining colony of future generation includes:
The intersection gene position of described individuality is generated, the most not repeatedly from described according to described default crossover probability
Colony of future generation selects two individualities, according to described intersection gene position, two individualities selected is carried out mutually
Intersecting, circulation is until all individualities in described colony of future generation all complete to intersect;
Traversal intersect after the class-mark of described individuality, according to all kinds of to described individuality of described default mutation probability
Number make a variation, obtain colony of future generation.
Chip hardware Trojan detecting method the most according to claim 1, it is characterised in that described cluster
The quantity of chipset is 2, and the chip described to be measured in described extraction described cluster chipset reversely divides
Analysis, and according to the result reversely analyzed, described cluster chipset is classified as wooden horse class chip or non-wooden horse class chip
Step include:
From cluster chipset any one described, randomly choose a chip to be measured reversely analyze;
Judge according to the result reversely analyzed whether the chip described to be measured of extraction exists hardware Trojan horse;
If so, the cluster chipset at the chip place described to be measured of extraction is classified as wooden horse class chip, another
Cluster chipset is classified as non-wooden horse class chip;
If it is not, the cluster chipset at the chip place described to be measured of extraction is classified as non-wooden horse class chip, another
Individual cluster chipset is classified as wooden horse class chip.
8. a chip hardware Trojan horse detection system, it is characterised in that including:
Parameter bypass test module, for all of chip to be measured carries out parameter bypass test, gathers described
Chip to be measured in working order under bypass data;
Cluster Analysis module, for merging the genetic algorithm of K mean algorithm according to described bypass data, employing
Described chip to be measured is carried out cluster analysis, obtains clustering chipset;
Trojan horse detection module, reversely analyzes for the chip described to be measured extracted in described cluster chipset,
And according to the result reversely analyzed, described cluster chipset is classified as wooden horse class chip or non-wooden horse class chip.
Chip hardware Trojan horse detection system the most according to claim 8, it is characterised in that described cluster
Analysis module includes:
Parent colony signal generating unit, for compiling all of described chip to be measured according to default clusters number
Code obtains the individuality that serial number represents, repeatedly circulation is until the quantity of described individuality is equal to presetting number of individuals, and
Using described group of individuals as parent colony;
Individual data items computing unit, for obtaining fitness and the selection of described individuality according to described bypass data
Operator;
Select copied cells, for selecting from described parent colony according to the described selection opertor of described individuality
Corresponding individual, repeatedly circulation is until the individual quantity selected is equal to described default number of individuals, and Replica Selection
Individuality as transitional population;
Cross and variation unit, for according to presetting crossover probability and default mutation probability in described transitional population
Individuality carry out intersecting and making a variation, obtain colony of future generation, and obtain the suitable of described individual in population of future generation
Response and selection opertor;
Optimum Classification unit, obtains new iterations for adding one by default iterations, it is judged that described
Whether new iterations is equal to presetting maximum iteration time;It is not equal to described pre-at described new iterations
If during maximum iteration time, using described colony of future generation as new parent colony, and it is multiple to control described selection
Unit processed selects corresponding individual from described parent colony according to the described selection opertor of described individuality, repeatedly follows
Ring is until the individual quantity selected is equal to described default number of individuals, and the individuality of Replica Selection is as intermediate group
Body;When described new iterations is equal to described default maximum iteration time, from described colony of future generation
Select the maximum individuality of fitness as optimum individual, according to the coding of described optimum individual to described core to be measured
Sheet is classified, and obtains described cluster chipset.
Chip hardware Trojan horse detection system the most according to claim 8, it is characterised in that described poly-
The quantity of class chipset is 2, described trojan horse detection module specifically for:
From cluster chipset any one described, randomly choose a chip to be measured reversely analyze, and according to
The result reversely analyzed judges whether the chip described to be measured of extraction exists hardware Trojan horse, there is hardware Trojan horse
Time, the cluster chipset at the chip place described to be measured of extraction is classified as wooden horse class chip, another clusters core
Sheet collection is classified as non-wooden horse class chip, when there is not hardware Trojan horse, by the chip place described to be measured of extraction
Cluster chipset is classified as non-wooden horse class chip, and another cluster chipset is classified as wooden horse class chip.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610188381.XA CN105893876A (en) | 2016-03-28 | 2016-03-28 | Chip hardware Trojan horse detection method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610188381.XA CN105893876A (en) | 2016-03-28 | 2016-03-28 | Chip hardware Trojan horse detection method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105893876A true CN105893876A (en) | 2016-08-24 |
Family
ID=57014641
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610188381.XA Pending CN105893876A (en) | 2016-03-28 | 2016-03-28 | Chip hardware Trojan horse detection method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105893876A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845287A (en) * | 2017-01-25 | 2017-06-13 | 天津大学 | Hardware Trojan horse detection method based on multi-parameter correlation |
CN107478978A (en) * | 2017-07-27 | 2017-12-15 | 天津大学 | Hardware Trojan horse optimal inspection vector generation method based on population |
CN107622214A (en) * | 2017-08-14 | 2018-01-23 | 天津大学 | Hardware Trojan horse optimal inspection vector generation method based on ant colony |
CN107783877A (en) * | 2017-09-20 | 2018-03-09 | 天津大学 | The test vector generating method that hardware Trojan horse based on analysis of variance effectively activates |
CN108154051A (en) * | 2017-11-23 | 2018-06-12 | 天津科技大学 | A kind of hardware Trojan horse detection method of discrimination based on support vector machines |
CN109583240A (en) * | 2018-10-23 | 2019-04-05 | 中国科学院计算技术研究所 | A kind of IC testing method and system |
CN109740348A (en) * | 2019-01-29 | 2019-05-10 | 福州大学 | A kind of hardware Trojan horse localization method based on machine learning |
CN110222505A (en) * | 2019-05-30 | 2019-09-10 | 北方工业大学 | Industrial control attack sample expansion method and system based on genetic algorithm |
CN114247661A (en) * | 2020-09-24 | 2022-03-29 | 长鑫存储技术有限公司 | Semiconductor product grading method and grading system |
CN115544947A (en) * | 2022-11-02 | 2022-12-30 | 西安电子科技大学广州研究院 | Large-scale integrated circuit layout optimization method based on genetic algorithm |
Citations (4)
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 |
CN104215894A (en) * | 2014-08-28 | 2014-12-17 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan horse detection method and system |
CN104316861A (en) * | 2014-10-16 | 2015-01-28 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan detection method and system |
CN104330721A (en) * | 2014-10-29 | 2015-02-04 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan horse detection method and integrated circuit hardware Trojan horse detection system |
-
2016
- 2016-03-28 CN CN201610188381.XA patent/CN105893876A/en active Pending
Patent Citations (4)
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 |
CN104215894A (en) * | 2014-08-28 | 2014-12-17 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan horse detection method and system |
CN104316861A (en) * | 2014-10-16 | 2015-01-28 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan detection method and system |
CN104330721A (en) * | 2014-10-29 | 2015-02-04 | 工业和信息化部电子第五研究所 | Integrated circuit hardware Trojan horse detection method and integrated circuit hardware Trojan horse detection system |
Non-Patent Citations (1)
Title |
---|
杨淑莹: "《图像模式识别:VC++技术实现》", 31 December 2005 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845287A (en) * | 2017-01-25 | 2017-06-13 | 天津大学 | Hardware Trojan horse detection method based on multi-parameter correlation |
CN107478978A (en) * | 2017-07-27 | 2017-12-15 | 天津大学 | Hardware Trojan horse optimal inspection vector generation method based on population |
CN107622214B (en) * | 2017-08-14 | 2020-03-27 | 天津大学 | Ant colony-based hardware Trojan horse optimization test vector generation method |
CN107622214A (en) * | 2017-08-14 | 2018-01-23 | 天津大学 | Hardware Trojan horse optimal inspection vector generation method based on ant colony |
CN107783877A (en) * | 2017-09-20 | 2018-03-09 | 天津大学 | The test vector generating method that hardware Trojan horse based on analysis of variance effectively activates |
CN107783877B (en) * | 2017-09-20 | 2023-12-22 | 天津大学 | Test vector generation method for effectively activating hardware Trojan based on variation analysis |
CN108154051A (en) * | 2017-11-23 | 2018-06-12 | 天津科技大学 | A kind of hardware Trojan horse detection method of discrimination based on support vector machines |
CN109583240B (en) * | 2018-10-23 | 2021-06-08 | 中国科学院计算技术研究所 | Integrated circuit testing method and system |
CN109583240A (en) * | 2018-10-23 | 2019-04-05 | 中国科学院计算技术研究所 | A kind of IC testing method and system |
CN109740348A (en) * | 2019-01-29 | 2019-05-10 | 福州大学 | A kind of hardware Trojan horse localization method based on machine learning |
CN109740348B (en) * | 2019-01-29 | 2022-06-14 | 福州大学 | Hardware Trojan horse positioning method based on machine learning |
CN110222505A (en) * | 2019-05-30 | 2019-09-10 | 北方工业大学 | Industrial control attack sample expansion method and system based on genetic algorithm |
CN114247661A (en) * | 2020-09-24 | 2022-03-29 | 长鑫存储技术有限公司 | Semiconductor product grading method and grading system |
CN115544947A (en) * | 2022-11-02 | 2022-12-30 | 西安电子科技大学广州研究院 | Large-scale integrated circuit layout optimization method based on genetic algorithm |
CN115544947B (en) * | 2022-11-02 | 2023-05-02 | 西安电子科技大学广州研究院 | Large-scale integrated circuit layout optimization method based on genetic algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105893876A (en) | Chip hardware Trojan horse detection method and system | |
Ulmas et al. | Segmentation of satellite imagery using u-net models for land cover classification | |
Ghorbani et al. | Comparing different resampling methods in predicting students’ performance using machine learning techniques | |
Sharma et al. | A survey on decision tree algorithms of classification in data mining | |
CN110346831B (en) | Intelligent seismic fluid identification method based on random forest algorithm | |
CN103544392B (en) | Medical science Gas Distinguishing Method based on degree of depth study | |
WO2019179403A1 (en) | Fraud transaction detection method based on sequence width depth learning | |
CN110163258A (en) | A kind of zero sample learning method and system reassigning mechanism based on semantic attribute attention | |
CN109214191A (en) | A method of utilizing deep learning forecasting software security breaches | |
CN106649275A (en) | Relation extraction method based on part-of-speech information and convolutional neural network | |
CN104408153A (en) | Short text hash learning method based on multi-granularity topic models | |
CN106201871A (en) | Based on the Software Defects Predict Methods that cost-sensitive is semi-supervised | |
CN106250442A (en) | The feature selection approach of a kind of network security data and system | |
CN102346829A (en) | Virus detection method based on ensemble classification | |
CN101614787A (en) | Analogical Electronics method for diagnosing faults based on M-ary textural classification device | |
CN106127242A (en) | Year of based on integrated study Extreme Precipitation prognoses system and Forecasting Methodology thereof | |
CN112364352B (en) | Method and system for detecting and recommending interpretable software loopholes | |
CN105389480A (en) | Multiclass unbalanced genomics data iterative integrated feature selection method and system | |
CN103886030B (en) | Cost-sensitive decision-making tree based physical information fusion system data classification method | |
CN110399286A (en) | A kind of automatic generation of test data based on independent pathway | |
CN105447520A (en) | Sample classification method based on weighted PTSVM (projection twin support vector machine) | |
CN103714148A (en) | SAR image search method based on sparse coding classification | |
CN110533116A (en) | Based on the adaptive set of Euclidean distance at unbalanced data classification method | |
CN109885482A (en) | Software Defects Predict Methods based on the study of few sample data | |
CN106570537A (en) | Random forest model selection method based on confusion matrix |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160824 |
|
RJ01 | Rejection of invention patent application after publication |