CN117034374A - LM-BPNN hardware Trojan detection method and system based on PSO - Google Patents
LM-BPNN hardware Trojan detection method and system based on PSO Download PDFInfo
- Publication number
- CN117034374A CN117034374A CN202311089596.2A CN202311089596A CN117034374A CN 117034374 A CN117034374 A CN 117034374A CN 202311089596 A CN202311089596 A CN 202311089596A CN 117034374 A CN117034374 A CN 117034374A
- Authority
- CN
- China
- Prior art keywords
- particle
- iteration
- neural network
- pso
- bpnn
- 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
- ZXQYGBMAQZUVMI-GCMPRSNUSA-N gamma-cyhalothrin Chemical compound CC1(C)[C@@H](\C=C(/Cl)C(F)(F)F)[C@H]1C(=O)O[C@H](C#N)C1=CC=CC(OC=2C=CC=CC=2)=C1 ZXQYGBMAQZUVMI-GCMPRSNUSA-N 0.000 title claims abstract description 73
- 238000001514 detection method Methods 0.000 title claims abstract description 56
- 238000013528 artificial neural network Methods 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000003062 neural network model Methods 0.000 claims abstract description 13
- 239000002245 particle Substances 0.000 claims description 95
- 239000011159 matrix material Substances 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 13
- 230000008859 change Effects 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013016 damping Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 21
- 239000000523 sample Substances 0.000 description 9
- 230000000694 effects Effects 0.000 description 4
- 238000011478 gradient descent method Methods 0.000 description 4
- 230000007423 decrease Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 241000283086 Equidae Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000007943 implant Substances 0.000 description 1
- 238000002513 implantation Methods 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0499—Feedforward networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention relates to the field of hardware security, in particular to a LM-BPNN hardware Trojan horse detection method and system based on PSO, wherein the method comprises the following steps: establishing a neural network model, and optimizing the neural network model; selecting part of chip circuits without Trojan and part of chip circuits with Trojan as samples, and collecting power consumption information of the samples; training the neural network by taking the power consumption information and a target output value corresponding to the power consumption information; the power consumption information of the chip circuit to be detected is acquired, and the acquired power consumption information is sent to a neural network which has completed training for discrimination and output; judging whether the chip circuit to be detected contains Trojan horse or not according to the output value. The method solves the problems that the convergence speed of the neural network in hardware Trojan detection is low, the whole neural network is too dependent on an initial value, local extremum is easy to fall into, and the hardware Trojan detection efficiency is low, and provides an effective way for finding Trojan in time.
Description
Technical Field
The invention relates to the field of hardware security, in particular to a LM-BPNN hardware Trojan horse detection method and system based on PSO.
Background
In recent years, with the rapid development of new generation information technology, FPGAs are increasingly widely applied in the fields of high-performance computing, 5G, artificial intelligence, internet of things and the like. Cloud acceleration servers based on FPGA are sequentially deployed in the Arian cloud, the Tengming cloud, the Baidu cloud and the like; microsoft uses FPGA to accelerate the must-search; in the field of cryptography, the FPGA completes cryptographic algorithms such as signature verification and the like in a coprocessor mode, so that the information safety is ensured. In the design and manufacturing stage of FPGA chips, FPGA manufacturers typically use third party IP cores in order to shorten the development cycle, and these untrusted third party IPs have the risk of introducing hardware trojans. Meanwhile, manufacturers outsource the manufacture and assembly of chips to other factories in order to reduce the cost, and the possibility that the FPGA equipment is implanted into a hardware Trojan is greatly increased. Secondly, in the application development process of the FPGA, an attacker can realize implantation of hardware Trojan horse in a plurality of stages, such as modifying a register transmission level code, synthesizing a netlist and laying out and wiring the netlist. Finally, during the application of the FPGA, an attacker may even modify the bitstream file to implant the Trojan horse.
Although the standard BP algorithm is widely applied as an effective hardware Trojan detection scheme, practice finds that the BP neural network-based side channel detection scheme has the defects of low convergence speed, long training time and the like, and many expert scholars at home and abroad improve the BP neural network-based side channel detection scheme based on the BP algorithm, and a plurality of methods such as a conjugate gradient method, a Newton method, a momentum method, an LM algorithm and the like are provided, wherein the LM algorithm has the advantages of high convergence speed and the like and is focused by more scholars. Although the LM algorithm is significantly improved in terms of convergence rate, training time, compared to the standard BP algorithm, there are still shortcomings. When the distance between the current optimal solution and the true optimal solution is relatively short, the damping coefficient mu gradually decreases, the convergence speed is very slow, and the algorithm efficiency is reduced. Secondly, the LM algorithm can not ensure convergence to a global minimum point, so that a weight matrix trained by the neural network still depends on the selection of initial weights, and if a local minimum value is encountered in the training process of the neural network, the neural network falls into the local minimum and can not jump out.
Disclosure of Invention
Aiming at the inadequacy of the existing method and the requirement of practical application, the training efficiency of the neural network and the success rate of Trojan horse detection are improved. In one aspect, the invention provides a LM-BPNN hardware Trojan detection method based on PSO, which comprises the following steps: establishing a neural network model, and optimizing the neural network model; selecting part of chip circuits without Trojan and part of chip circuits with Trojan as samples, and collecting power consumption information of the samples; training the neural network by taking the power consumption information and a target output value corresponding to the power consumption information; the power consumption information of the chip circuit to be detected is acquired, and the acquired power consumption information is sent to a neural network which has completed training for discrimination and output; judging whether the chip circuit to be detected contains Trojan horse or not according to the output value. The invention solves the problems that the convergence speed of the neural network in hardware Trojan detection is low, the whole neural network is too dependent on an initial value, local extremum is easy to fall into, and the detection efficiency of the hardware Trojan is low by optimizing the neural network algorithm, and provides an effective way for finding the Trojan in time.
Optionally, optimizing the neural network model includes the steps of: initializing a weight space, setting the size of a particle swarm, and recording the initial position of the particles; iterating the positions of the particles, and finding out individual optimal position variation and global optimal position variation of particle iteration; and updating the weight space according to the individual optimal position variation and the global optimal position variation of the particle iteration. According to the invention, the neural network model is optimized, so that the neural network gets rid of dependence on the initial weight, and local minimum value is jumped out in the training process, thereby improving the success rate of Trojan horse detection.
Optionally, the initializing the weight space, setting the size of the particle swarm, and recording the initial position of the particle includes: randomly initializing a D-dimensional weight search space, wherein m particles representing potential solutions of problems are arranged in the weight search space, and the particles form a population; wherein the position of the ith particle is recorded as(i＝1,2,…,m)；
Wherein t is the current iteration number;coordinates of the ith particle in the first dimension at the t-th iteration; />Coordinates of the ith particle in the second dimension at the t-th iteration; />Coordinates of the ith particle in the D dimension at the t-th iteration; t represents the formula as a matrix.
Alternatively, the two-iteration position change of the particle is recorded as
Wherein t is the current iteration number;the coordinate variation of the ith particle in the first dimension at the t-th iteration; />The coordinate variation of the ith particle in the second dimension at the t-th iteration; />The coordinate variation of the ith particle in the D dimension at the t-th iteration; t represents the formula as a matrix.
Optionally, the position change amount satisfies the following formula:
ΔX＝(J T J+μI) -1 ×J T V
wherein J is a Jacobian matrix; j (J) T A transpose of the jacobian matrix; mu is the damping coefficient, and mu>0 and is a constant; i is an identity matrix; v is the error vector.
Alternatively, the individual optimal position variation of the particle iteration is noted asThe global optimal position change of the particle iteration is recorded as
Wherein,representing individual optimal position variation of the particle iteration in a first dimension; />Representing individual optimal position variation of the particle iteration in a second dimension; />Representing individual optimal position variation of the particle iteration in the D dimension; />Representing a global optimum position variation of the particle iteration in the first dimension; />Representing a global optimum position variation of the particle iteration in the second dimension; />Representing a global optimum position variation of the particle iteration in the D-th dimension; t represents the formula as a matrix.
Optionally, the updating the weight space according to the individual optimal position variation and the global optimal position variation of the particle iteration satisfies the following formula:
wherein,representing the position of the (t+1) th iteration of the (i) th particle; />Representing the position of the ith particle at the t-th iteration; />Representing the position change amount of the ith particle from the t th to the t+1st iteration; c 1 And c 2 Called perturbation factor, c 1 ∈(0,1]，c 2 ∈(0,1]；/>Representing individual optimal position variation of particle iteration; />Representing the global optimum position variation of the particle iteration.
Optionally, the training process of the neural network includes: and inputting the power consumption information extraction feature matrix of the sample into the neural network, and training the neural network. According to the invention, the corresponding feature matrix is extracted from the power consumption information of the sample, and the feature matrix is input into the neural network, so that the effect of training the neural network is achieved.
Optionally, the training process of the neural network further includes: and adjusting the current weight of the neural network through training error feedback, and updating the actual output weight of the neural network. According to the invention, the weight of the neural network is adjusted through training error feedback, so that the actual output of a sample is closer to the expected one, and the success rate of Trojan detection is improved.
In a second aspect, in order to be able to efficiently execute the LM-BPNN hardware Trojan detection method based on a PSO provided by the present invention, the present invention further provides a LM-BPNN hardware Trojan detection system based on a PSO, which includes a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, and the memory is used to store a computer program, where the computer program includes program instructions, and the processor is configured to invoke the program instructions to execute the LM-BPNN hardware Trojan detection method based on a PSO according to the first aspect of the present invention. The LM-BPNN hardware Trojan detection system based on the PSO has the advantages of compact structure and stable performance, and the LM-BPNN hardware Trojan detection method based on the PSO provided by the invention can be stably executed, so that the overall applicability and practical application capability of the invention are improved.
Drawings
FIG. 1 is a flowchart of a LM-BPNN hardware Trojan horse detection method based on PSO provided by the embodiment of the invention;
FIG. 2 is a flowchart of an optimized neural network model according to an embodiment of the present invention;
FIG. 3 is a block diagram of a PSO-based LM-BPNN hardware Trojan detection system provided by an embodiment of the invention.
Detailed Description
Specific embodiments of the invention will be described in detail below, it being noted that the embodiments described herein are for illustration only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that: no such specific details are necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been described in detail in order not to obscure the invention.
Throughout the specification, references to "one embodiment," "an embodiment," "one example," or "an example" mean: a particular feature, structure, or characteristic described in connection with the embodiment or example is included within at least one embodiment of the invention. Thus, the appearances of the phrases "in one embodiment," "in an embodiment," "one example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combination and/or sub-combination in one or more embodiments or examples. Moreover, those of ordinary skill in the art will appreciate that the illustrations provided herein are for illustrative purposes and that the illustrations are not necessarily drawn to scale.
In an alternative embodiment, please refer to fig. 1, fig. 1 is a flowchart of a PSO-based LM-BPNN hardware Trojan detection method according to an embodiment of the present invention. As shown in FIG. 1, the LM-BPNN hardware Trojan horse detection method based on PSO comprises the following steps:
s1, building a neural network model, and optimizing the neural network model. The BP neural network comprises three parts, namely an input layer, an implicit layer and an output layer. The training process of the BP algorithm consists of two processes, namely forward propagation of the signal and reverse propagation of the error. The signal is transmitted from the input layer of the neural network, passes through the hidden layer and is transmitted to the output layer, an output signal and an error signal are generated at the output layer, and if the error signal meets the requirement, the calculation is finished; if the error signal does not meet the given accuracy requirement, the transfer signal is counter-propagated. Counter-propagation of an error signal is the process by which the error signal generated from the output propagates layer by layer along the output layer, hidden layer, and input layer. In the process, the error feedback can adjust the weight of the neural network, and the actual output of the network gradually approaches the expected output in the continuous updating process of the weight.
Referring to fig. 2, in an alternative embodiment, optimizing the neural network model includes the steps of:
s11, initializing a weight space, setting the size of the particle swarm, and recording the initial position of the particle. In this embodiment, initializing the weight space, setting the size of the particle swarm, and recording the initial position of the particle includes: randomly initializing a D-dimensional weight search space, wherein m particles representing potential solutions of problems are arranged in the weight search space, and the particles form a population; wherein the position of the ith particle is recorded as(i＝1,2,…,m)；
Wherein t is the current iteration number;coordinates of the ith particle in the first dimension at the t-th iteration; />Coordinates of the ith particle in the second dimension at the t-th iteration; />Coordinates of the ith particle in the D dimension at the t-th iteration; t represents the formula as a matrix.
S12, iterating the positions of the particles, and finding out individual optimal position variation and global optimal position variation of particle iteration. In the present embodiment, the two-iteration position change amount of the particle is recorded as
Wherein t is the current iteration number;the coordinate variation of the ith particle in the first dimension at the t-th iteration; />The coordinate variation of the ith particle in the second dimension at the t-th iteration; />The coordinate variation of the ith particle in the D dimension at the t-th iteration; t represents the formula as a matrix.
The two-time iteration position change quantity of the particles meets the following formula:
ΔX＝(J T J+μI) -1 ×J T V
wherein J is a Jacobian matrix; j (J) T A transpose of the jacobian matrix; mu is the damping coefficient, and mu>0 and is a constant; i is an identity matrix; v is the error vector.
Recording the individual optimal position change of the particle iteration asThe global optimum position variation of the particle iteration is marked as +.>
Wherein,representing individual optimal position variation of the particle iteration in a first dimension; />Representing individual optimal position variation of the particle iteration in a second dimension; />Representing individual optimal position variation of the particle iteration in the D dimension; />Representing a global optimum position variation of the particle iteration in the first dimension; />Representing a global optimum position variation of the particle iteration in the second dimension; />Representing a global optimum position variation of the particle iteration in the D-th dimension; t represents the formula as a matrix.
S13, updating the weight space according to the individual optimal position variation and the global optimal position variation of the particle iteration.
Specifically, the updating of the weight space according to the individual optimal position variation and the global optimal position variation of the particle iteration satisfies the following formula:
wherein,representing the position of the (t+1) th iteration of the (i) th particle; />Representing the position of the ith particle at the t-th iteration; />Representing the position change amount of the ith particle from the t th to the t+1st iteration; c 1 And c 2 Called perturbation factor, c 1 ∈(0,1]，c 2 ∈(0,1]；/>Representing individual optimal position variation of particle iteration; />Representing the global optimum position variation of the particle iteration.
S2, selecting part of chip circuits without Trojan and part of chip circuits with Trojan as samples, and collecting power consumption information of the samples.
In this embodiment, the power consumption information collection platform is set up to collect the power consumption information of the Trojan-free chip circuit and the Trojan-containing chip circuit which are selected as the samples. The voltage, current, electromagnetic information and the like of the sample chip circuit can be acquired according to actual requirements. The power consumption information acquisition platform can be composed of a high-sensitivity electromagnetic probe, a digital oscilloscope, a signal amplifier, a high-precision fixed platform, a computer and the like. After the chip is loaded and applied, the high-sensitivity electromagnetic probe is controlled by the high-precision fixed platform to measure the power consumption of the chip point by point on the surface of the chip, and detected power consumption information is transmitted back to the computer. Parameters such as the path of movement of the probe, the area scanned, and the sampling rate can be set by software on the computer.
And S3, taking the power consumption information and a target output value corresponding to the power consumption information to train the neural network.
In this embodiment, the training process of the neural network includes: and taking the global optimal solution of the particle swarm as an initial weight of the neural network.
Further, the power consumption information extraction feature matrix of the sample is input into the neural network, and the neural network is trained. In order to reduce the calculated amount of the neural network and improve the training speed of the neural network, the feature matrix can be input into the neural network after the dimension is reduced, and the training effect of the neural network is achieved by establishing a one-to-one correspondence mapping relation between the power consumption information of the sample and the actual condition of the chip circuit.
Further, the current weight of the neural network is adjusted through training error feedback. According to the invention, the weight of the neural network is adjusted through training error feedback, so that the actual output value is more approximate to the expected value, and the success rate of Trojan horse detection is improved.
S4, collecting power consumption information of the chip circuit to be detected, and sending the collected power consumption information to the neural network after training is completed to judge and output.
In this embodiment, the built power consumption information collection platform is used to collect the power consumption information of the chip circuit to be detected. Further, the feature matrix is extracted from the acquired power consumption information, the dimension of the extracted feature matrix can be further reduced in order to reduce the calculated amount of the neural network, the feature matrix after the dimension reduction is input into the neural network after training, and the neural network performs calculation and analysis on the feature matrix to generate an output value.
And S5, judging whether the chip circuit to be detected contains Trojan horse or not according to the output value.
Further, in order to verify the stability of the detection result of the hardware Trojan horse with limited training samples, 10 hardware Trojan horses with different areas are detected by using a gradient descent algorithm, a momentum gradient descent algorithm, a variable learning rate momentum gradient descent algorithm, an LM algorithm and an LM algorithm random initial weight and a threshold value based on PSO respectively, and each type of Trojan horse to be detected is detected 100 times to finally obtain the detection success rate of the Trojan horse circuit, as shown in table 1. As can be seen from table 1, the Trojan detection success rate generally decreases as the Trojan area ratio gradually decreases. The gradient descent method and the momentum gradient descent method have poor Trojan detection effects, and the T2100 detection success rate with the largest area does not reach 100%. The detection success rate of the Trojan horse circuit with the area ratio of more than 0.53% by the variable learning rate gradient descent method and the variable learning rate momentum gradient descent method can reach 100%. The detection success rate of the LM algorithm on the Trojan circuit with the area ratio of more than 0.30% can reach 100%, but the detection success rate of the LM algorithm on the Trojan circuit with the Trojan area ratio of 0.19% is only 86.67%, and the detection success rate of the Trojan detection method on the small hardware Trojan is poor. The detection success rate of the LM algorithm based on PSO is 100%, and the result shows that the detection effect of the LM-BPNN hardware Trojan detection method based on PSO under the condition of limited training samples is obviously better than that of other neural networks. In summary, under the condition of limited training samples, when the distance between the current optimal solution and the true optimal solution is relatively short, the LM algorithm has the problems of slow convergence speed and easy falling into local optimal state and incapability of jumping. According to the invention, the PSO algorithm is used for optimizing the model, so that the convergence can be accelerated, the local optimum can be jumped out, the convergence of the model can be completed with fewer training samples and time, and the detection efficiency of the hardware Trojan horse is effectively improved.
Table 1 success rate of detection for various types of circuits containing Trojan horse
Referring to fig. 3, in an alternative embodiment, to be capable of efficiently executing the PSO-based LM-BPNN hardware Trojan detection method provided by the present invention, the present invention further provides a PSO-based LM-BPNN hardware Trojan detection system, where the system includes a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, where the memory is configured to store a computer program, where the computer program includes program instructions, and the processor is configured to invoke the program instructions to execute the specific steps of the related embodiments of the PSO-based LM-BPNN hardware Trojan detection method provided by the present invention. The LM-BPNN hardware Trojan detection system based on the PSO has complete, objective and stable structure, can efficiently execute the LM-BPNN hardware Trojan detection method based on the PSO, and improves the overall applicability and practical application capability of the invention.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention, and are intended to be included within the scope of the appended claims and description.
Claims (10)
1. The LM-BPNN hardware Trojan horse detection method based on PSO is characterized by comprising the following steps:
establishing a neural network model, and optimizing the neural network model;
selecting part of chip circuits without Trojan and part of chip circuits with Trojan as samples, and collecting power consumption information of the samples;
training the neural network by taking the power consumption information and a target output value corresponding to the power consumption information;
the power consumption information of the chip circuit to be detected is acquired, and the acquired power consumption information is sent to a neural network which has completed training for discrimination and output;
judging whether the chip circuit to be detected contains Trojan horse or not according to the output value.
2. The PSO-based LM-BPNN hardware Trojan detection method as in claim 1, wherein optimizing the neural network model includes the steps of:
initializing a weight space, setting the size of a particle swarm, and recording the initial position of the particles;
iterating the positions of the particles, and finding out individual optimal position variation and global optimal position variation of particle iteration;
and updating the weight space according to the individual optimal position variation and the global optimal position variation of the particle iteration.
3. The PSO-based LM-BPNN hardware Trojan detection method as in claim 2, wherein said initializing a weight space, setting a size of a particle swarm, recording an initial position of the particle includes:
randomly initializing a D-dimensional weight search space, wherein m particles representing potential solutions of problems are arranged in the weight search space, and the particles form a population;
wherein the position of the ith particle is recorded as
Wherein t is the current iteration number;coordinates of the ith particle in the first dimension at the t-th iteration; />For the ith particle at the t-th iterationCoordinates in a second dimension; />Is the coordinates of the ith particle in the D dimension at the t-th iteration.
4. The PSO-based LM-BPNN hardware Trojan detection method as in claim 2, wherein said two iterative position variation of particles is noted as
Wherein t is the current iteration number;the coordinate variation of the ith particle in the first dimension at the t-th iteration;the coordinate variation of the ith particle in the second dimension at the t-th iteration; />The variation of the coordinates of the ith particle in the D dimension at the t-th iteration.
5. The PSO-based LM-BPNN hardware Trojan detection method as in claim 4, wherein said position variation satisfies the following formula:
ΔX＝(J T J+μI) -1 ×J T V
wherein J is a Jacobian matrix; j (J) T A transpose of the jacobian matrix; mu is the damping coefficient, and mu>0 and is a constant; i is an identity matrix; v is the error vector.
6. The PSO-based LM-BPNN hardware Trojan detection method as in claim 2, wherein individual optimal positions of said particle iterationsThe change is recorded asThe global optimum position variation of the particle iteration is marked as +.>
Wherein,representing individual optimal position variation of the particle iteration in a first dimension; />Representing individual optimal position variation of the particle iteration in a second dimension; />Representing individual optimal position variation of the particle iteration in the D dimension;representing a global optimum position variation of the particle iteration in the first dimension; />Representing a global optimum position variation of the particle iteration in the second dimension; />Representing the global optimum position variation of the particle iteration in the D-th dimension.
7. The PSO-based LM-BPNN hardware Trojan detection method as in claim 2, wherein said updating said weight space according to the individual optimum position variation and the global optimum position variation of said particle iteration satisfies the following formula:
wherein,representing the position of the (t+1) th iteration of the (i) th particle; />Representing the position of the ith particle at the t-th iteration;representing the position change amount of the ith particle from the t th to the t+1st iteration; c 1 And c 2 Called perturbation factor, c 1 ∈(0,1]，c 2 ∈(0,1]；/>Representing individual optimal position variation of particle iteration; />Representing the global optimum position variation of the particle iteration.
8. The PSO-based LM-BPNN hardware Trojan detection method as in claim 1, wherein the training process of the neural network includes:
and inputting the power consumption information extraction feature matrix of the sample into the neural network, and training the neural network.
9. The PSO-based LM-BPNN hardware Trojan detection method as in claim 1, wherein the training process of said neural network further comprises:
and adjusting the current weight of the neural network through training error feedback, and updating the actual output weight of the neural network.
10. A PSO-based LM-BPNN hardware Trojan detection system, characterized in that the system comprises a processor, an input device, an output device, and a memory, the processor, the input device, the output device, and the memory being interconnected, wherein the memory is for storing a computer program, the computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the PSO-based LM-BPNN hardware Trojan detection method as in any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311089596.2A CN117034374A (en) | 2023-08-28 | 2023-08-28 | LM-BPNN hardware Trojan detection method and system based on PSO |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311089596.2A CN117034374A (en) | 2023-08-28 | 2023-08-28 | LM-BPNN hardware Trojan detection method and system based on PSO |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117034374A true CN117034374A (en) | 2023-11-10 |
Family
ID=88637244
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311089596.2A Pending CN117034374A (en) | 2023-08-28 | 2023-08-28 | LM-BPNN hardware Trojan detection method and system based on PSO |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117034374A (en) |
Citations (9)
| 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 |
| 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 |
| CN107656839A (en) * | 2017-08-11 | 2018-02-02 | 天津大学 | Research on Integrated Circuit Security is assessed and detection method |
| 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 |
| CN108052840A (en) * | 2017-11-13 | 2018-05-18 | 天津大学 | Hardware Trojan horse detection method based on neutral net |
| CN108446555A (en) * | 2018-02-11 | 2018-08-24 | 复旦大学 | The method that hardware Trojan horse is monitored in real time and is detected |
| KR101962227B1 (en) * | 2018-11-26 | 2019-03-26 | 국방과학연구소 | Method and Apparatus for FPGA Hardware-Trojan Trigger Detection Based on Machine Learning |
| CN109543522A (en) * | 2018-10-18 | 2019-03-29 | 天津大学 | Hardware Trojan horse noise preprocess method based on gaussian kernel function SVM |
| US20220026580A1 (en) * | 2020-07-24 | 2022-01-27 | Amida Technology Solutions, Inc. | Method, system, and apparatus for verification of operation using gps receiver input within integrated circuits and electronic systems using an operation navigation system |
-
2023
- 2023-08-28 CN CN202311089596.2A patent/CN117034374A/en active Pending
Patent Citations (9)
| 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 |
| 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 |
| CN107656839A (en) * | 2017-08-11 | 2018-02-02 | 天津大学 | Research on Integrated Circuit Security is assessed and detection method |
| 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 |
| CN108052840A (en) * | 2017-11-13 | 2018-05-18 | 天津大学 | Hardware Trojan horse detection method based on neutral net |
| CN108446555A (en) * | 2018-02-11 | 2018-08-24 | 复旦大学 | The method that hardware Trojan horse is monitored in real time and is detected |
| CN109543522A (en) * | 2018-10-18 | 2019-03-29 | 天津大学 | Hardware Trojan horse noise preprocess method based on gaussian kernel function SVM |
| KR101962227B1 (en) * | 2018-11-26 | 2019-03-26 | 국방과학연구소 | Method and Apparatus for FPGA Hardware-Trojan Trigger Detection Based on Machine Learning |
| US20220026580A1 (en) * | 2020-07-24 | 2022-01-27 | Amida Technology Solutions, Inc. | Method, system, and apparatus for verification of operation using gps receiver input within integrated circuits and electronic systems using an operation navigation system |
Non-Patent Citations (1)
| Title |
|---|
| 宋晨晨: "基于侧信道分析的硬件木马检测技术", 《中国优秀硕士学位论文全文数据库信息科技辑》, pages 37 - 49 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | A novel chaotic particle swarm optimization based fuzzy clustering algorithm | |
| Li et al. | Toward convolutional neural networks on pulse repetition interval modulation recognition | |
| CN106899440B (en) | Network intrusion detection method and system for cloud computing | |
| US20180357335A1 (en) | Systems for solving general and user preference-based constrained multi-objective optimization problems | |
| KR101889451B1 (en) | Weight benefit evaluator for training data | |
| CN111079780A (en) | Training method of space map convolution network, electronic device and storage medium | |
| CN109684834A (en) | A kind of gate leve hardware Trojan horse recognition method based on XGBoost | |
| CN108508351A (en) | A kind of single-particle direct fault location emulation mode based on double-two fingers number current source | |
| US20220122343A1 (en) | Image feature extraction and network training method, apparatus, and device | |
| CN113298152B (en) | Model training method, device, terminal equipment and computer readable storage medium | |
| CN105895089A (en) | Speech recognition method and device | |
| CN110288075A (en) | A kind of feature selection approach based on improvement shuffled frog leaping algorithm | |
| CN111353391A (en) | Radar interference effect evaluation method and device, electronic equipment and storage medium thereof | |
| CN109121133B (en) | Location privacy protection method and device | |
| CN112231775A (en) | Hardware Trojan horse detection method based on Adaboost algorithm | |
| KR20200117690A (en) | Method and Apparatus for Completing Knowledge Graph Based on Convolutional Learning Using Multi-Hop Neighborhoods | |
| CN117034374A (en) | LM-BPNN hardware Trojan detection method and system based on PSO | |
| CN110441815A (en) | Decline improved simulated annealing Rayleigh waves inversion method based on differential evolution and block coordinate | |
| CN113673349B (en) | Method, system and device for generating Chinese text by image based on feedback mechanism | |
| Wang et al. | FCM algorithm and index CS for the signal sorting of radiant points | |
| CN104516858B (en) | Phase diagram matrix method for nonlinear dynamic behavior analysis | |
| CN112446435A (en) | City data classification method and system | |
| CN112285541B (en) | Fault diagnosis method for current frequency conversion circuit | |
| Huang et al. | Projection pursuit flood disaster classification assessment method based on multi-swarm cooperative particle swarm optimization | |
| CN113011446A (en) | Intelligent target identification method based on multi-source heterogeneous data learning |
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 |