CN108710912B - Time sequence logic approximate model detection method and system based on two-classification machine learning - Google Patents
Time sequence logic approximate model detection method and system based on two-classification machine learning Download PDFInfo
- Publication number
- CN108710912B CN108710912B CN201810487191.7A CN201810487191A CN108710912B CN 108710912 B CN108710912 B CN 108710912B CN 201810487191 A CN201810487191 A CN 201810487191A CN 108710912 B CN108710912 B CN 108710912B
- Authority
- CN
- China
- Prior art keywords
- model
- machine learning
- detection
- model detection
- training
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of artificial intelligence and model detection, and particularly relates to a sequential logic approximate model detection method and system based on two-classification machine learning, wherein the method comprises the following steps: a plurality of system models and time sequence logic formulas are given, model detection is carried out on the system models and the time sequence logic formulas by utilizing a model detection algorithm, and a plurality of data records consisting of three fields of the system models, the time sequence logic formulas, model detection results and the like are obtained; and then, training the record set data by using a two-classification machine learning algorithm, and predicting and outputting a model detection result of a given system model and a time sequence logic formula by using a trained machine learning model. The approximate model detection method disclosed by the invention has the advantages that on the premise of basically keeping the prediction accuracy, the calculation efficiency is improved in a magnitude order, the problem of state space explosion commonly existing in the existing model detection technology is solved, and the method has a wide application prospect in the fields of chip design, network protocols, security protocols, malicious program detection and the like.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence and model detection, and particularly relates to a time sequence logic approximate model detection method and system based on two-classification machine learning.
Background
Model detection, proposed by professor Clarke, et al, is a generic technique for automatically verifying whether a system satisfies given properties. The method is widely applied to multiple fields such as CPU design, security protocol verification, malicious code detection … … and the like, and the developed products are used by information technology leaders such as INTEL and IBM. The general principle of model detection is: in model inspection systems, finite state automata (or Kripke structure) is used to model the system, and sequential logic equations are used to describe the properties that the system needs to satisfy (therefore, model inspection is also often referred to as sequential logic model inspection). If the finite state machine meets the corresponding logic formula, the finite state machine meets the property; otherwise, it is not satisfied. Linear sequential logic and computational tree logic are two major types of sequential logic currently used in the computer industry to describe the linear sequential nature and branch sequential nature of a system. State space explosion is always one of the major technical bottlenecks for model detection. To alleviate this problem, a number of optimization techniques are proposed, including: symbolization, partial order, equivalence, combination, symmetry, and the like. These optimization techniques achieve certain results. In one case of automatically verifying the circuit properties of the chip, even up to 10120 states can be symbolically model-detected. However, the above optimization techniques do not seek to solve the problem by non-traversing the state space, and thus fail to fundamentally avoid state explosion.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a time sequence logic approximate model detection method and system based on two-classification machine learning, which predict the detection result of the model by using a machine learning algorithm and effectively avoid the situations of state explosion and the like in the detection of the model.
According to the design scheme provided by the invention, a sequential logic approximate model detection method based on two-classification machine learning comprises the steps of firstly, obtaining model detection results of a plurality of sequential logic formulas and a system model by using a model detection algorithm; then, performing two-classification machine learning training on data such as a sequential logic formula, a system model, a model detection result and the like to obtain a machine learning model; and finally, predicting the model detection result by using a time sequence logic formula of the machine learning model to the waiting model detection and the system model, wherein the prediction result is the approximate model detection result.
Further, the time sequence logic approximate model detection method based on the two-classification machine learning comprises the following steps:
the method comprises the following steps of (1) obtaining n system models and m sequential logic formulas;
respectively carrying out model detection on n system models and m sequential logic formulas through a model detection algorithm to obtain n x m model detection records, wherein each model detection record at least comprises three fields of a system model, a logic formula and a detection result;
dividing n x m model detection records into training samples and testing samples, and performing two-classification machine learning training on the training samples to obtain a machine learning model; predicting the test sample through a machine learning model to obtain prediction data of a detection result of each model;
and (4) comparing each piece of predicted data with the actual data, if the comparison requirements are met, judging that the obtained machine learning model is the machine learning model for implementing approximation for model detection, otherwise, adjusting parameters of the machine learning model, and repeating the step of performing two-classification machine learning training on the training samples to obtain a new machine learning model until the comparison requirements are met, and ending.
The system model is a Kripke structure, or an automaton, or a petri net model, or a process algebra model.
In the above, the sequential logic formula is a linear sequential logic formula, or a computational tree (sequential) logic formula.
As described above, the model detection algorithm is used to automatically detect whether a given system model satisfies a given sequential logic formula.
As described above, the system model is constructed by numeric characters as the first field of the model test record, a character string representing the sequential logic formula as the second field of the model test record, and a true or false character of the test result as the third field of the model test record.
In the above two-classification machine learning training, the learning training is performed by using a decision tree machine learning algorithm or a random forest machine learning algorithm or a support vector machine learning algorithm or a neural network machine learning algorithm or a bayesian classification machine learning algorithm or a K-nearest neighbor machine learning algorithm or a K-means machine learning algorithm or a logistic regression machine learning algorithm.
As described above, comparing each piece of predicted data with the actual data includes: obtaining the average prediction accuracy and the average prediction time of the test sample according to the prediction data; if the average prediction accuracy is higher than the set threshold and the average prediction time is lower than the preset threshold, the comparison requirement is determined to be met.
A sequential logic approximation model detection system based on two-class machine learning, comprising: an acquisition module, a detection module, a training module, a test module and an output module, wherein,
the acquisition module is used for acquiring n system models and m sequential logic formulas;
the detection module is used for respectively carrying out model detection on the n system models and the m sequential logic formulas through a model detection algorithm, acquiring n model detection records, and dividing the n model detection records into training samples and testing samples, wherein each model detection record at least comprises three fields of a system model, a logic formula and a detection result;
the training module is used for performing two-classification machine learning training aiming at the training samples to obtain a machine learning model;
the testing module is used for predicting the testing sample through the machine learning model and obtaining the prediction data of the detection result of each model;
and the output module is used for comparing each piece of predicted data with the actual data, outputting the obtained machine learning model if the comparison requirements are met, otherwise, performing two-classification machine learning training steps on the training samples again to obtain a new machine learning model until the comparison requirements are met, and outputting the new machine learning model.
In the system, the output module comprises a comparison data acquisition unit, a data judgment unit and an output unit; wherein the content of the first and second substances,
the comparison data acquisition unit is used for acquiring the average prediction accuracy and the average prediction time of the test sample according to the prediction data;
and the data judgment unit is used for comparing the average prediction accuracy and the average prediction time with a preset threshold, transmitting the machine learning model acquired by the training module to the output unit for output if the average prediction accuracy is higher than the preset threshold and the average prediction time is lower than the preset threshold, and otherwise, triggering the training module to perform two-classification machine learning training and acquiring the machine learning model again for testing.
The invention has the beneficial effects that:
the invention introduces a machine learning method to the technical field of model detection, and by defining the problem of model detection as a machine learning two-classification problem and using a two-classification machine learning algorithm to predict the model detection result, the invention realizes the function of carrying out approximate simulation on the model detection technology by using machine learning; the model detection approximation technology is not implemented in a state space searching mode, so that combined state explosion on a large state space is fundamentally avoided in principle, and the increase of the state space caused by the increase of the problem scale is avoided, so that the improvement effect of the method on the operation efficiency is more obvious when the problem scale is large; the model detection is approximately simulated, and the method has wide application prospects in various application fields of the model detection technology, such as chip design, network protocol design, security protocol design, malicious program detection and the like.
Description of the drawings:
FIG. 1 is a flow chart of a method in an embodiment;
FIG. 2 is a block diagram of the system in the example;
FIG. 3 is a block diagram of an output module in an embodiment;
FIG. 4 is a schematic diagram of the detection of an approximate model in an embodiment;
FIG. 5 is a structural diagram of Kripke in the embodiment;
FIG. 6 is a result of operation of the detection of the Kripke structure and the LTL formula NuSMV model in the embodiment;
FIG. 7 is a result of the operation of the Kripke structure and another LTL formula NuSMV model test in the embodiment;
fig. 8 shows the operation result of the machine learning part in the embodiment.
The specific implementation mode is as follows:
in order to make the objects, technical solutions and advantages of the present invention clearer and more obvious, the present invention is further described in detail below with reference to the accompanying drawings and technical solutions.
For the situations of bottleneck state space explosion and the like in the existing model detection technology, in a first embodiment of the present invention, referring to fig. 1, a sequential logic approximation model detection method based on two-class machine learning is provided, including:
s01) acquiring n system models and m sequential logic formulas;
s02) respectively carrying out model detection on the n system models and the m sequential logic formulas through a model detection algorithm to obtain n x m model detection records, wherein each model detection record at least comprises three fields of a system model, a logic formula and a detection result;
s03) dividing n x m model detection records into training samples and testing samples, and performing two-classification machine learning training on the training samples to obtain a machine learning model; predicting the test sample through a machine learning model to obtain prediction data of a detection result of each model;
s04) comparing each piece of prediction data with the actual data, if the comparison requirements are met, judging that the obtained machine learning model is the machine learning model for carrying out approximate model detection, otherwise, adjusting the parameters of the machine learning model, and carrying out the step of two-classification machine learning training on the training samples again to obtain a new machine learning model until the comparison requirements are met, and ending.
By introducing a machine learning method into the technical field of model detection, the problem of model detection is reduced to the problem of machine learning two-classification, and a model detection result is predicted by using a two-classification machine learning algorithm, so that the effect of performing approximate simulation on a model detection technology by using machine learning is realized; the model detection approximation technique is not implemented by searching the state space, thus fundamentally avoiding the explosion of combined states over large state spaces. The model detection algorithm builds a model for the system through an automaton or Kripke structure or petri net or process algebra, describes the properties required to be met by the system through a time sequence logic formula, and judges whether the properties are met on the automaton or Kripke structure or petri net or process algebra.
The system model can be Kripke structure, automaton, petri net, process algebra or other various formalized models that can be used for model detection. The Kripke structure and the automaton are common models. The sequential logic may be linear sequential logic, computational tree (sequential) logic, or other various sequential logics that can be used for model detection, both of which are common sequential logic. The model detection algorithm is any algorithm capable of carrying out model detection on a system model and a time sequence logic formula under the condition of the given system model and the time sequence logic formula. Let the given system model be K and the given sequential logic formula be f, then the model detection algorithm can automatically determine if K satisfies f. In the two-classification machine learning training, the machine learning algorithm training of two classifications can be carried out on the data set through decision trees, random forests, support vector machines, neural networks, Bayesian classification or any other machine learning algorithm. In the above, the system model is constructed by numeric characters as the first field of the model test record, a character string representing the sequential logic formula as the second field of the model test record, and a true or false character of the test result as the third field of the model test record.
Comparing each piece of predicted data with the actual data, which can be obtained by: obtaining the average prediction accuracy and the average prediction time of the test sample according to the prediction data; if the average prediction accuracy is higher than the set threshold and the average prediction time is lower than the preset threshold, the comparison requirement is determined to be met.
Based on the above approximate model detection method, an embodiment of the present invention further provides a time series logic approximate model detection system based on two-class machine learning, as shown in fig. 2, including: an acquisition module 001, a detection module 002, a training module 003, a test module 004, and an output module 005, wherein,
the acquisition module 001 is used for acquiring n system models and m sequential logic formulas;
the detection module 002 is configured to perform model detection on the n system models and the m sequential logic formulas respectively through a model detection algorithm, obtain n × m model detection records, and divide the n × m model detection records into a training sample and a test sample, where each model detection record at least includes three fields, namely a system model, a logic formula, and a detection result;
the training module 003 is used for performing two-class machine learning training on the training samples to obtain a machine learning model;
the test module 004 is used for predicting the test samples through the machine learning model to obtain the prediction data of the detection result of each model;
and the output module 005 is used for comparing each piece of predicted data with the actual data, outputting the obtained machine learning model if the comparison requirements are met, and otherwise, performing two-classification machine learning training steps on the training samples again to obtain a new machine learning model until the comparison requirements are met and outputting the new machine learning model.
In the system, for the output module, as shown in fig. 3, the system includes a comparison data obtaining unit 5001, a data determining unit 5002 and an output unit 5003; wherein the content of the first and second substances,
a comparison data obtaining unit 5001 configured to obtain an average prediction accuracy and an average prediction time of the test sample according to the prediction data;
and the data determination unit 5002 is configured to compare the average prediction accuracy and the average prediction time with a preset threshold, transmit the machine learning model acquired by the training module to the output unit for output if the average prediction accuracy is higher than the preset threshold and the average prediction time is lower than the preset threshold, and otherwise, trigger the training module to perform the two-class machine learning training and acquire the machine learning model again for testing.
To verify the effectiveness of the technical solution of the present invention, model detection is performed on a linear time series logic formula (LTL) below to determine whether a Kripke structure K satisfies an LTL formula f, and the flow is shown in the upper half of fig. 4. A machine learning algorithm, a lifting tree (BT) algorithm, performs approximation on LTL model detection, the process is shown in the lower half of FIG. 4, the BT algorithm is used for predicting the LTL model detection result, and the specific steps are as follows:
step 1.1, randomly generating 20 Kripke structures K and 20 LTL formulas f, and combining to obtain 400 binary groups (K, f);
step 1.2, using an LTL model detection algorithm to respectively detect each pair of K and f in 400 binary groups (K, f) on a model detection tool NuSMV, and recording a model detection result r;
1.3.1, expressing K by using a digital character string, wherein in each state of K, a satisfied atom proposition is expressed by 1, an unsatisfied atom proposition is expressed by 0, each state transition of K is uniquely determined by a starting state sequence number and an ending state sequence number, and the character string constructed in this way is used as a first field of record for expressing K;
step 1.3.2 a sequential logic formula is a character string, which is used as the second field of the record for representing f;
step 1.3.3 model test results r have two possibilities: "yes (true)", and "no (false)", the former is denoted by 1, the latter is denoted by 0, as the third field of the record, and is used for representing r, and the invention is to classify the data according to the value of r.
FIG. 5 shows an example of a Kripke structure K, which has 5 states and 8 state transitions, all three atom propositions p, q, r in the state s0 are not satisfied, only the atom proposition q in the state s1 satisfies … …, the state s0 can transition to the state s1, the state s1 can transition to the state s0 or the state s2 … …, and the K can be represented by a character string 0000100100101110110122124303243, wherein the first 15 bits describe whether 3 atom propositions in 5 states are satisfied, and the last 16 bits describe the starting state sequence number and the ending state sequence number of 8 state transitions; for K ═ 0000100100101110110122124303243 ", f ═ |! X ((! F ((! p & q | r) U (p | q | r))) U (F (p & q & | r))) ", the NuSMV platform model test run result is shown in FIG. 6, where it can be seen that the model test result is" true ", i.e., K satisfies F, so the three fields corresponding to the record are" 0000100100101110110122124303243 "," |)! X ((! F ((! p & q | r) U (p | q | r))) U (F (p & q & | r))) "," 1 ", and it can be seen that the NuSMV model detects that the run time is 0.018 seconds; for K ═ 0000100100101110110122124303243 ", f ═ X! ((F (G! p | q & r))) U ((p & q | r) U (| -p | q & r))) ", the result of the platform model test run of the NuSMV is shown in FIG. 7, where it can be seen that the result of the model test is" false ", i.e., K does not satisfy F, so that the three fields corresponding to the record are" 0000100100101110110122124303243 "," X |)! ((F (G! p | q & r))) U ((p & q | r) U (| -p | q & r))) "," 0 ", and it can be seen that the NuSMV model detects that the run time is 0.017 seconds.
Step 1.3.4 for each of the 400 groups K, f, r, step 1.3.1, step 1.3.2, step 1.3.3 were repeated to obtain 400 records, which were used as the data set a for the machine learning algorithm of the subsequent step, and the average running time of the NuSMV for the model test of the 400 groups was 0.014 seconds.
Step 2.1, performing parameter adjustment training on the A by using a BT algorithm on a machine learning tool Graph Lab, setting parameters as shown in Table 1, automatically dividing a training set and a testing set by the Graph Lab according to the parameters, and operating to obtain a machine learning model M;
table 1 parameters of BT algorithm in example
Step 2.2, automatically using the trained M to test the test set in the A by the Graph Lab, and calculating the average prediction accuracy and the average prediction time, wherein the result is shown in FIG. 8;
step 2.3 as shown in fig. 8, the average prediction accuracy exceeds 95.5%, the average prediction time of a single record reaches 0.00035 seconds, and compared with the average running time (0.014 seconds) of the NuSMV in 400 groups, the efficiency of the present invention in this embodiment is improved by 40 times for 0.014/0.00035; the model M is a found machine learning model that can predict the detection result of the LTL model.
Through the test results, the invention can be verified to improve the calculation efficiency by orders of magnitude on the premise of basically keeping the prediction accuracy on the basis of effective approximate model detection, avoids the problem of state space explosion commonly existing in the existing model detection technology, and has wide application prospect in the fields of chip design, network protocol, security protocol, malicious program detection and the like.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The modules and method steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and the components and steps of the examples have been described in a functional generic sense in the foregoing description for the purpose of clearly illustrating the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
Those skilled in the art will appreciate that all or part of the steps of the above methods may be implemented by instructing the relevant hardware through a program, which may be stored in a computer-readable storage medium, such as: read-only memory, magnetic or optical disk, and the like. Alternatively, all or part of the steps of the foregoing embodiments may also be implemented by using one or more integrated circuits, and accordingly, each module/unit in the foregoing embodiments may be implemented in the form of hardware, and may also be implemented in the form of a software functional module. The present invention is not limited to any specific form of combination of hardware and software.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A time sequence logic approximate model detection method based on two-classification machine learning is characterized in that: firstly, obtaining a model detection result about a plurality of sequential logic formulas and a system model by using a model detection algorithm; then, performing two-classification machine learning training on the time sequence logic formula, the system model and the model detection result data to obtain a machine learning model; finally, a time sequence logic formula of the machine learning model for waiting model detection and a system model are utilized to carry out model detection result prediction, and the prediction result is an approximate model detection result; the model detection problem is reduced to a machine learning two-classification problem, and a two-classification machine learning algorithm is used for predicting a model detection result so as to realize approximate simulation of a model detection technology by machine learning;
expressing a system model K by using a digital character string, wherein in each state of the K, satisfied atom propositions are expressed by 1, unsatisfied atom propositions are expressed by 0, and each state transition in the K is uniquely determined by a starting state serial number and an ending state serial number, so that the constructed character string is used as a first field of a record and is used for expressing the K; a sequential logic formula is a character string as a second field of the record, and is used for representing f; there are two possibilities for the model detection result r: "yes (true)", and "no (false)", the former is denoted by 1, the latter is denoted by 0, and is used as the third field of the record for representing r, and the data is classified two according to the value of r.
2. The method for detecting the sequential logic approximation model based on the binary machine learning as claimed in claim 1, characterized by comprising the following steps:
the method comprises the following steps of (1) obtaining n system models and m sequential logic formulas;
respectively carrying out model detection on n system models and m sequential logic formulas through a model detection algorithm to obtain n x m model detection records, wherein each model detection record at least comprises three fields of a system model, a logic formula and a detection result;
dividing n x m model detection records into training samples and testing samples, and performing two-classification machine learning training on the training samples to obtain a machine learning model; predicting the test sample through a machine learning model to obtain prediction data of a detection result of each model;
and (4) comparing each piece of predicted data with the actual data, if the comparison requirements are met, judging the obtained machine learning model to be the machine learning model which implements approximation for model detection, otherwise, performing the step of two-classification machine learning training on the training sample again to obtain a new machine learning model until the comparison requirements are met, and ending.
3. The method of claim 1, wherein the system model is a Kripke structure, or an automaton, or a petri net model, or a process algebra model.
4. The method for detecting sequential logic approximation model based on binary machine learning of claim 1, wherein the sequential logic formula is a linear sequential logic formula or a computational tree sequential logic formula.
5. The method of claim 1, wherein a model detection algorithm is used to automatically detect whether a given system model satisfies a given sequential logic formula.
6. The method of claim 1, wherein the system model is constructed by numeric characters as a first field of the model test record, a character string representing the sequential logic formula as a second field of the model test record, and true and false characters of the test result as a third field of the model test record.
7. The method for detecting sequential logic approximation model based on binary machine learning according to claim 1, wherein the binary machine learning training is performed by using a decision tree machine learning algorithm or a random forest machine learning algorithm or a support vector machine learning algorithm or a neural network machine learning algorithm or a Bayesian classification machine learning algorithm or a K-nearest neighbor machine learning algorithm or a K-means machine learning algorithm or a logistic regression machine learning algorithm.
8. The method of claim 1, wherein comparing each predicted datum to an actual datum comprises: obtaining the average prediction accuracy and the average prediction time of the test sample according to the prediction data; if the average prediction accuracy is higher than the set threshold and the average prediction time is lower than the preset threshold, the comparison requirement is determined to be met.
9. A sequential logic approximation model detection system based on two-classification machine learning is characterized by comprising: an acquisition module, a detection module, a training module, a test module and an output module, wherein,
the acquisition module is used for acquiring n system models K and m sequential logic formulas;
the detection module is used for respectively carrying out model detection on the n system models and the m sequential logic formulas through a sequential logic model detection algorithm, acquiring n x m model detection records, and dividing the n x m model detection records into training samples and testing samples, wherein each model detection record at least comprises three fields of a system model, a logic formula and a detection result;
the training module is used for performing two-classification machine learning training aiming at the training samples to obtain a machine learning model;
the testing module is used for predicting the testing sample through the machine learning model and obtaining the prediction data of the detection result of each model; the output module is used for comparing each piece of predicted data with the actual data, outputting the obtained machine learning model if the comparison requirements are met, otherwise, adjusting the training parameters of the machine learning model, and performing two-classification machine learning training steps on the training samples again to obtain a new machine learning model until the comparison requirements are met and outputting the new machine learning model;
expressing K by using a numeric character string, wherein in each state in the K, a satisfied atom proposition is expressed by 1, an unsatisfied atom proposition is expressed by 0, and each state transition in the K is uniquely determined by a starting state serial number and an ending state serial number, so that the constructed character string is used as a first field of a record and is used for expressing the K; a sequential logic formula is a character string as a second field of the record, and is used for representing f; there are two possibilities for the model detection result r: "yes (true)", and "no (false)", the former is denoted by 1, the latter is denoted by 0, and is used as the third field of the record for representing r, and the data is classified two according to the value of r.
10. The system according to claim 9, wherein the output module comprises a comparison data obtaining unit, a data determining unit and an output unit; wherein the content of the first and second substances,
the comparison data acquisition unit is used for acquiring the average prediction accuracy and the average prediction time of the test sample according to the prediction data;
and the data judgment unit is used for comparing the average prediction accuracy and the average prediction time with a preset threshold, transmitting the machine learning model acquired by the training module to the output unit for output if the average prediction accuracy is higher than the preset threshold and the average prediction time is lower than the preset threshold, and otherwise, triggering the training module to perform two-classification machine learning training and acquiring the machine learning model again for testing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810487191.7A CN108710912B (en) | 2018-05-21 | 2018-05-21 | Time sequence logic approximate model detection method and system based on two-classification machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810487191.7A CN108710912B (en) | 2018-05-21 | 2018-05-21 | Time sequence logic approximate model detection method and system based on two-classification machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108710912A CN108710912A (en) | 2018-10-26 |
| CN108710912B true CN108710912B (en) | 2021-04-02 |
Family
ID=63869160
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810487191.7A Active CN108710912B (en) | 2018-05-21 | 2018-05-21 | Time sequence logic approximate model detection method and system based on two-classification machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108710912B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070121B (en) * | 2019-04-15 | 2021-08-10 | 西北工业大学 | Rapid approximate K nearest neighbor method based on tree strategy and balanced K mean clustering |
| CN113268890B (en) * | 2021-06-25 | 2023-06-23 | 华北电力大学（保定） | Method for mapping behavior tree model to Kripke structure |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104657610A (en) * | 2015-02-13 | 2015-05-27 | 南京邮电大学 | Temporal logic robustness assessment method for information physical fusion system |
| CN105677307A (en) * | 2014-11-19 | 2016-06-15 | 上海烟草集团有限责任公司 | Big data processing method and system of mobile terminal |
| CN106532940A (en) * | 2016-11-10 | 2017-03-22 | 中国电力科学研究院 | Alarm signal action and time sequence logic relationship analysis method of intelligent substation |
| CN106682118A (en) * | 2016-12-08 | 2017-05-17 | 华中科技大学 | Social network site false fan detection method achieved on basis of network crawler by means of machine learning |
| CN108052528A (en) * | 2017-11-09 | 2018-05-18 | 华中科技大学 | A kind of storage device sequential classification method for early warning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10896096B2 (en) * | 2016-04-21 | 2021-01-19 | International Business Machines Corporation | Temporal logical transactions |
| US10958547B2 (en) * | 2016-09-09 | 2021-03-23 | Hewlett Packard Enterprise Development Lp | Verify a network function by inquiring a model using a query language |
-
2018
- 2018-05-21 CN CN201810487191.7A patent/CN108710912B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105677307A (en) * | 2014-11-19 | 2016-06-15 | 上海烟草集团有限责任公司 | Big data processing method and system of mobile terminal |
| CN104657610A (en) * | 2015-02-13 | 2015-05-27 | 南京邮电大学 | Temporal logic robustness assessment method for information physical fusion system |
| CN106532940A (en) * | 2016-11-10 | 2017-03-22 | 中国电力科学研究院 | Alarm signal action and time sequence logic relationship analysis method of intelligent substation |
| CN106682118A (en) * | 2016-12-08 | 2017-05-17 | 华中科技大学 | Social network site false fan detection method achieved on basis of network crawler by means of machine learning |
| CN108052528A (en) * | 2017-11-09 | 2018-05-18 | 华中科技大学 | A kind of storage device sequential classification method for early warning |
Non-Patent Citations (2)
| Title |
|---|
| How to find a GSMem malicious activity via an AI approach;ZHU Wei-Jun等;《Arxiv》;20180108;全文 * |
| 基于DNA计算的线性时序逻辑模型检测方法;朱维军等;《计算机学报》;20161231;第39卷(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108710912A (en) | 2018-10-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109241740B (en) | Malicious software benchmark test set generation method and device | |
| WO2021139279A1 (en) | Data processing method and apparatus based on classification model, and electronic device and medium | |
| CN111600919B (en) | Method and device for constructing intelligent network application protection system model | |
| CN112492059A (en) | DGA domain name detection model training method, DGA domain name detection device and storage medium | |
| CN109145030B (en) | Abnormal data access detection method and device | |
| CN109657461B (en) | RTL hardware Trojan horse detection method based on gradient lifting algorithm | |
| CN114692156B (en) | Memory segment malicious code intrusion detection method, system, storage medium and equipment | |
| CN108710912B (en) | Time sequence logic approximate model detection method and system based on two-classification machine learning | |
| CN112464233A (en) | RNN-based malicious software detection method on cloud platform | |
| CN112491891A (en) | Network attack detection method based on hybrid deep learning in Internet of things environment | |
| CN112035345A (en) | Mixed depth defect prediction method based on code segment analysis | |
| CN114285587B (en) | Domain name identification method and device and domain name classification model acquisition method and device | |
| CN112966507A (en) | Method, device, equipment and storage medium for constructing recognition model and identifying attack | |
| CN116663018A (en) | Vulnerability detection method and device based on code executable path | |
| CN115858002B (en) | Binary code similarity detection method and system based on graph comparison learning and storage medium | |
| CN111737694A (en) | Behavior tree-based malicious software homology analysis method | |
| CN111786999A (en) | Intrusion behavior detection method, device, equipment and storage medium | |
| CN111221704B (en) | Method and system for determining running state of office management application system | |
| CN111027318B (en) | Industry classification method, device and equipment based on big data and storage medium | |
| CN110766338A (en) | DPOS (distributed data processing) bifurcation prediction model method based on artificial intelligence and EOS (Ethernet over Ethernet) and IO (input/output) of block chain technology | |
| Rickli | Learning Policies for Path Selection in Attack Graphs | |
| CN117792794B (en) | Network threat information analysis method, device and system | |
| CN113886579B (en) | Construction method and system, identification method and system for positive and negative surface models of industry information | |
| CN113836292B (en) | Structuring method, system, device and medium for biomedical literature abstract | |
| CN115116594B (en) | Method and device for detecting effectiveness of medical device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |