CN107967208A - A kind of Python resource sensitive defect code detection methods based on deep neural network - Google Patents
A kind of Python resource sensitive defect code detection methods based on deep neural network Download PDFInfo
- Publication number
- CN107967208A CN107967208A CN201610915633.4A CN201610915633A CN107967208A CN 107967208 A CN107967208 A CN 107967208A CN 201610915633 A CN201610915633 A CN 201610915633A CN 107967208 A CN107967208 A CN 107967208A
- Authority
- CN
- China
- Prior art keywords
- code
- resource
- version
- mrow
- pattern
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 69
- 238000001514 detection method Methods 0.000 title claims abstract description 27
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 33
- 238000012360 testing method Methods 0.000 claims abstract description 30
- 238000012423 maintenance Methods 0.000 claims abstract description 21
- 238000003062 neural network model Methods 0.000 claims abstract description 15
- 238000011161 development Methods 0.000 claims abstract description 8
- 230000002596 correlated effect Effects 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 11
- 230000000875 corresponding effect Effects 0.000 claims description 10
- 239000000203 mixture Substances 0.000 claims description 10
- 239000000284 extract Substances 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 9
- 230000005856 abnormality Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000009472 formulation Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 239000012634 fragment Substances 0.000 claims 1
- 230000003068 static effect Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000000877 morphologic effect Effects 0.000 description 3
- 230000007935 neutral effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005538 encapsulation Methods 0.000 description 1
- 238000002513 implantation Methods 0.000 description 1
- 235000015110 jellies Nutrition 0.000 description 1
- 239000008274 jelly Substances 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000001568 sexual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Stored Programmes (AREA)
Abstract
The present invention is a kind of Python resource sensitive defect code detection methods based on deep neural network, is comprised the following steps：1) source code of old version and the source code of version to be measured of same software are obtained；2) use pattern infers the resource sensitive code pattern for extracting each version；3) correlated characteristic of resource sensitive code pattern is extracted；4) each characteristic similarity between defect code pattern and security code pattern, defect code pattern and code pattern to be measured is calculated, generates feature vector, and obtain training set and test set；5) feature merging is carried out with training set training deep neural network model, then to the pattern in test set to calculating the degree of correlation with deep neural network model and sorting；6) in program development and maintenance phase, the resource object operation that mistake may occur is reminded according to relevancy ranking result, auxiliary development and maintenance；The present invention solves the problems such as automatic mode that presently, there are and lack for the identification of Python resource sensitive code and defect code detection, and then reduces software hazard, improves software quality, so as to improve developer and guardian's exploitation and the efficiency for safeguarding software.
Description
Technical field
The invention belongs to field of computer technology, especially software technology field, and in particular to one kind based on deep
Spend the Python resource sensitive aacode defect code detection methods of neutral net.
Background technology
With the continuous development of software application technology, requirement of the user to software quality is higher and higher, software developer
Met the needs of users by various technologies.Resource sensitive code is the code block or sentence of a process resource object.
All there is abnormal hidden danger in exploitation and maintenance phase in software, many resource sensitive codes, often just having in maintenance process can
It can be found.With the continuous prevalence of agile development technology, version upgrading is frequent, causes resource sensitive code to trigger exception suddenly
Situation constantly occur.It is for the most traditional solution method of resource sensitive code abnormality processing：It is crucial using try-except
Word is captured and handled.However, developer in the development phase, often ignores abnormality processing, so as to cause program to happen suddenly
Sexual abnormality, causes application crashes.Thus the identification to resource object risky operation and detection are that program development and maintenance phase must
Indispensable step, it can effectively improve program quality, and help is developed finds procedural problem in time with maintenance personnel, so that
Formulate more effective solution.
At present, Python has become the programming language that developer extremely favors.Now, major open source community Python should
With continuing to bring out, a huge ecosystem is formd.Python is a kind of object-oriented, explanation type program language, is had
There is succinct, graceful, practical feature.As a kind of dynamic language, Python is more applied to design the Internet, applications, figure
User interface and script implantation etc., so as to be related to various types of resources.Due to the dynamic language characteristic of Python,
Developer often dynamically changes types of variables, causes uneasy full operation numerous.On the other hand, Python is carried out to resource object
During operation, usually since various exceptions occur in the reasons such as resource distribution, and it is not easy the problem of the operation generation of this resource sensitive
It is found.At present, developer is detected using condition, the mode such as abnormality processing controls these aacode defects.
At this stage, the method for identifying and detecting resource object can substantially be divided into two classes.One kind is based on program analysis number
According to method, it can be according to logic and semantic analysis locating resource object risky operation.In contrast, another kind of is using letter
The method of retrieval is ceased, resource object and detection defect code are identified by means of the mode of machine learning.First method is based on language
Justice analysis, generation that can be quickly is as a result, still have the problems such as accuracy rate is low, semantic rules is difficult to definition.And second of side
Method, by the mode extraction feature such as context, is then learnt and is predicted using the mode of machine learning, although producing result
It is relatively slow, but have the characteristics that accuracy rate is high, highly practical.The present invention is exactly to be detected by the way of machine learning.
In maintenance phase, developer submits every time may repair the defects of many places are identical at the same time, so that same version
Defect code has very strong correlation.The present invention tells defect code and security code, and profit according to history restoration information
With the correlation between defect code, thus it is speculated that with historic defects code, similar code is likely to existing defects, further
Provide a kind of Python resource sensitive defect code detection methods based on deep neural network.
The content of the invention
The present invention provides a kind of Python resource sensitive defect code detection methods based on deep neural network.We
Method is found out code similar to its in code to be measured, is reminded by excavating and comparing the defects of being repaired in old version code
Developer and guardian pay attention to there may be it is identical the problem of, to repair as early as possible.This method is from software version control system
Collect the old version of same Python softwares and version to be measured；For old version, then identified by type inference
Resource sensitive code pattern, and extract corresponding pattern feature, according to history restoration information by drawbacks described above code pattern and peace
Full code pattern composition associative mode pair and irrelevant pattern pair, and characteristic similarity generation feature vector is calculated, trained
Collection；For version to be measured, using identical method extraction different mode and individual features, by old version defect code pattern and
Version compositional model pair to be measured, and characteristic similarity generation feature vector is calculated, obtain test set.Then, using training
Collection training deep neural network model, carries out feature merging to test set by trained deep neural network model, is treated
Survey the degree of correlation between code and defect code.Finally, be ranked up according to the degree of correlation, identify in code to be measured with history version
Originally the closely similar potential danger code of the resource sensitive code that is repaired, so as to propose to build to program developer and guardian
View, prevents abnormal generation.Present invention seek to address that presently, there are shortage for Python resource sensitive code identification and
The problems such as automatic mode of defect code detection, and then software hazard is reduced, software quality is improved, is opened so as to improve developer
Feel like jelly the efficiency of part.
To reach above-mentioned purpose, the present invention proposes a kind of Python resource sensitive defect codes based on deep neural network
Detection method method comprises the following steps：
1) source code of old version and the source code of version to be measured of same software are obtained；
2) use pattern infers the resource sensitive code pattern for extracting each version；
3) correlated characteristic of resource sensitive code pattern is extracted；
4) calculate each between defect code pattern and security code pattern, defect code pattern and code pattern to be measured
Characteristic similarity, generates feature vector, and obtains training set and test set；
5) feature merging is carried out with training set training deep neural network model, then to the pattern in test set to depth
Degree neural network model calculates the degree of correlation and sorts；
6) in program development and maintenance phase, the resource object that mistake may occur is operated according to relevancy ranking result
Reminded, auxiliary development and maintenance.
Further, wherein above-mentioned steps 1) comprise the following steps that：
Step 1) -1：Initial state；
Step 1) -2：According to filename and version information, from the history version for the same software of version control system acquisition of increasing income
The source program and the source program of version to be measured being repaired in this；
Step 1) -3：The collection of software different editions source program finishes.
Further, wherein above-mentioned steps 2) comprise the following steps that：
Step 2) -1：Initial state；
Step 2) -2：Morphological analysis and syntactic analysis are carried out to the source program of each version respectively, use Python java standard libraries
In ast modules generate the corresponding abstract syntax tree of each version；
Step 2) -3：According to the abstract syntax defined in Python java standard libraries, each type of Python, each class are encapsulated
Type has a mapping table table, built-in attribute name or api interface name comprising the type.
Step 2) -4：Ergodic abstract syntax tree, and according to the type and module of encapsulation, infer the possibility class of each variable
Type.Extract the variable of resource object types.
Step 2) -5：For unidentified type, if the variable is an interface name, and there is resource object class in its parameter
Type, then be identified as resource object types, if if it is not, then the variable is its dependent variable member；If calling variable is resource
Object type, is also identified as resource object types.
Step 2) -6：The code snippet of resource object types variable will be called as sensitive resource code pattern.
Step 2) -8：Resource sensitive code pattern information is collected and finished.
Further, wherein above-mentioned steps 3) comprise the following steps that：
Step 3) -1：Initial state；
Step 3) -2：According to resource language pattern information, locating resource Object Operations position, extraction API (parameter type,
Reference order), resource name, call structure and function internal structure etc. be used as feature.
Step 3) -3：By API (parameter type, quantity), resource name, structure and function structure is called to unitize name.
Step 3) -4：Resource language pattern feature information extraction finishes.
Further, wherein above-mentioned steps 4) comprise the following steps that：
Step 4) -1：Initial state；
Step 4) -2：The resource sensitive code pattern that will identify that is divided into three classes, and is respectively defect code pattern, safe generation
Pattern and code pattern to be measured；
Step 4) -3：For old version, will be similar according to history restoration information the defects of code pattern match two-by-two, group
Into associative mode pair；Defect code pattern and the security code pattern similar to its are matched two-by-two, form irrelevant pattern pair；
Step 4) -4：For version to be measured, defect code pattern and code pattern to be measured are matched two-by-two, form mould to be measured
Formula pair；
Step 4) -5：Each characteristic similarity of different mode pair is calculated, and generates feature vector；
Step 4) -6：Training set is obtained to the set of eigenvectors formed by the code pattern of old version, by version to be measured
Code pattern test set is obtained to the set of eigenvectors of composition；
Step 4) -7：Training set test set information, which is collected, to be finished；
Further, wherein above-mentioned steps 5) comprise the following steps that：
Step 5) -1：Initial state；
Step 5) -2：With the similar degrees of data training deep neural network of the training set of generation in step 4), model is obtained
Each parameter value；
Step 5) -3：Using the test set of generation in step 4) as input, pass through trained deep neural network mould
Type, obtains relevance degree；
Step 5) -4：According to the relevance degree calculated, the degree of correlation between all codes pair is arranged from big to small
Sequence, take before k test pattern to as resource sensitive code detection as a result, edition code to be measured therein is labeled as possible
Resource sensitive defect code.
Step 5) -5：Possible resource sensitive defect code mark finishes.
Further, wherein above-mentioned steps 6) comprise the following steps that：
Step 6) -1：Initial state；
Step 6) -2：Code for being labeled as sensitive resource, prompts exploitation and maintenance personnel's history version related to this
The position occurred in this, it is proposed that make it modify, and provide a kind of recovery scenario.
Step 6) -3：In program development and maintenance phase, system is automatically to submitting code to be detected, for there are potential
The operation of dangerous resource, provides warning.
Step 6) -4：Using the version program newly submitted as old version data, for comparing next time so that detection knot
Fruit is more accurate.
Step 6) -5：Resource sensitive defect code prompting in code to be measured finishes.
The present invention carries out feature merging based on deep neural network, and code to be measured is weighed using the metric of a standard
Correlation levels between the defects of old version code, so as to navigate to resource sensitive defect code block, deeply
To basic statement level.After resource sensitive code is identified according to type inference, solved according in the old version similar to its
Certainly scheme, is repaired and prompts developer and guardian automatically.By the above method, have identified resource sensitive code and its
Risky operation, improves the efficiency of software development, is beneficial to the software application product for developing high quality.
Brief description of the drawings
Fig. 1 is a kind of Python resource sensitive defect codes detection side based on deep neural network of the embodiment of the present invention
The general frame figure of method.
Fig. 2 is a kind of Python resource sensitive defect codes detection side based on deep neural network of the embodiment of the present invention
The flow chart of method.
Fig. 3 is the possible abstract syntax tree schematic diagram of a loop control structure.
Embodiment
The method of the present invention by the software version control system such as CVS, have collected all of same Python softwares first
The source code that old version is repaired.Then morphological analysis and grammer point are carried out to the source code of old version and version to be measured
Analysis, according to the abstract syntax tree of generation, carries out type inference, marks out the variable of resource object operation, identify resource language
Pattern, and picked out from the resource sensitive code pattern of each version of history according to history restoration information defect code pattern and
Security code pattern, composition associative mode pair and irrelevant pattern pair.Then, version resource sensitive code pattern to be measured and will go through
History defect code pattern forms test pattern pair.Then, according to the pattern feature of extraction, each pattern is calculated to each feature
Similarity, and feature vector is generated, obtain corresponding training set and test set.Then, using above-mentioned training set training depth god
Through network model, trained deep neural network model is subjected to feature merging to test set, obtain code pattern to be measured with
The corresponding degree of correlation between historic defects code pattern.Finally, it is ranked up according to the degree of correlation, k relevant patterns before selection
Pair as a result, the code to be measured of code centering is labeled as the sensitive resource sensitive code for having latent defect, with this in program
Auxiliary development and maintenance personnel are developed and are safeguarded in exploitation and maintenance process, prevent abnormal generation.
In order to which the technology contents of the present invention are better described, spy coordinates appended diagram to be explained as follows.
The general frame figure of the present invention is as shown in Figure 1, flow chart is as shown in Figure 2.One kind proposed by the present invention is based on depth
The Python resource sensitive defect code detection methods of neutral net, including following 6 steps：
Step 1：Obtain the source code of source code that same software old version is repaired and version program to be measured.CVS etc.
All versions of the program are saved in software version control system, and are labelled with version number.Can according to the version number of formulation,
Obtain the old version of same Python softwares and version source code to be measured.
Step 2：Usage type infers that mode extracts the resource language pattern of each version program source code.First, first to step
The source code of each version obtained in rapid 1 carries out morphological analysis and syntactic analysis, utilizes the ast module phases in Python java standard libraries
Function is answered to generate abstract syntax tree.In abstract syntax tree, each tree interior joint and subtree correspond to a source code entity.
In order to preferably carry out type inference, type that we define according to Python encapsulates several abstract types Types.Often
There are a table attribute, expression and the title in current type attribute or the relevant abstract syntax tree of calling in a Types,
Such as append；For each node in abstract syntax tree, we are provided with type and value, while are provided with node only
One identifier id.For each tree interior joint, t (x) represents the type of the type, i.e. node of node, such as assignment statement.V (x) tables
Show the value of node, be the text representation of the node, such as the particular content of the assignment statement.Id (x) represents unique mark of node
Symbol is known, for distinguishing node.
Such as：Assignment statement is a simple sentence, corresponding to a leaf node in abstract syntax tree, the leaf
The type of node is " assign statement ", value are the content of assignment statement；While Do statements correspond to abstract
A stalk tree in syntax tree, the type of the root node of the subtree is " while statement ", value are while statement
Rule of judgment, child nodes are while internal statements content and the sentence content for jumping out circulation.Fig. 3 is a Do statement knot
The possible abstract syntax tree of structure.
Finally, the whole abstract syntax tree of postorder traversal, maps according to table in the type information and node of abstract syntax tree
The information such as the relevant attribute of each type, infer type of variables, the code piece for the calling resource object variable that will conclude that
Section is labeled as resource sensitive code pattern.Resource sensitive code pattern refers to resource object (file object, graphical user interface
Object etc.) code snippet that is operated.
Such as：
In the code snippet, self is a resource object, and have invoked switch_backends function pairs, it is grasped
Make.Therefore, here it is a resource sensitive code pattern.
Step 3：By step 2, we have extracted resource language pattern from source code.The resource that the present invention extracts is quick
Sense code pattern correlated characteristic be：API (parameter type, reference order), resource name, call structure and function structure.
Then, by the feature Naming conventions of extraction.Wherein, for API features, parameter type and reference order meter are used
Calculate characteristic similarity；For resource name feature, the word order column count characteristic similarity in resource name is used；It is special for calling structure
Sign, characteristic similarity is used as using structural similarity is called；For function structure feature, feature is used as using function structure feature
Similarity.
Step 4：Firstly, for old version, will be similar according to history restoration information the defects of code pattern match two-by-two,
Form associative mode pair；Defect code pattern and the security code pattern similar to its are matched two-by-two, form irrelevant pattern
It is right.For version to be measured, defect code pattern and code pattern to be measured are matched two-by-two, form test pattern pair.Pass through step
3, we with the characteristic information of decimation pattern, can calculate each characteristic similarity of different mode pair.
The characteristic similarity of API uses rVSM algorithms, wherein for parameter type, power is calculated using the algorithm of TF-IDF
Weight, formula are as follows：
Wherein TF appears in the frequency in API, Total for the typeapiFor API sums, ContaintypeTo include such
The API quantity of type.The present invention uses weight of the above method as the API feature vectors formed, meanwhile, adopted for type sequence
Measured with 2-Grams, change of this method for type sequence has robustness.By type sequence and the degree of parameter type
Amount one feature vector of composition.For the feature vector of two version generations, similarity is calculated using rVSM algorithms.In this method
In, COS distance between old version feature vector a and version feature vector b to be measured represents similarity, and formula is as follows：
Wherein,WithOld version feature vector a and version feature vector b to be measured is represented respectively,Represent two
The inner product of feature vector.
Resource name characteristic similarity uses text similarity measurement algorithm.First, resource name is parsed into by a sequence word combination
The form formed.Next, for the resource name R in old version1With the resource name R in version to be measured2, calculation formula is as follows：
Wherein, lcs (R1, R2) represent R1In all sub- word in R2In appearance number, so as to obtain resource name
Quantized value, generate relevant vector.Such as " length " and " getLength ", itsAnd
" getLength " " getlength ", its
For function structure characteristic similarity and architectural feature similarity is called, according to the abstract syntax obtained in step 2
Tree, travels through tree construction, obtains corresponding similarity by the identical number of tree node, calculating probability, is required.Finally, by going through
The code pattern of history version obtains training set to the set of eigenvectors of composition, by the code pattern of version to be measured to the feature that forms
Vector set obtains test set.
Step 5：By step 4, we can obtain the training set and test set being made of feature vector.Since it can not
Entirety indicate whether it is related to some dangerous resource object operation, thus, here we using the algorithm of deep neural network come
Realize that feature merges, and calculate the degree of correlation.
First, using the training set training deep neural network of generation.The neutral net that the present invention designs is divided into three layers,
Respectively input layer, hidden layer -1, hidden layer -2 and output layer.Wherein hidden layer -1 is twice of input layer number, hidden
Hide the half that the node of layer -2 is input layer number.The each node H1 of hidden layer -1iCalculation formula it is as follows：
Wherein w1i, b is to need trained parameter, InputiFor input node value.Similarly hidden layer -2 by the formula by
Hidden layer -1 is derived from.Training for w and b, the method that the present invention is declined using batch gradient, step are as follows：
1) initialize：Δw(l)=0, Δ b(l)Then random initializtion is smaller numerical value by=0, w and b；
2) assume that iterations is m, for i from 1 to m, calculate gradient using BP algorithm and add up：
Wherein,
3) undated parameter：
Wherein, λ is optional parameters, takes 2 in the present invention.By above-mentioned training method, deep neural network mould is trained
Type.
In detection-phase, using the feature vector for testing each pattern pair in set as input, pass through above-mentioned node formula
Calculated.Last output is a relevance degree, represents the degree of correlation of the pattern pair.Used here as nonlinear nerve
Network method can preferably react correlation levels than using linear information retrieval method, significant effect.
In deep neural network, intermediate layer and the weights each linked of input layer are trained by the data of old version
Obtain, corresponding weights are as the same.Meanwhile by largely training, change partial link and weights among neuron, so that excellent
Change output result.
For obtained relevance degree, we according to being ranked up from big to small, and before choosing k pattern to as output
As a result.
Step 6：According to the very big sensitive code to be measured of the obtained degree of correlation, the position for reminding exploitation to occur with maintenance personnel
With history resource operation related to this, and to the abnormality processing scheme of this resource before providing, and give a warning.For
The Python source codes detected are used to detect next time as old version data, and Detection accuracy is improved with this.For firm
The Python source codes of submission, are detected automatically, and send alarm to exploitation and maintenance personnel according to result.
Such as：In old version, the operation to resource object of certain is as follows：
In the old version, self variables are a resource objects, are to carry out reading operation to the object here.Exploitation
There is exception in order to prevent in person, and try_catch abnormality processings are added in sentence periphery.
And occurs sentence as follows in the source code of version to be measured：
Def read_bytes (self, num_bytes, callback=None, streaming_callback=
None,
Partial=False)：
self._try_inline_read()
Here read operation has been carried out and to resource object, and has used identical API, but does not carry out exception
Reason.By above-mentioned two code combination into code pair, by means of the invention it is also possible to identify and detect whether both related,
So that it is determined that whether code to be measured is sensitive resource code, and remind developer and guardian to be handled, while provide correlation
Old version code information.
In conclusion the present invention provides a kind of Python resource sensitives defect code detection based on deep neural network
Method, solves automatic mode for lacking identified for Python resource sensitive code detection and risky operation at present etc.
Problem, improves software application quality, ensure that the controllability in software evolution process.
Claims (7)
1. a kind of Python resource sensitive defect code detection methods based on deep neural network, it is characterised in that from software
Version control system, collects the old version of same Python softwares and version to be measured；For old version, then pass through class
Type, which is inferred, identifies resource sensitive code pattern, and extracts corresponding pattern feature, according to history restoration information by drawbacks described above
Code pattern and security code pattern composition associative mode pair and irrelevant pattern pair, and calculate characteristic similarity generate feature to
Amount, obtains training set；For version to be measured, using identical method extraction different mode and individual features, old version is lacked
Code pattern and version compositional model pair to be measured are fallen into, and calculates characteristic similarity generation feature vector, obtains test set；So
Afterwards, deep neural network model is trained using training set, trained deep neural network model is subjected to feature to test set
Merge, obtain the degree of correlation between code to be measured and defect code；Finally, it is ranked up according to the degree of correlation, k correlation before selection
Code pair as a result, the code to be measured of code centering to be labeled as to the resource sensitive code of latent defect, detect to endanger
Dangerous resource object operation, and auxiliary information is provided；This method comprises the following steps：
1) source code of old version and the source code of version to be measured of same software are obtained；The software version control system such as CVS
In save all versions of the program and submit, and have standardized version number, can be obtained same according to the version number of formulation
The old version of Python softwares and version source code to be measured；
2) use pattern infers the resource sensitive code pattern for extracting each version；To the old version gathered in step 1
Morphology and syntactic analysis are carried out with version source code to be measured, it is corresponding abstract using the ast modules generation in Python java standard libraries
Syntax tree, Python types are abstracted, and set type and value to each node, reuse global type inference
Method, extract resource sensitive code pattern；
Resource sensitive code pattern refers to the code operated to resource object (file object, graphical user interface object etc.)
Fragment；
Such as：
In the code snippet, self is a resource object, and have invoked switch_backends function pairs, it is operated；
Therefore, here it is a resource sensitive code pattern；
Define 1：Python java standard libraries are issued with Python, contain the interior modeling for providing various system level functions
Block；
Define 2：Type inference is a kind of by carrying out static analysis to source code, infers the side of types of variables in dynamic language
Method；
Define 3：Type is used to identify the node type information in abstract syntax tree, is worth the abstract syntax defined from Python,
Including function_call etc.；
Define 4：Value is the text representation of the node content in abstract syntax tree, such as the Rule of judgment of while control structures
Deng；
3) correlated characteristic of resource sensitive code pattern is extracted；By step 2, we have extracted resource sensitive from source code
Code pattern；The correlated characteristic for the resource sensitive code pattern that the present invention extracts is：API (parameter type, reference order), resource
Name, call structure and function structure；Finally, by the feature Naming conventions of extraction；
Define 1：For API features, characteristic similarity is calculated using parameter type and reference order；
Define 2：For resource name feature, the word order column count characteristic similarity in resource name is used；
Define 3：For calling architectural feature, characteristic similarity is used as using structural similarity is called；
Define 4：For function structure feature, characteristic similarity is used as using function structure feature；
4) each feature between defect code pattern and security code pattern, defect code pattern and code pattern to be measured is calculated
Similarity, generates feature vector, and obtains training set and test set；, will be similar according to history restoration information for old version
The defects of code pattern match two-by-two, form associative mode pair；By defect code pattern and the security code pattern similar to its
Match two-by-two, form irrelevant pattern pair；For version to be measured, defect code pattern and code pattern to be measured are matched two-by-two,
Form test pattern pair；Then, each characteristic information extracted according to step 3, each feature for calculating different mode pair are similar
Degree, and generate feature vector；Finally, training set is obtained to the set of eigenvectors formed by the code pattern of old version, by treating
The code pattern for surveying version obtains test set to the set of eigenvectors of composition；
Define 1：Defect code pattern refers to the resource sensitive defect code pattern being then repaired in history restoration information；
Define 2：Security code pattern refer to it is similar to defect code pattern but without find defect resource sensitive code mould
Formula；
Define 3：The characteristic similarity of API uses rVSM algorithms, wherein for parameter type, power is calculated using the algorithm of TF-IDF
Weight, formula are as follows：
<mrow>
<mi>T</mi>
<mi>F</mi>
<mo>-</mo>
<mi>I</mi>
<mi>D</mi>
<mi>F</mi>
<mo>=</mo>
<mi>T</mi>
<mi>F</mi>
<mo>*</mo>
<mi>l</mi>
<mi>o</mi>
<mi>g</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msub>
<mi>Total</mi>
<mrow>
<mi>a</mi>
<mi>p</mi>
<mi>i</mi>
</mrow>
</msub>
</mrow>
<mrow>
<msub>
<mi>Contain</mi>
<mrow>
<mi>t</mi>
<mi>y</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
</msub>
<mo>+</mo>
<mn>1</mn>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
</mrow>
Wherein TF appears in the frequency in API, Total for the typeapiFor API sums, ContaintypeTo include the type
API quantity；The present invention uses weight of the above method as the API feature vectors formed, meanwhile, use 2- for type sequence
Grams is measured, and change of this method for type sequence has robustness；By type sequence and the set of measurements of parameter type
Into a feature vector；The feature vector of generation for two versions, similarity is calculated using rVSM algorithms；In this method
In, COS distance between old version feature vector a and version feature vector b to be measured represents similarity, and formula is as follows：
<mrow>
<mi>S</mi>
<mi>i</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mo>,</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>C</mi>
<mi>o</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mo>,</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mover>
<msub>
<mi>V</mi>
<mi>a</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
<mo>&CenterDot;</mo>
<mover>
<msub>
<mi>V</mi>
<mi>b</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
</mrow>
<mrow>
<mo>|</mo>
<mover>
<msub>
<mi>V</mi>
<mi>a</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
<mo>|</mo>
<mo>|</mo>
<mover>
<msub>
<mi>V</mi>
<mi>b</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
Wherein,WithOld version feature vector a and version feature vector b to be measured is represented respectively,Represent two features
The inner product of vector；
Define 4：The characteristic similarity of resource name uses text similarity measurement algorithm；First, resource name is parsed into by a sequence word
The form being composed；Next, for the resource name R in old version1With the resource name R in version to be measured2, calculation formula
It is as follows：
<mrow>
<mi>S</mi>
<mi>i</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<mi>l</mi>
<mi>c</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mo>+</mo>
<mo>|</mo>
<mi>l</mi>
<mi>c</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>|</mo>
<mo>+</mo>
<mo>|</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
Wherein, lcs (R1, R2) represent R1In all sub- word in R2In appearance number, so as to obtain the amount of resource name
Change value, generates relevant vector；
Define 5：RVSM algorithms are vector space model, are a kind of algorithms for calculating similarity；
5) feature merging is carried out with training set training deep neural network model, then to the pattern in test set to refreshing with depth
The degree of correlation is calculated through network model and is sorted；The training set training deep neural network model generated using step 2), then
The test set of step 2) generation is subjected to feature merging using trained deep neural network model, and calculates the degree of correlation；
Finally by the relevance degree between defect code pattern and code pattern to be measured according to being ranked up from big to small, and choose k generation
Code is to as output result；
6) in program development and maintenance phase, the resource object that mistake may occur is operated according to relevancy ranking result and is carried out
Remind, auxiliary development and maintenance；According to the very big resource sensitive code to be measured of the obtained degree of correlation, exploitation and maintenance personnel are reminded
The position of appearance and history resource operation related to this, and to the abnormality processing scheme of this resource, concurrent responding before providing
Accuse；It is used to detect next time as old version data for the Python source codes crossed after testing, it is accurate to improve detection with this
True rate；For the Python source codes just submitted, it is detected automatically, and announcement is sent to exploitation and maintenance personnel according to result
It is alert.
2. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that, in step 1), according to the version number of formulation, obtains the old version of same Python softwares and version source to be measured
Code.
3. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that, in step 2), morphology and syntactic analysis is carried out to the old version gathered and version source code to be measured,
Corresponding abstract syntax tree is generated using the ast modules in Python java standard libraries, Python types are abstracted, are reused
The method of global type inference, extracts resource sensitive code pattern.
4. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that, in step 3), according to the resource language pattern information of collection, extracts following correlated characteristic：API (parameter type,
Reference order), resource name, call structural similarity and function structure similarity；Finally, by the feature Naming conventions of extraction.
5. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that, in step 4), calculates defect code pattern and security code pattern, defect code pattern and code pattern to be measured
Between each characteristic similarity, generate feature vector, and obtain training set and test set；For old version, according to history
The defects of restoration information will be similar code pattern matches two-by-two, forms associative mode pair；By defect code pattern and similar to its
Security code pattern match two-by-two, form irrelevant pattern pair；For version to be measured, by defect code pattern and code to be measured
Pattern is matched two-by-two, forms test pattern pair；Then, each characteristic information extracted according to step 3, calculates different mode pair
Each characteristic similarity, and generate feature vector；Finally, the set of eigenvectors formed is obtained by the code pattern of old version
Training set, test set is obtained by the code pattern of version to be measured to the set of eigenvectors formed.
6. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that, in step 5), training deep neural network model carries out feature merging, to test code to being arranged according to the degree of correlation
Sequence；The training set training deep neural network model generated using step 2), is then made the test set that step 2) generates
Feature merging is carried out with trained deep neural network model, and is calculated related between code to be measured and historic defects code
Degree；Finally, according to obtained relevance degree, we according to being ranked up from big to small, and before choosing k code to as output
As a result.
7. the Python resource sensitive defect code detection methods according to claim 1 based on deep neural network, its
It is characterized in that,, can according to relevancy ranking result pair in program development and maintenance phase according to ranking results in step 6)
The resource object operation that mistake can occur is reminded, auxiliary development and maintenance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610915633.4A CN107967208B (en) | 2016-10-20 | 2016-10-20 | Python resource sensitive defect code detection method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610915633.4A CN107967208B (en) | 2016-10-20 | 2016-10-20 | Python resource sensitive defect code detection method based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107967208A true CN107967208A (en) | 2018-04-27 |
| CN107967208B CN107967208B (en) | 2020-01-17 |
Family
ID=61996517
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610915633.4A Active CN107967208B (en) | 2016-10-20 | 2016-10-20 | Python resource sensitive defect code detection method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107967208B (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241739A (en) * | 2018-07-19 | 2019-01-18 | 中国科学院信息工程研究所 | Android malware detection methods, device and storage medium based on API |
| CN109446078A (en) * | 2018-10-18 | 2019-03-08 | 网易（杭州）网络有限公司 | Code test method and device, storage medium, electronic equipment |
| CN109657461A (en) * | 2018-11-26 | 2019-04-19 | 浙江大学 | RTL hardware Trojan horse detection method based on gradient boosting algorithm |
| CN109726120A (en) * | 2018-12-05 | 2019-05-07 | 北京计算机技术及应用研究所 | A kind of software defect confirmation method based on machine learning |
| CN110162245A (en) * | 2019-04-11 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Analysis method, device, electronic equipment and the storage medium of graphic operation |
| CN110175128A (en) * | 2019-05-29 | 2019-08-27 | 北京百度网讯科技有限公司 | A kind of similar codes case acquisition methods, device, equipment and storage medium |
| CN110349477A (en) * | 2019-07-16 | 2019-10-18 | 湖南酷得网络科技有限公司 | A kind of misprogrammed restorative procedure, system and server based on history learning behavior |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN110780878A (en) * | 2019-10-25 | 2020-02-11 | 湖南大学 | Method for carrying out JavaScript type inference based on deep learning |
| CN110825642A (en) * | 2019-11-11 | 2020-02-21 | 浙江大学 | Software code line-level defect detection method based on deep learning |
| CN111427775A (en) * | 2020-03-12 | 2020-07-17 | 扬州大学 | Method level defect positioning method based on Bert model |
| CN111459789A (en) * | 2019-08-28 | 2020-07-28 | 南京意博软件科技有限公司 | Detection method and device for application programming interface |
| CN111858323A (en) * | 2020-07-11 | 2020-10-30 | 南京工业大学 | Code representation learning-based instant software defect prediction method |
| CN111913874A (en) * | 2020-06-22 | 2020-11-10 | 西安交通大学 | Software defect tracing method based on syntactic structure change analysis |
| CN111913718A (en) * | 2020-06-22 | 2020-11-10 | 西安交通大学 | Binary function differential analysis method based on basic block context information |
| CN112131120A (en) * | 2020-09-27 | 2020-12-25 | 北京软安科技有限公司 | Source code defect detection method and device |
| CN112328475A (en) * | 2020-10-28 | 2021-02-05 | 南京航空航天大学 | Defect positioning method for multiple suspicious code files |
| CN113407442A (en) * | 2021-05-27 | 2021-09-17 | 杭州电子科技大学 | Pattern-based Python code memory leak detection method |
| CN113408597A (en) * | 2021-06-10 | 2021-09-17 | 北京工业大学 | Java method name recommendation method based on two-stage framework |
| CN113722239A (en) * | 2021-11-01 | 2021-11-30 | 南昌航空大学 | Airborne embedded software quality detection method, device, medium and electronic equipment |
| CN113836020A (en) * | 2021-09-24 | 2021-12-24 | 中国电信股份有限公司 | Code detection method, device and storage medium |
| US11449610B2 (en) * | 2018-03-20 | 2022-09-20 | WithSecure Corporation | Threat detection system |
| CN115454855A (en) * | 2022-09-16 | 2022-12-09 | 中国电信股份有限公司 | Code defect report auditing method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1609855A (en) * | 2003-06-23 | 2005-04-27 | 微软公司 | Query optimizer system and method |
| CN101441571A (en) * | 2008-12-02 | 2009-05-27 | 南京大学 | Gridding system implementing method based on Python language |
| CN105159715A (en) * | 2015-09-01 | 2015-12-16 | 南京大学 | Python code change reminding method on basis of abstract syntax tree node change extraction |
-
2016
- 2016-10-20 CN CN201610915633.4A patent/CN107967208B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1609855A (en) * | 2003-06-23 | 2005-04-27 | 微软公司 | Query optimizer system and method |
| CN100517307C (en) * | 2003-06-23 | 2009-07-22 | 微软公司 | Query optimizer system and method |
| CN101441571A (en) * | 2008-12-02 | 2009-05-27 | 南京大学 | Gridding system implementing method based on Python language |
| CN105159715A (en) * | 2015-09-01 | 2015-12-16 | 南京大学 | Python code change reminding method on basis of abstract syntax tree node change extraction |
Non-Patent Citations (2)
| Title |
|---|
| ZHIFEI CHEN 等: "Tracking Down Dynamic Feature Code Changes Against Python Software Evolution", 《2016 THIRD INTERNATIONAL CONFERENCE ON TRUSTWORTHY SYSTEMS AND THEIR APPLICATIONS》 * |
| 李清言: "Pyreview：一个基于抽象语法树差异提取的Python源代码分析工具", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11449610B2 (en) * | 2018-03-20 | 2022-09-20 | WithSecure Corporation | Threat detection system |
| CN109241739A (en) * | 2018-07-19 | 2019-01-18 | 中国科学院信息工程研究所 | Android malware detection methods, device and storage medium based on API |
| CN109241739B (en) * | 2018-07-19 | 2021-01-05 | 中国科学院信息工程研究所 | API-based android malicious program detection method and device and storage medium |
| CN109446078A (en) * | 2018-10-18 | 2019-03-08 | 网易（杭州）网络有限公司 | Code test method and device, storage medium, electronic equipment |
| CN109446078B (en) * | 2018-10-18 | 2022-02-18 | 网易（杭州）网络有限公司 | Code testing method and device, storage medium and electronic equipment |
| CN109657461A (en) * | 2018-11-26 | 2019-04-19 | 浙江大学 | RTL hardware Trojan horse detection method based on gradient boosting algorithm |
| CN109657461B (en) * | 2018-11-26 | 2020-12-08 | 浙江大学 | RTL hardware Trojan horse detection method based on gradient lifting algorithm |
| CN109726120A (en) * | 2018-12-05 | 2019-05-07 | 北京计算机技术及应用研究所 | A kind of software defect confirmation method based on machine learning |
| CN109726120B (en) * | 2018-12-05 | 2022-03-08 | 北京计算机技术及应用研究所 | Software defect confirmation method based on machine learning |
| CN110162245A (en) * | 2019-04-11 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Analysis method, device, electronic equipment and the storage medium of graphic operation |
| CN110175128A (en) * | 2019-05-29 | 2019-08-27 | 北京百度网讯科技有限公司 | A kind of similar codes case acquisition methods, device, equipment and storage medium |
| CN110349477A (en) * | 2019-07-16 | 2019-10-18 | 湖南酷得网络科技有限公司 | A kind of misprogrammed restorative procedure, system and server based on history learning behavior |
| CN110349477B (en) * | 2019-07-16 | 2022-01-07 | 长沙酷得网络科技有限公司 | Programming error repairing method, system and server based on historical learning behaviors |
| CN111459789B (en) * | 2019-08-28 | 2023-11-03 | 南京意博软件科技有限公司 | Detection method and device for application programming interface |
| CN111459789A (en) * | 2019-08-28 | 2020-07-28 | 南京意博软件科技有限公司 | Detection method and device for application programming interface |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN110780878A (en) * | 2019-10-25 | 2020-02-11 | 湖南大学 | Method for carrying out JavaScript type inference based on deep learning |
| CN110825642A (en) * | 2019-11-11 | 2020-02-21 | 浙江大学 | Software code line-level defect detection method based on deep learning |
| CN111427775A (en) * | 2020-03-12 | 2020-07-17 | 扬州大学 | Method level defect positioning method based on Bert model |
| CN111427775B (en) * | 2020-03-12 | 2023-05-02 | 扬州大学 | Method level defect positioning method based on Bert model |
| CN111913874B (en) * | 2020-06-22 | 2021-12-28 | 西安交通大学 | Software defect tracing method based on syntactic structure change analysis |
| CN111913874A (en) * | 2020-06-22 | 2020-11-10 | 西安交通大学 | Software defect tracing method based on syntactic structure change analysis |
| CN111913718A (en) * | 2020-06-22 | 2020-11-10 | 西安交通大学 | Binary function differential analysis method based on basic block context information |
| CN111858323B (en) * | 2020-07-11 | 2021-06-01 | 南京工业大学 | Code representation learning-based instant software defect prediction method |
| CN111858323A (en) * | 2020-07-11 | 2020-10-30 | 南京工业大学 | Code representation learning-based instant software defect prediction method |
| CN112131120B (en) * | 2020-09-27 | 2022-09-30 | 北京智联安行科技有限公司 | Source code defect detection method and device |
| CN112131120A (en) * | 2020-09-27 | 2020-12-25 | 北京软安科技有限公司 | Source code defect detection method and device |
| CN112328475B (en) * | 2020-10-28 | 2021-11-30 | 南京航空航天大学 | Defect positioning method for multiple suspicious code files |
| CN112328475A (en) * | 2020-10-28 | 2021-02-05 | 南京航空航天大学 | Defect positioning method for multiple suspicious code files |
| CN113407442A (en) * | 2021-05-27 | 2021-09-17 | 杭州电子科技大学 | Pattern-based Python code memory leak detection method |
| CN113408597A (en) * | 2021-06-10 | 2021-09-17 | 北京工业大学 | Java method name recommendation method based on two-stage framework |
| CN113408597B (en) * | 2021-06-10 | 2024-06-04 | 北京工业大学 | Java method name recommendation method based on two-stage framework |
| CN113836020A (en) * | 2021-09-24 | 2021-12-24 | 中国电信股份有限公司 | Code detection method, device and storage medium |
| CN113722239A (en) * | 2021-11-01 | 2021-11-30 | 南昌航空大学 | Airborne embedded software quality detection method, device, medium and electronic equipment |
| CN115454855A (en) * | 2022-09-16 | 2022-12-09 | 中国电信股份有限公司 | Code defect report auditing method and device, electronic equipment and storage medium |
| CN115454855B (en) * | 2022-09-16 | 2024-02-09 | 中国电信股份有限公司 | Code defect report auditing method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107967208B (en) | 2020-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107967208A (en) | A kind of Python resource sensitive defect code detection methods based on deep neural network | |
| Li et al. | Dear: A novel deep learning-based approach for automated program repair | |
| Kovbasistyi et al. | Method for detection of non-relevant and wrong information based on content analysis of web resources | |
| Di Lucca et al. | An approach to identify duplicated web pages | |
| CN109214191A (en) | A method of utilizing deep learning forecasting software security breaches | |
| Shen et al. | A survey of automatic software vulnerability detection, program repair, and defect prediction techniques | |
| CN117951701A (en) | Method for determining flaws and vulnerabilities in software code | |
| CN109657473A (en) | A kind of fine granularity leak detection method based on depth characteristic | |
| CN114297654A (en) | Intelligent contract vulnerability detection method and system for source code hierarchy | |
| CN111459799A (en) | Software defect detection model establishing and detecting method and system based on Github | |
| CN107066262A (en) | Source code file clone's adjacency list merges detection method | |
| CN115495755B (en) | Codebert and R-GCN-based source code vulnerability multi-classification detection method | |
| CN103914374B (en) | The aacode defect detection method and device extracted based on program slice and frequent mode | |
| CN109067800A (en) | A kind of cross-platform association detection method of firmware loophole | |
| CN115033895B (en) | Binary program supply chain safety detection method and device | |
| CN105279086A (en) | Flow chart-based method for automatically detecting logic loopholes of electronic commerce websites | |
| CN106682507A (en) | Virus library acquiring method and device, equipment, server and system | |
| CN106330861A (en) | Website detection method and apparatus | |
| CN114900346A (en) | Network security testing method and system based on knowledge graph | |
| Yang et al. | Smart contract vulnerability detection based on abstract syntax tree | |
| CN114398069A (en) | Method and system for identifying accurate version of public component library based on cross fingerprint analysis | |
| CN110049052A (en) | The malice domain name detection method of label and attribute similarity based on dom tree | |
| Kim | Enhancing code clone detection using control flow graphs. | |
| CN109670311A (en) | Malicious code analysis and detection method based on high-level semantics | |
| Zhao et al. | Suzzer: A vulnerability-guided fuzzer based on deep 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |