CN108108622A - Leakage location based on depth convolutional network and controlling stream graph - Google Patents
Leakage location based on depth convolutional network and controlling stream graph Download PDFInfo
- Publication number
- CN108108622A CN108108622A CN201711325630.6A CN201711325630A CN108108622A CN 108108622 A CN108108622 A CN 108108622A CN 201711325630 A CN201711325630 A CN 201711325630A CN 108108622 A CN108108622 A CN 108108622A
- Authority
- CN
- China
- Prior art keywords
- code
- mrow
- msub
- unit
- neural network
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/50—Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
- G06F21/57—Certifying or maintaining trusted computer platforms, e.g. secure boots or power-downs, version controls, system software checks, secure updates or assessing vulnerabilities
- G06F21/577—Assessing vulnerabilities and evaluating computer system security
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/40—Transformation of program code
- G06F8/53—Decompilation; Disassembly
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
A kind of leakage location based on depth convolutional network and controlling stream graph, including:Preprocessing module, training module and prediction module, wherein:Preprocessing module reads in the object code in bug code storehouse and changes into bivector after generating controlling stream graph, and feature learning module carries out feature extraction and training with depth convolutional network from bivector;The present invention obtains the characteristic model of loophole by analyzing substantial amounts of loophole sample, with the mode of deep learning, and unknown loophole is found with this model, while code to be measured and the code of known bugs are carried out similitude and compared, to find approximate bug code.By artificial intelligence, the difficulty of bug excavation can be reduced, loophole feature is found by machine, screens out the code there is no loophole, promote the efficiency of safety engineer.
Description
Technical field
The present invention relates to a kind of technologies of image processing field, and in particular to and it is a kind of that code is changed into controlling stream graph, it will
Controlling stream graph is re-encoded as image, and the technology of code vulnerabilities feature is extracted by the deep learning to image.
Background technology
Loophole refers to some functional or security logic flaws present in system, causes to threaten including all, damage
All factors of bad computer system security are computer systems in hardware, software, the specific implementation of agreement or system safety
Defect present on strategy and deficiency.Currently used bug excavation technology includes:Artificial detection, Fuzz technologies, binary system pair
Than, static analysis and dynamic analysis.Under real engineering-environment, bug excavation be all using the judgement of people as guide, with reference to
Upper common technology, therefore the efficiency of bug excavation depends primarily upon the ability of safety engineer.By artificial intelligence, can drop
The difficulty of low bug excavation finds loophole feature by machine, screens out the code there is no loophole, promotes the effect of engineer
Rate.
The content of the invention
The present invention is directed to the defects of prior art and deficiency, proposes a kind of leakage based on depth convolutional network and controlling stream graph
Hole detecting system, controlling stream graph is changed by code, and controlling stream graph is re-encoded as image, and figure is carried out with deep learning model
The feature learning of picture, so as to obtain the characteristics of image of loophole.Specific practice is:Source code is changed into assembler language first, two into
Code processed carries out dis-assembling to assembly code.The controlling stream graph of assembly code is generated, controlling stream graph is encoded, so as to control
Flow graph processed changes into vector.By these vector combinations, become bivector, an image has been synthesized so as to be equivalent to.The present invention
Loophole is found by two ways.One kind is the substantial amounts of loophole sample of analysis, and the spy of loophole is obtained with the mode of deep learning
Model is levied, unknown loophole is found with this model;Another kind is that code to be measured is similar to the code progress of known bugs
Property compare, to find approximate bug code.
The present invention is achieved by the following technical solutions:
The present invention relates to a kind of leakage location based on depth convolutional network and controlling stream graph, including:Pre-process mould
Block, training module and prediction module, wherein:Preprocessing module reads in the object code in bug code storehouse and generates control stream
Bivector is changed into after figure, feature learning module carries out feature extraction with depth convolutional network and instructed from bivector
Practice, training module includes:Convolutional neural networks unit for feature learning and the BP neural network for similitude comparison
Unit, feature extraction is carried out to bivector with depth convolutional network for convolutional neural networks unit and training obtains identification loophole
The neural network parameter of feature;BP neural network unit is similar with bug code progress to object code with BP neural network
Property analyze and train to obtain the neural network parameter that differentiates code similitude, prediction module carries out object code prediction work,
The parameter obtained according to training module judges whether comprising loophole, for being determined as that leaky code building loophole is reported.
The bivector, when object code is source code, then preprocessing module is directly translated into assembly code, works as target
Code is binary code, then preprocessing module first carries out dis-assembling and operates to obtain assembly code;It is generated and controlled according to assembly code
Each node of controlling stream graph is mapped to merge after a vector after flow graph processed and is obtained.
The bug code storehouse is the bug code collected from CVE databases.
The preprocessing module includes:Code compilation will be collected for the compiler of assembly code, by Open-Source Tools
Code changes into the controlling stream graph generation unit of controlling stream graph and each nodes encoding of controlling stream graph is become vector and group
Synthesize the coding unit of bivector.
The training module includes:Convolutional neural networks unit, the parameter adjustment unit of convolutional neural networks, BP nerves
The parameter adjustment unit of network element and BP neural network unit, wherein:Convolutional neural networks unit is connected with preprocessing module,
Receive the neural network parameter that bivector carries out feature extraction and training obtains identifying loophole feature, the ginseng of convolutional neural networks
Number adjustment units be connected with convolutional neural networks unit and transmit adjustment after neural network parameter, BP neural network unit in advance
Processing module is connected, and receives the neutral net ginseng that bivector carries out similarity analysis and training obtains differentiating code similitude
Number, the parameter adjustment unit of BP neural network unit, which is connected with BP neural network unit and transmits the neutral net after adjustment, joins
Number.
Technique effect
Compared with prior art, the present invention is based on deep learning, code is automatically detected, is found latent in code
In loophole.Differentiate whether code comprising loophole has machine to complete, and unconventional artificial detection, the present invention can save largely
Cost of labor.
Description of the drawings
Fig. 1 is the depth convolutional neural networks cell schematics for feature learning;
Fig. 2 is the BP neural network schematic diagram for similitude comparison;
Fig. 3 is embodiment flow diagram.
Specific embodiment
The present embodiment includes:Preprocessing module, training module and prediction module, wherein:Preprocessing module reads in loophole
Code library if object code is source code, changes into assembly code, if object code is binary code, carries out anti-
Compilation operation, obtains assembly code.Controlling stream graph is generated from assembly code, controlling stream graph is carried out encoded translated into vector, general
These vector combinations, become bivector, are equivalent to and synthesize an image;Training module is made of two submodules:For
The convolutional neural networks unit of feature learning carries out the process of feature extraction and training with depth convolutional network;For phase
Like the BP neutral net units that property compares, with cluster analysis into the similarity analysis of line code, so as to find bug code.
Prediction module carries out prediction work to code to be measured, judges whether comprising loophole, for being determined as that leaky code is final
It is detected by manually
The bug code storehouse is the bug code collected from CVE databases.
The bug code that the present embodiment is directed to is C language and C Plus Plus.
Code is changed into image by the preprocessing module, is specially:When object code is C language and C Plus Plus, then
Assembly code is compiled as with gcc;When object code is binary code, then by manual type by crossover tool by two into
Code processed is converted into assembly code, and further obtains the controlling stream graph of assembly code by crossover tool, to controlling stream graph into
One step is analyzed to obtain the attribute of controlling stream graph.With structure2vec algorithms, each node of controlling stream graph is mapped to one
A vector, vector are combined into bivector, that is, being piece image.
The controlling stream graph refers to represent all roads that can be traversed in a program process with the form of figure
Footpath is the important tool of Program Static Analysis.
The attribute of the controlling stream graph refers to the attribute between the attribute and node inside flow chart node, specifically
Parameter see the table below:
Table 1
The node of the controlling stream graph is converted into vector especially by structure2vec algorithms, which uses and change
The mode in generation, not only retains the attributive character of node itself, while also takes into account the relation between adjacent node, institute specific as follows
Show:
xvRepresent the d dimensional vectors of the attribute composition in table 1, N (v) represents that node v is scheming adjacent node all in G, often
One node is from μv (0)Start iteration, μv (0)Initial value be filled with for 0 vector, W1It is d × p matrix, p represents vector filling
Length, Pi(i=1 ..., n) it is p × p matrix.
F(xv,∑u∈N(v)μu)=tanh (W1xv+σ(∑u∈N(v)μu))
σ (l)=P1×ReLU(P2×…ReLU(Pnl))
ReLU (x)=max { 0, x }
The structure2vec algorithms choose whole attributes in table 1, i.e. d=8 in the present embodiment.
The training module includes:Convolutional neural networks unit, the parameter adjustment unit of convolutional neural networks, BP nerves
The parameter adjustment unit of network element and BP neural network unit, wherein:Convolutional neural networks unit is connected with preprocessing module,
Receive the neural network parameter that bivector carries out feature extraction and training obtains identifying loophole feature, the ginseng of convolutional neural networks
Number adjustment units be connected with convolutional neural networks unit and transmit adjustment after neural network parameter, BP neural network unit in advance
Processing module is connected, and receives the neutral net ginseng that bivector carries out similarity analysis and training obtains differentiating code similitude
Number, the parameter adjustment unit of BP neural network unit, which is connected with BP neural network unit and transmits the neutral net after adjustment, joins
Number.
The convolutional neural networks unit automatically extracts the feature of bivector, such as Fig. 1 by convolutional neural networks
Shown, the convolutional neural networks unit includes:Sequentially connected input layer, the first convolutional layer, the first active coating, the first pond
Change layer, the second convolutional layer, the second active coating, the second pond layer, the first full articulamentum, the second full articulamentum and active coating, wherein:
The effect of convolutional layer is extraction feature, and the effect of pond layer is compression of images extraction main feature, and the effect of active coating is to introduce
It is non-linear, the effect of full articulamentum be all features of connection for classifying, i.e., by the convolution kernel of two layers 3*3 and the training of pond layer
And after extracting feature, using two layers of full articulamentum, loophole is determined whether.
As shown in Fig. 2, the BP neural network unit by by structure2vec algorithms convert two dimension to
Amount carries out cosine computings, and 1 is exported if object code is approximate with bug code, and difference goes out -1, is specially:
Wherein:G represents controlling stream graph,Represent structure2vec algorithms.
The prediction module includes:Discrimination module, journal module and artificial cognition module, wherein:Discrimination module and instruction
Practice module be connected and receive feature learning and binary system comparison as a result, journal module is connected with discrimination module, for being judged as
Leaky code generates the data flow diagram and controlling stream graph of input, convenient for engineer's artificial detection, artificial cognition module and day
Will module is connected, and when model is determined as that the code of loophole generates data flow diagram by journal module, is submitted to artificial detection.Model
The loophole detected whether be using loophole, will be finally by manually judging to complete.
The result includes:[0,0], [0,1], [1,0], [1,1], wherein:0 represents not including loophole, and 1 indicates leakage
Hole.Two parameters carry out OR operation, that is, the output for there was only [0,0] can be just filtered.
It is compared with traditional static analysis, the present invention adds in structure feature and the instruction features of code to differentiate, extends
Feature space can have higher recall rate.It is compared with dynamic analysis, the present invention has a clear superiority in detection speed.
Above-mentioned specific implementation can by those skilled in the art on the premise of without departing substantially from the principle of the invention and objective with difference
Mode carry out local directed complete set to it, protection scope of the present invention is subject to claims and not by above-mentioned specific implementation institute
Limit, each implementation within its scope is by the constraint of the present invention.
Claims (8)
1. a kind of leakage location based on depth convolutional network and controlling stream graph, which is characterized in that including:Pre-process mould
Block, training module and prediction module, wherein:Preprocessing module reads in the object code in bug code storehouse and generates control stream
Bivector is changed into after figure, feature learning module carries out feature extraction with depth convolutional network and instructed from bivector
Practice, training module includes:Convolutional neural networks unit for feature learning and the BP neural network for similitude comparison
Unit, feature extraction is carried out to bivector with depth convolutional network for convolutional neural networks unit and training obtains identification loophole
The neural network parameter of feature;BP neural network unit carries out similitude with BP neural network to object code and bug code
It analyzes and trains to obtain the neural network parameter for differentiating code similitude, prediction module carries out prediction work, root to object code
The parameter obtained according to training module judges whether comprising loophole, for being determined as that leaky code building loophole is reported;
The bivector, when object code is source code, then preprocessing module is directly translated into assembly code, works as object code
For binary code, then preprocessing module first carries out dis-assembling and operates to obtain assembly code;Control stream is generated according to assembly code
Each node of controlling stream graph is mapped to merge after a vector after figure and is obtained.
2. system according to claim 1, it is characterized in that, the bug code storehouse is to be collected from CVE databases
The bug code arrived.
3. system according to claim 1, it is characterized in that, the preprocessing module includes:It is compilation by code compilation
Assembly code is changed into the controlling stream graph generation unit of controlling stream graph by Open-Source Tools and will controlled by the compiler of code
Each nodes encoding of flow graph becomes vector and is combined into the coding unit of bivector.
4. system according to claim 3, it is characterized in that, when object code is binary code, then pass through manual type
Binary code is converted into assembly code, reconvert obtains the controlling stream graph of assembly code, and controlling stream graph is further analyzed
Obtain the attribute of controlling stream graph;With structure2vec algorithms, each node of controlling stream graph is mapped to a vector,
Vector is combined into bivector.
5. system according to claim 4, it is characterized in that, the structure2vec algorithms are specially:
<mrow>
<mi>F</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>v</mi>
</msub>
<mo>,</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>u</mi>
<mo>&Element;</mo>
<mi>N</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>&mu;</mi>
<mi>u</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>tanh</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<msub>
<mi>x</mi>
<mi>v</mi>
</msub>
<mo>+</mo>
<mi>&sigma;</mi>
<mo>(</mo>
<mrow>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>u</mi>
<mo>&Element;</mo>
<mi>N</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>&mu;</mi>
<mi>u</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
Wherein:σ (l)=P1×ReLU(P2×…ReLU(PnL)), ReLU (x)=max { 0, x }, xvRepresent the set of properties in table one
Into d dimensional vectors, N (v) represents that node v is scheming adjacent node all in G, each node is from μv (0)Start iteration, μv (0)'s
Initial value is filled with as 0 vector, W1It is d × p matrix, p represents the length of vector filling, PiIt is p × p matrix, i=1 ..., n,
<mrow>
<msup>
<msub>
<mi>&mu;</mi>
<mi>v</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mi>F</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>v</mi>
</msub>
<mo>,</mo>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>u</mi>
<mo>&Element;</mo>
<mi>N</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</mrow>
</msub>
<msup>
<msub>
<mi>&mu;</mi>
<mi>u</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>v</mi>
<mo>&Element;</mo>
<mi>V</mi>
<mo>.</mo>
</mrow>
6. system according to claim 1, it is characterized in that, the training module includes:Convolutional neural networks unit, volume
The parameter adjustment unit of the product parameter adjustment unit of neutral net, BP neural network unit and BP neural network unit, wherein:Volume
Product neutral net unit is connected with preprocessing module, receives bivector and carries out feature extraction and train to obtain identification loophole feature
Neural network parameter, the parameter adjustment unit of convolutional neural networks be connected with convolutional neural networks unit and transmit adjustment after
Neural network parameter, BP neural network unit are connected with preprocessing module, receive bivector and carry out similarity analysis and training
Obtain differentiating the neural network parameter of code similitude, parameter adjustment unit and the BP neural network unit of BP neural network unit
Be connected and transmit adjustment after neural network parameter.
7. the system according to claim 1 or 6, it is characterized in that, the convolutional neural networks unit includes:It is sequentially connected
Input layer, the first convolutional layer, the first active coating, the first pond layer, the second convolutional layer, the second active coating, the second pond layer,
One full articulamentum, the second full articulamentum and active coating, i.e., by two layers 3*3 the training of convolution kernel and pond layer and extract feature
Afterwards, using two layers of full articulamentum, loophole is determined whether.
8. the system according to claim 1 or 6, it is characterized in that, the BP neural network unit will pass through
Bivector that structure2vec algorithms convert carries out cosine computings, exports 1 if code approximation, and difference goes out-
1, be specially:
Wherein:G represents controlling stream graph,Represent structure2vec algorithms.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711325630.6A CN108108622B (en) | 2017-12-13 | 2017-12-13 | Vulnerability detection system based on deep convolutional network and control flow graph |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711325630.6A CN108108622B (en) | 2017-12-13 | 2017-12-13 | Vulnerability detection system based on deep convolutional network and control flow graph |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108108622A true CN108108622A (en) | 2018-06-01 |
CN108108622B CN108108622B (en) | 2021-03-16 |
Family
ID=62215679
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711325630.6A Active CN108108622B (en) | 2017-12-13 | 2017-12-13 | Vulnerability detection system based on deep convolutional network and control flow graph |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108108622B (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109214191A (en) * | 2018-09-18 | 2019-01-15 | 北京理工大学 | A method of utilizing deep learning forecasting software security breaches |
CN110147235A (en) * | 2019-03-29 | 2019-08-20 | 中国科学院信息工程研究所 | Semantic comparison method and device between a kind of source code and binary code |
CN110175454A (en) * | 2019-04-19 | 2019-08-27 | 肖银皓 | A kind of intelligent contract safety loophole mining method and system based on artificial intelligence |
CN110598408A (en) * | 2019-08-23 | 2019-12-20 | 华中科技大学 | App clone detection method and system based on function layer coding |
CN110866254A (en) * | 2019-09-29 | 2020-03-06 | 华为终端有限公司 | Vulnerability detection method and electronic equipment |
CN110943981A (en) * | 2019-11-20 | 2020-03-31 | 中国人民解放军战略支援部队信息工程大学 | Cross-architecture vulnerability mining method based on hierarchical learning |
CN110995713A (en) * | 2019-12-06 | 2020-04-10 | 北京理工大学 | Botnet detection system and method based on convolutional neural network |
CN111552969A (en) * | 2020-04-21 | 2020-08-18 | 中国电力科学研究院有限公司 | Embedded terminal software code vulnerability detection method and device based on neural network |
CN111639344A (en) * | 2020-07-31 | 2020-09-08 | 中国人民解放军国防科技大学 | Vulnerability detection method and device based on neural network |
CN113434858A (en) * | 2021-05-25 | 2021-09-24 | 天津大学 | Malicious software family classification method based on disassembly code structure and semantic features |
CN113449303A (en) * | 2021-06-28 | 2021-09-28 | 杭州云象网络技术有限公司 | Intelligent contract vulnerability detection method and system based on teacher-student network model |
CN113468525A (en) * | 2021-05-24 | 2021-10-01 | 中国科学院信息工程研究所 | Similar vulnerability detection method and device for binary program |
CN114020628A (en) * | 2021-11-09 | 2022-02-08 | 中国工商银行股份有限公司 | Code vulnerability detection method and device |
US11960610B2 (en) | 2018-12-03 | 2024-04-16 | British Telecommunications Public Limited Company | Detecting vulnerability change in software systems |
US11973778B2 (en) | 2018-12-03 | 2024-04-30 | British Telecommunications Public Limited Company | Detecting anomalies in computer networks |
US11989289B2 (en) | 2018-12-03 | 2024-05-21 | British Telecommunications Public Limited Company | Remediating software vulnerabilities |
US11989307B2 (en) | 2018-12-03 | 2024-05-21 | British Telecommunications Public Company Limited | Detecting vulnerable software systems |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106648818B (en) * | 2016-12-16 | 2019-06-14 | 华东师范大学 | A kind of object code controlling stream graph generation system |
CN106681917B (en) * | 2016-12-21 | 2019-06-18 | 南京大学 | A kind of front end appraisal procedure neural network based |
-
2017
- 2017-12-13 CN CN201711325630.6A patent/CN108108622B/en active Active
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109214191A (en) * | 2018-09-18 | 2019-01-15 | 北京理工大学 | A method of utilizing deep learning forecasting software security breaches |
US11960610B2 (en) | 2018-12-03 | 2024-04-16 | British Telecommunications Public Limited Company | Detecting vulnerability change in software systems |
US11973778B2 (en) | 2018-12-03 | 2024-04-30 | British Telecommunications Public Limited Company | Detecting anomalies in computer networks |
US11989289B2 (en) | 2018-12-03 | 2024-05-21 | British Telecommunications Public Limited Company | Remediating software vulnerabilities |
US11989307B2 (en) | 2018-12-03 | 2024-05-21 | British Telecommunications Public Company Limited | Detecting vulnerable software systems |
CN110147235A (en) * | 2019-03-29 | 2019-08-20 | 中国科学院信息工程研究所 | Semantic comparison method and device between a kind of source code and binary code |
CN110147235B (en) * | 2019-03-29 | 2021-01-01 | 中国科学院信息工程研究所 | Semantic comparison method and device between source code and binary code |
CN110175454A (en) * | 2019-04-19 | 2019-08-27 | 肖银皓 | A kind of intelligent contract safety loophole mining method and system based on artificial intelligence |
CN110598408A (en) * | 2019-08-23 | 2019-12-20 | 华中科技大学 | App clone detection method and system based on function layer coding |
CN110598408B (en) * | 2019-08-23 | 2021-03-26 | 华中科技大学 | App clone detection method and system based on function layer coding |
CN110866254A (en) * | 2019-09-29 | 2020-03-06 | 华为终端有限公司 | Vulnerability detection method and electronic equipment |
CN110943981A (en) * | 2019-11-20 | 2020-03-31 | 中国人民解放军战略支援部队信息工程大学 | Cross-architecture vulnerability mining method based on hierarchical learning |
CN110995713A (en) * | 2019-12-06 | 2020-04-10 | 北京理工大学 | Botnet detection system and method based on convolutional neural network |
CN111552969A (en) * | 2020-04-21 | 2020-08-18 | 中国电力科学研究院有限公司 | Embedded terminal software code vulnerability detection method and device based on neural network |
CN111639344A (en) * | 2020-07-31 | 2020-09-08 | 中国人民解放军国防科技大学 | Vulnerability detection method and device based on neural network |
CN113468525A (en) * | 2021-05-24 | 2021-10-01 | 中国科学院信息工程研究所 | Similar vulnerability detection method and device for binary program |
CN113434858A (en) * | 2021-05-25 | 2021-09-24 | 天津大学 | Malicious software family classification method based on disassembly code structure and semantic features |
CN113449303A (en) * | 2021-06-28 | 2021-09-28 | 杭州云象网络技术有限公司 | Intelligent contract vulnerability detection method and system based on teacher-student network model |
CN114020628A (en) * | 2021-11-09 | 2022-02-08 | 中国工商银行股份有限公司 | Code vulnerability detection method and device |
Also Published As
Publication number | Publication date |
---|---|
CN108108622B (en) | 2021-03-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108108622A (en) | Leakage location based on depth convolutional network and controlling stream graph | |
CN110136170A (en) | A kind of remote sensing image building change detecting method based on convolutional neural networks | |
CN105095856B (en) | Face identification method is blocked based on mask | |
CN109977790A (en) | A kind of video smoke detection and recognition methods based on transfer learning | |
CN103824309B (en) | Automatic extracting method of urban built-up area border | |
Li et al. | Sewer pipe defect detection via deep learning with local and global feature fusion | |
CN109741328A (en) | A kind of automobile apparent mass detection method based on production confrontation network | |
CN109902018B (en) | Method for acquiring test case of intelligent driving system | |
CN107220603A (en) | Vehicle checking method and device based on deep learning | |
CN104077613A (en) | Crowd density estimation method based on cascaded multilevel convolution neural network | |
CN108108751A (en) | A kind of scene recognition method based on convolution multiple features and depth random forest | |
CN114373128A (en) | Remote sensing monitoring method for four mess of rivers and lakes based on category self-adaptive pseudo label generation | |
CN112884758B (en) | Defect insulator sample generation method and system based on style migration method | |
CN115331012B (en) | Joint generation type image instance segmentation method and system based on zero sample learning | |
CN109543744B (en) | Multi-category deep learning image identification method based on Loongson group and application thereof | |
Hoang et al. | Fast local Laplacian‐based steerable and Sobel filters integrated with adaptive boosting classification tree for automatic recognition of asphalt pavement cracks | |
CN110677437A (en) | User disguised attack detection method and system based on potential space countermeasure clustering | |
Yang et al. | Superpixel image segmentation-based particle size distribution analysis of fragmented rock | |
CN114926767A (en) | Prediction reconstruction video anomaly detection method fused with implicit space autoregression | |
Hashemi et al. | Runtime monitoring for out-of-distribution detection in object detection neural networks | |
CN108596121A (en) | A kind of face critical point detection method based on context and structural modeling | |
CN111415326A (en) | Method and system for detecting abnormal state of railway contact net bolt | |
Hepburn et al. | Enforcing perceptual consistency on generative adversarial networks by using the normalised laplacian pyramid distance | |
CN116630989A (en) | Visual fault detection method and system for intelligent ammeter, electronic equipment and storage medium | |
Ramachandra | Causal inference for climate change events from satellite image time series using computer vision and 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 |