CN110072205B - Hierarchical aggregation method for abnormal data detection of wireless sensor network - Google Patents

Hierarchical aggregation method for abnormal data detection of wireless sensor network Download PDF

Info

Publication number
CN110072205B
CN110072205B CN201910226592.1A CN201910226592A CN110072205B CN 110072205 B CN110072205 B CN 110072205B CN 201910226592 A CN201910226592 A CN 201910226592A CN 110072205 B CN110072205 B CN 110072205B
Authority
CN
China
Prior art keywords
cluster
data
abnormal
detection
node
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910226592.1A
Other languages
Chinese (zh)
Other versions
CN110072205A (en
Inventor
吴蒙
许春杰
杨立君
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN201910226592.1A priority Critical patent/CN110072205B/en
Publication of CN110072205A publication Critical patent/CN110072205A/en
Application granted granted Critical
Publication of CN110072205B publication Critical patent/CN110072205B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/04—Arrangements for maintaining operational condition
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention provides a hierarchical aggregation method for detecting abnormal data of a wireless sensor network in the field of sensor networks, which transmits a cluster abstract representing a bottom node upwards, carries out cluster combination by a father node, and finally carries out iteration until a top gateway node, the gateway node carries out detection of an abnormal cluster, and distinguishes the abnormal cluster and a normal cluster, calculating threshold data for local detection according to the data of the normal cluster and returning the threshold data to other nodes, therefore, global abnormal values are identified at the bottom node, the invention is based on the abnormal data detection of the wireless sensor network, the existing HSCBD algorithm is improved, the cluster merging algorithm is improved, and the technical scheme of the invention can realize high abnormal data detection rate through the artificially constructed Gaussian data set and the data set collected in the real environment, and is greatly reduced in energy loss.

Description

Hierarchical aggregation method for abnormal data detection of wireless sensor network
Technical Field
The invention relates to an abnormal data detection method, in particular to a hierarchical polymerization method for abnormal data detection.
Background
Anomaly detection mechanisms can generally be divided into three categories of approaches, depending on the type of background knowledge of the data available. The first category of methods, which do not require a priori knowledge to find outliers, generally use clustering or unsupervised learning methods and assume that outliers can be well separated from normal. The second is supervised anomaly detection, which requires a data set containing data that has been explicitly labeled as normal or abnormal. Through supervised training, a trained classifier can gain the ability to classify new data as normal or abnormal. This method requires the classifier to be retrained if the features of the normal and abnormal data in the system change. The third method is a semi-supervised anomaly detection method. The training data set used by the classifier is not contaminated by abnormal data, and the abnormality is identified according to the training device. When data is available, the normal model can be learned step by step to accommodate changes in the data distribution, and then the likelihood of the resulting learned model generating a test case for the test is detected.
The existing sensor network anomaly detection technologies at home and abroad at present can be roughly classified into the following types:
density-based methods.
And detecting isolated points of the high-dimensional data based on the subspaces and the correlations.
A kind of support vector machine.
A neural network is replicated.
Outlier detection based on cluster analysis.
Deviation from association rules and frequent itemsets.
The performance of the detection method is greatly influenced by different values of the data set and the parameters, the above methods have no absolute difference between the merits and the demerits, and the detection performance is related to the specific data set and the parameter values.
The following describes an anomaly detection method based on data clustering in recent years.
Data clustering is the process of finding clusters of similar data points, the result of which enables each set of data points to be well separated. Data clustering-based anomaly detection basically clusters data first and then performs an anomaly detection algorithm using the clusters.
Rajasegara proposes an algorithm based on fixed-width clustering, which creates spherical clusters with fixed radii for data. If the clusters are closer in distance to other clusters, then the clusters are labeled as normal data points; otherwise, the scheme will flag it as abnormal. If any new traffic data points fall outside the normal cluster, they are marked as anomalous. However, the accuracy of this scheme is limited by the choice of the cluster radius, and as one of the centralized schemes, this scheme consumes a lot of communication costs.
And S, Rajasegarar provides a distributed abnormal data detection technology based on multilayer aggregation for a sensor network and provides a wireless sensor network model with a layered topology. In the centralized scheme, all the collected data are transmitted to the gateway node by each sensor node, and the gateway node executes an abnormal detection algorithm according to the received data and returns an abnormal cluster abstract to distinguish the abnormal data. In the proposed distributed method, the sensor nodes perform clustering operation on the data received by the sensor nodes to generate local clusters, the clustering algorithm adopts the simplest width-based clustering, and the sensor nodes then send the summary merging information of the clusters to their respective father nodes, so that the energy consumption for data transmission can be reduced. The process is iteratively performed up to the gateway node. And at the gateway node, executing an abnormal cluster detection algorithm according to the received cluster abstract set as data so as to mark the cluster as normal or abnormal. However, this solution also has a problem that the parameters are difficult to determine, and as the iteration progresses, an accumulated error is generated.
Disclosure of Invention
The invention aims to provide a hierarchical aggregation method for abnormal data detection of a wireless sensor network, which combines a K-means + + algorithm, a cluster merging algorithm and a KNN-based abnormal detection algorithm to realize higher detection efficiency and lower energy consumption, remarkably reduces the influence of data communication on the wireless sensor network, and can meet the characteristic that the wireless sensor node is limited in bandwidth, energy and the like.
The invention provides a hierarchical aggregation method for detecting abnormal data of a wireless sensor network, which is characterized in that a cluster abstract representing a bottom node is transmitted upwards, a father node performs cluster combination, and finally, iteration execution is performed until a top gateway node, the gateway node performs detection of an abnormal cluster, the abnormal cluster and a normal cluster are distinguished, threshold data for local abnormal detection is calculated according to data of the normal cluster and returned to other nodes, and therefore a global abnormal value is identified at the bottom node.
As a further limitation of the present invention, the method for acquiring the cluster abstract comprises: clustering the standardized detection data vectors by using a K-means + + algorithm to obtain summary information of each cluster, and for the cluster CiThe abstract is as follows:
Figure BDA0002005372960000031
as a further limitation of the present invention, the specific method for cluster merging includes: the summary information is used for
Figure BDA0002005372960000032
Uploading to a direct father node, executing a cluster merging algorithm at the father node, wherein ciIs the center of the ith cluster, | Ci| is the total amount of data contained in the ith cluster,
Figure BDA0002005372960000033
the maximum distance from the data point in the ith cluster to the center of the cluster;
step.1 summary of clusters received by parent node
Figure BDA0002005372960000034
And
Figure BDA0002005372960000035
if dis (c)i，cj) If the cluster size is less than or equal to w, executing merging operation to obtain a new cluster CkWherein c isk＝(ci+cj)/2，|Ck|＝|Ci|+|Cj|，
Figure BDA0002005372960000036
Otherwise, executing step.2;
step.2 if dis (c)i，cj) If not, the combination operation is not executed, the summary information of the cluster is continuously transmitted upwards to obtain a matrix formed by the distance between the center points of the clusters, the elements in the matrix are compared with w, and if dis (c)i，cj) If w is smaller than w, step.1 is executed, otherwise step.3 is executed;
step.3, continuously uploading the obtained summary information of the clusters layer by layer until a gateway node, executing a cluster merging algorithm by a father node in the middle, executing a K neighbor algorithm by the gateway node, calculating the average value of the weighted distances of K cluster centers which are close to the current considered cluster center, and comparing the average value with a specific threshold value to find an abnormal cluster.
As a further limitation of the present invention, the abnormal cluster detection algorithm specifically includes:
definition of a Cluster CiThe closeness to other clusters is:
Figure BDA0002005372960000037
wherein k is 0.3 x | C | and j is the cluster center label nearest to i;
defining an abnormal cluster as
Figure BDA0002005372960000038
Figure BDA0002005372960000039
The different values of (a) correspond to different confidence degrees;
then, the abstract information of the normal cluster is integrated to calculate the overall abstract information
Figure BDA0002005372960000041
Wherein
Figure BDA0002005372960000042
|Cg|＝∑i|Ci|，
Figure BDA0002005372960000043
i is a normal cluster number, and the gateway node returns the global abstract information
Figure BDA0002005372960000044
To each cluster head, wherein
Figure BDA0002005372960000045
As a threshold for global anomaly detection.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects: the invention is based on the abnormal data detection of the wireless sensor network, improves the existing HSCBD algorithm, provides a K-means + + based clustering algorithm, then improves the clustering algorithm, proves that the technical scheme of the invention can realize high abnormal data detection rate through a manually constructed Gaussian data set and a data set collected in a real environment, and greatly reduces the energy loss.
Drawings
FIG. 1 is a flow chart of the operation of the present invention.
FIG. 2 is a schematic diagram of cluster merging in the present invention.
FIG. 3 is a diagram illustrating global abnormal data detection in a distributed approach according to the present invention.
FIG. 4 is a schematic diagram of an artificial Gaussian dataset distribution.
Fig. 5 is a graph of the effect of detection by the HSCBD protocol (gaussian dataset).
Fig. 6 is a graph of the detection efficiency of the a _ HSCBD scheme (gaussian dataset).
Fig. 7 is a map of a _ HSCBD detection efficiency (k 14, gaussian dataset).
Fig. 8 shows the communication cost saving rate (gaussian data set) for the two schemes HSCBD and a _ HSCBD.
Fig. 9 is an intel berkeley laboratory node distribution plot.
FIG. 10 is a scatter plot of the sampled data.
FIG. 11 shows the HSCBD scheme detection efficiency (IBRL data set).
FIG. 12 shows the detection efficiency (IBRL data set) of the A _ HSCBD scheme.
FIG. 13 shows the communication cost saving rates (IBRL data sets) for both HSCBD and A _ HSCBD schemes.
Detailed Description
The technical scheme of the invention is further explained in detail by combining the embodiment and the attached drawings:
examples
1. Data pre-processing
At time t, each sensor node siMeasuring a feature vector
Figure BDA0002005372960000051
There are m measurements in a time window T within which each sensor node siMeasured data of
Figure BDA0002005372960000052
The timing of the anomaly detection is at the end of each time window in order to derive from the joint data set X Ui＝1..nXiWhere all anomalous data points (data that is inconsistent with most of the other data in the dataset) are found.
Data in a wireless sensor network needs to be preprocessed before use, converted into a form suitable for distance-based clustering, and generally, two data vectors x are calculated1，x2The distance therebetween is represented by the Euclidean distance dis (x)1，x2). However, different dimensional feature vectors of sensor nodes are often in different dynamic ranges, so each data vector v is usedkj(k is time, j represents the j-th dimension of the measurement data) into its standard form: u. ofkj＝(vkj-uvj)/σvjWherein u isvjAnd σvjAre respectively a data feature vector vkjMean and standard deviation of (d). Further each feature vector may be normalized to [0, 1%]，
Figure BDA0002005372960000053
Figure BDA0002005372960000054
Wherein maxujAnd minujMaximum and minimum of each feature vector of the converted data, and original data vector
Figure BDA0002005372960000055
Can finally be expressed as
Figure BDA0002005372960000056
Figure BDA0002005372960000057
2. Cluster division based on K-means + + algorithm
The preprocessed data in the upper ring section is clustered, and the partitioning of a cluster set and the abstract of the cluster can be obtained at a node level by adopting the K-means + + algorithm-based clustering;
the basic description of the algorithm is as follows:
inputting: normalized measurement vector Dataset, number of cluster centers k
And (3) outputting: digest of cluster per node, and cluster C ═ Cr：r＝1...Nc}
Figure BDA0002005372960000058
Figure BDA0002005372960000061
Figure BDA0002005372960000071
3. Iterative cluster merging algorithm
If the Euclidean distance between the centers of two clusters is smaller than a given threshold value w, carrying out a merging operation on the two clusters, as shown in FIG. 2;
the algorithm MergeCluster (C)m) The basic description is as follows:
inputting: cluster C ═ Cr：r＝1...NcH, merging threshold value w
And (3) outputting: merged cluster Cm
Figure BDA0002005372960000072
Figure BDA0002005372960000081
4. KNN-based abnormal cluster detection scheme
Each cluster executes a local anomaly detection algorithm according to the information returned by the gateway node; calculating the corresponding cluster center from the data contained in the cluster groupAnd the distance is compared with the distance obtained from the global summary
Figure BDA0002005372960000082
Comparing, namely distinguishing whether the data is abnormal data;
Figure BDA0002005372960000083
FIG. 3 is a graph obtained by
Figure BDA0002005372960000084
And (3) performing a global abnormal data detection schematic diagram, firstly, upwards transmitting the local abstract of the node, and returning a constant serving as a threshold value to the gateway node for detecting the global abnormal.
5. Performance analysis
The experimental environment is Intel (R) core (TM) i3-2370M CPU @2.40GHz, windows10 operating system, and the whole experiment is realized based on Python 2.7.
5.1 simulation analysis based on artificially constructed Gaussian dataset
The first data set is an artificially constructed gaussian data set, the data set has two eigenvectors, each eigenvector obeys a normal distribution, the mean value of the normal distribution is randomly selected from [0.3, 0.35, 0.45], the standard deviation is 0.03, then noise (anomaly) is introduced into each eigenvector, the abnormal data points are uniformly distributed on [0.5, 1], the gaussian data set consists of an array of 10 sensor nodes, each node contains 100 normal data and 5 abnormal data, and the data needs to be normalized to the [0, 1] interval for use. Since the data set does not contain topology information of the sensor nodes, a single-layer network structure is established.
The gaussian data set is distributed as shown in fig. 4, the lower left corner is normal positive-tai distribution data, the upper right corner is abnormal data points, and the data set is composed of two-dimensional vectors, which are temperature and humidity, respectively.
The Detection Rate (DR) is a ratio of the number of abnormal data detected by the abnormal detection algorithm to the total number of actual abnormal data. The false alarm rate (FPR) is the proportion of the number of normal data which are misjudged as abnormal values by the algorithm to the total amount of the actual normal data.
The detection efficiency of the HSCBD detection scheme proposed by rajasegragar S is shown in fig. 5. It can be seen from the figure that when w ∈ [0.05, 0.018], the algorithm has a higher detection rate and a lower false alarm rate. When w ∈ [0.03, 0.04], abnormal data points may be classified into normal data clusters, thereby affecting detection efficiency and resulting in a higher false alarm rate. It can be seen that the HSCBD protocol is sensitive to the w parameter.
Based on a Gaussian data set, a single variable control method is utilized, the clustering quantity k is taken from [5, 30], a graph 6 is obtained, it is seen from the graph that when k belongs to [13, 15], the improved A _ HSCBD scheme can also achieve high detection efficiency, when k is taken as 14, a detection image like the graph 7 is obtained, it is seen from the graph 7, the scheme can achieve detection efficiency equivalent to HSCBD, and in addition, based on the total data quantity and the quantity of clusters obtained by clustering operation, the communication cost saving rate of HSCBD and A _ HSCBD relative to a centralized scheme can be measured.
It can be seen from fig. 8 that the improved a _ HSCBD method can achieve a higher communication cost saving rate. As can be seen from the figure, the improved scheme further improves the communication cost saving rate compared with the HSCBD scheme.
5.2 simulation analysis based on IBRL data set
The second data set was collected in a real environment from intel berkeley laboratories, ca, usa; the sensor nodes that collected this data were deployed in an indoor environment, and fig. 9 is a distribution plot of the laboratory sensor nodes, with the sensors collecting five measurements at 31 second intervals: temperature, light intensity, temperature relative humidity, voltage (in volts), and topology information; due to the enormous volume of data, only a four hour time window (0: 00:00-03:59:59 on 3/1/2004) was studied. Because the nodes 5 and 15 do not contain any data, the data are ignored, and the topology information in the data is combined with the cluster head election strategy for layering, so that a network layered structure with relatively saved energy can be obtained. Fig. 10 is a scatter diagram distribution of the experimental data, and regions that are apparently isolated are marked as abnormal data for determination of the detection rate. FIG. 11 shows the simulation result of the HSCBD scheme based on IBRL data, and it can be seen from the figure that when w belongs to [0.005, 0.019], the HSCBD scheme realizes a better detection result, the abnormal data detection rate reaches more than 96%, and the false alarm rate FPR is also lower; fig. 12 is a simulation result of the a _ HSCBD scheme based on the IBRL data set, and it can be seen from the image that when w ∈ [0.013, 0.015], the detection efficiency of the a _ HSCBD scheme is further improved compared with the HSBCS scheme, the detection rate reaches above 98%, and the false alarm rate FPR is also low.
Since the main drawback of the centralized scheme is high communication consumption, fig. 13 shows the energy saving rate of the HSCBD scheme and the a _ HSCBD scheme compared to the centralized scheme, and as can be seen from fig. 13, the a _ HSCBD scheme proposed herein has a further improved energy saving rate compared to the HSCBD scheme.
In order to avoid the influence of one experiment on the accuracy of the proposed scheme, the scheme is carried out for 10 times, and finally the comparison of the experimental results of the two implementation schemes shown in the table 1 is obtained; as can be seen from table 1, the improved a _ HSCBD scheme has improved detection efficiency compared with the centralized scheme and the HSCBD scheme.
TABLE 1 comparison of the ten-fold average detection efficiency of HSCBD and A _ HSCBD protocols
Figure BDA0002005372960000101
As can be seen from comprehensive analysis, the improved a _ HSCBD scheme provided herein achieves improvement in detection efficiency and improvement in communication efficiency as compared with the centralized scheme and the HSCBD scheme, but requires a learning process due to determination of the clustering algorithm k, which increases time complexity.
According to the scheme, a layered aggregation wireless sensor network model is established on the basis of data collected by an IBRL laboratory, data received by bottom nodes are clustered firstly, a novel cluster merging method is provided, then summary information of all clusters is transmitted to a gateway node to execute an abnormal cluster recognition algorithm based on KNN, and finally normal cluster information obtained by the gateway node is transmitted to other nodes to perform distributed abnormal data detection.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can understand that the modifications or substitutions within the technical scope of the present invention are included in the scope of the present invention, and therefore, the scope of the present invention should be subject to the protection scope of the claims.

Claims (1)

1. A hierarchical aggregation method for abnormal data detection of a wireless sensor network is characterized in that a cluster abstract representing a bottom node is transmitted upwards, a father node performs cluster combination, and finally iteration is performed until a top gateway node, the gateway node performs abnormal cluster detection, the abnormal cluster and a normal cluster are distinguished, threshold data for local detection is calculated according to data of the normal cluster and returned to other nodes, and therefore a global abnormal value is identified at the bottom node;
the method for acquiring the cluster abstract comprises the following steps: clustering the standardized detection data vectors by using a K-means + + algorithm to obtain summary information of each cluster, and for the cluster CiThe abstract is as follows:
Figure FDA0003482027880000011
wherein c isiIs the center of the ith cluster, | Ci| is the total amount of data contained in the ith cluster,
Figure FDA0003482027880000012
the maximum distance from the data point in the ith cluster to the center of the cluster;
the specific method for cluster merging is as follows: the summary information is used for
Figure FDA0003482027880000013
Uploading to a direct father node, and executing a cluster merging algorithm on the father node;
step.1 summary of clusters received by parent node
Figure FDA0003482027880000014
And
Figure FDA0003482027880000015
if dis (c)i，cj) If the cluster size is less than or equal to w, executing merging operation to obtain a new cluster CkWherein c isk＝(ci+cj)/2，|Ck|＝|Ci|+|Cj|，
Figure FDA0003482027880000016
Otherwise, executing step.2;
step.2 if dis (c)i，cj) If not, the combination operation is not executed, the summary information of the cluster is continuously transmitted upwards to obtain a matrix formed by the distance between the center points of the clusters, the elements in the matrix are compared with w, and if dis (c)i，cj) If w is smaller than w, step.1 is executed, otherwise step.3 is executed;
step.3, continuously uploading the obtained summary information of the clusters layer by layer until a gateway node, executing a cluster merging algorithm by a father node in the middle, executing a k neighbor algorithm by the gateway node, calculating an average value of weighted distances of k cluster centers which are close to a currently considered cluster center, and comparing the average value with a specific threshold value to find an abnormal cluster;
the abnormal cluster detection algorithm specifically comprises the following steps:
definition of a Cluster CiThe closeness to other clusters is:
Figure FDA0003482027880000017
wherein k is 0.3 x | C | and j is the cluster center label nearest to i;
defining an abnormal cluster as
Figure FDA0003482027880000018
Figure FDA0003482027880000019
The different values of (a) correspond to different confidence degrees;
then, the abstract information of the normal cluster is integrated to calculate the overall abstract information
Figure FDA0003482027880000021
Wherein
Figure FDA0003482027880000022
|Cg|＝∑i|Ci|，
Figure FDA0003482027880000023
i is a normal cluster number, and the gateway node returns the global abstract information
Figure FDA0003482027880000024
To each cluster head, wherein
Figure FDA0003482027880000025
As a threshold for global anomaly detection.
CN201910226592.1A 2019-03-25 2019-03-25 Hierarchical aggregation method for abnormal data detection of wireless sensor network Active CN110072205B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910226592.1A CN110072205B (en) 2019-03-25 2019-03-25 Hierarchical aggregation method for abnormal data detection of wireless sensor network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910226592.1A CN110072205B (en) 2019-03-25 2019-03-25 Hierarchical aggregation method for abnormal data detection of wireless sensor network

Publications (2)

Publication Number Publication Date
CN110072205A CN110072205A (en) 2019-07-30
CN110072205B true CN110072205B (en) 2022-03-15

Family

ID=67366564

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910226592.1A Active CN110072205B (en) 2019-03-25 2019-03-25 Hierarchical aggregation method for abnormal data detection of wireless sensor network

Country Status (1)

Country Link
CN (1) CN110072205B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111654874B (en) * 2020-06-03 2023-02-24 枣庄学院 Wireless sensor network anomaly detection method
CN114362994B (en) * 2021-11-26 2023-01-06 北京交通大学 Multilayer different-granularity intelligent aggregation railway system operation behavior safety risk identification method
CN114827211B (en) * 2022-05-13 2023-12-29 浙江启扬智能科技有限公司 Abnormal monitoring area detection method driven by node data of Internet of things

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102045708A (en) * 2011-01-25 2011-05-04 河海大学常州校区 Energy prediction-based wireless sensor network intrusion detection method
CN106604267A (en) * 2017-02-21 2017-04-26 重庆邮电大学 Dynamic self-adapting wireless sensor network invasion detection intelligence algorithm
WO2018063701A1 (en) * 2016-10-01 2018-04-05 Intel Corporation Unsupervised machine learning ensemble for anomaly detection

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102045708A (en) * 2011-01-25 2011-05-04 河海大学常州校区 Energy prediction-based wireless sensor network intrusion detection method
WO2018063701A1 (en) * 2016-10-01 2018-04-05 Intel Corporation Unsupervised machine learning ensemble for anomaly detection
CN106604267A (en) * 2017-02-21 2017-04-26 重庆邮电大学 Dynamic self-adapting wireless sensor network invasion detection intelligence algorithm

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Intrusion detection framework of cluster-based wireless sensor network;Hichem Sedjelmaci;《2012 IEEE Symposium on Computers and Communications(ISCC)》;20120726;全文 *
基于K-means聚类的WSN异常数据检测算法;费欢;《计算机工程》;20150731;全文 *
自适应多趟聚类在检测无线传感器网络安全中的应用;滕少华;《传感器与微系统》;20150228;全文 *

Also Published As

Publication number Publication date
CN110072205A (en) 2019-07-30

Similar Documents

Publication Publication Date Title
CN110110802B (en) Airborne laser point cloud classification method based on high-order conditional random field
CN108023876B (en) Intrusion detection method and intrusion detection system based on sustainability ensemble learning
CN108093406B (en) Wireless sensor network intrusion detection method based on ensemble learning
Messaoud et al. A survey on machine learning in Internet of Things: Algorithms, strategies, and applications
Gao et al. Ship-handling behavior pattern recognition using AIS sub-trajectory clustering analysis based on the T-SNE and spectral clustering algorithms
CN110072205B (en) Hierarchical aggregation method for abnormal data detection of wireless sensor network
O'Reilly et al. Anomaly detection in wireless sensor networks in a non-stationary environment
CN110309861B (en) Multi-modal human activity recognition method based on generation of confrontation network
Amini et al. On density-based data streams clustering algorithms: A survey
CN109612513B (en) Online anomaly detection method for large-scale high-dimensional sensor data
CN108985380B (en) Point switch fault identification method based on cluster integration
Wang et al. An isolation-based distributed outlier detection framework using nearest neighbor ensembles for wireless sensor networks
Hu et al. Fuzzy clustering data given on the ordinal scale based on membership and likelihood functions sharing
CN111460441A (en) Network intrusion detection method based on batch normalization convolutional neural network
CN115100709B (en) Feature separation image face recognition and age estimation method
CN115096627B (en) Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment
CN110826642A (en) Unsupervised anomaly detection method for sensor data
Wang et al. Measurement data classification optimization based on a novel evolutionary kernel clustering algorithm for multi-target tracking
Chen et al. Comogcn: Coherent motion aware trajectory prediction with graph representation
Guan et al. Application of a novel PNN evaluation algorithm to a greenhouse monitoring system
CN114879628A (en) Multi-mode industrial process fault diagnosis method based on local maximum mean difference resistance
CN116206208B (en) Forestry plant diseases and insect pests rapid analysis system based on artificial intelligence
CN112307894A (en) Pedestrian age identification method based on wrinkle features and posture features in community monitoring scene
Abdulla et al. kEFCM: kNN-based dynamic evolving fuzzy clustering method
CN112529025A (en) Data processing method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant