CN110061961B - Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine - Google Patents
Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine Download PDFInfo
- Publication number
- CN110061961B CN110061961B CN201910163503.3A CN201910163503A CN110061961B CN 110061961 B CN110061961 B CN 110061961B CN 201910163503 A CN201910163503 A CN 201910163503A CN 110061961 B CN110061961 B CN 110061961B
- Authority
- CN
- China
- Prior art keywords
- node
- network
- connection
- tracking
- candidate
- 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.)
- Expired - Fee Related
Links
- 238000010276 construction Methods 0.000 title claims abstract description 23
- 238000005457 optimization Methods 0.000 claims abstract description 68
- 238000004364 calculation method Methods 0.000 claims abstract description 58
- 238000001514 detection method Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 37
- 230000005540 biological transmission Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012423 maintenance Methods 0.000 claims description 3
- 230000002068 genetic effect Effects 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000035772 mutation Effects 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 108090000623 proteins and genes Proteins 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000002028 premature Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 241000234282 Allium Species 0.000 description 1
- 235000002732 Allium cepa var. cepa Nutrition 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 238000009401 outcrossing Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0893—Assignment of logical groups to network elements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/10—Network architectures or network communication protocols for network security for controlling access to devices or network resources
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/20—Network architectures or network communication protocols for network security for managing network security; network security policies in general
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses an anti-tracking network topology intelligent construction method and system based on a limited Boltzmann machine. The anti-tracking network node deploys a connection optimization algorithm, network state parameters of corresponding nodes are obtained and calculated, the connection state of the corresponding nodes is judged according to the calculation result, and the local topology is updated and optimized, so that the anti-tracking network can dynamically evolve according to different network states, and intelligent construction and optimization of the network topology are realized. The invention improves the robustness of the anti-tracking network, realizes network load balance, improves the anti-tracking performance to a certain extent, and can increase the difficulty of network topology detection of enemies and the difficulty of network flow tracking.
Description
Technical Field
The invention belongs to the field of network space security, and relates to an anti-tracking network topology intelligent construction method and system based on a limited Boltzmann machine, which realizes the intelligent construction and optimization of an anti-tracking network topology, effectively improves the robustness of a network and the network load balancing capability, and improves the anti-tracking capability of the network to a certain extent.
Background
With the development and popularization of the internet, the network has penetrated the aspects of daily life of people, however, the network threat facing the privacy of network users is rampant day by day. The malicious behaviors of network hackers and lawless persons such as stealing user privacy, selling user information, stealing user property and the like are increasingly severe, and even network information security events facing important organizations such as large-scale enterprises and government offices appear. Meanwhile, the development of a network tracking and tracing technology reduces the threshold that lawless persons infringe personal privacy by utilizing a network to a great extent. Therefore, it is very urgent to research how to resist the invasion of personal privacy by using a network tracing and tracing technology to ensure personal information security of network users.
The anti-tracking network is a network system capable of effectively resisting network tracking and tracing, and the network system can ensure that the network information is difficult to detect and track in the transmission process, so that the identity of a network user is hidden, and the privacy of the user is protected. Compared with an anonymous network, the anti-tracking network not only provides a user identity anonymous service similar to the anonymous network, but also emphasizes the anti-tracking characteristic in the information transmission process. Currently, the focus of research on anti-tracking networks is mainly on constructing a network system by taking a P2P network as a platform and taking a user to join the network voluntarily. Taking anonymous network Tor as an example, Tor network is a rerouting anonymous communication system based on the Onion protocol, which provides anonymous service for users and data transmission. However, the Tor network is extremely vulnerable to tracking and attack of the directory server, and has weak resistance to malicious nodes and malicious behaviors. There are also related researches focusing on implementing anti-tracing of message passing based on P2P network, and due to the wide distribution of P2P network nodes, the anti-tracing performance of the network can be improved by the jump of message passing between different nodes. However, with the introduction of the graph tracing method, the greatest challenge is faced by the tracking resistance of the P2P-based anti-tracing network, mainly due to frequent information transfer between P2P network nodes and "mutual connection" between the network nodes, which is a dilemma that the network is easily traced. This is mainly due to the fact that the existing anti-tracking network topology is generally a static update structure and has no capability of self-adapting to the network state and self-adjusting optimization. The adjustment and optimization of the existing anti-tracking network topology structure also mainly adopts a centralized command control mode, and a controller monitors the network topology and issues a command to update the network topology, so that the exposure probability of network nodes is increased invisibly, and the soft rib is used for defending network tracking and tracing.
The anti-tracking network may be classified into a centralized topology, a P2P type topology, and a hybrid topology according to its topology. The representative work of the construction, updating, optimization and the like of the anti-tracking network topology is as follows.
Most of the anti-tracking networks of the centralized topology structure are realized by adopting an agent mechanism or a VPN technology. And forwarding the message through a proxy server or a VPN server through a proxy mechanism or a VPN channel so as to hide a network system of a message sender. However, the anti-tracing network of the mechanism can only provide certain anonymous service of the sender, the existing network tracing and tracing technology can effectively trace both parties of message delivery, and the anti-tracing capability is weak.
In the aspect of anti-tracking network research of a P2P type topology, some researchers propose a path selection and optimization algorithm for the Tor network, and select a new path with bandwidth optimized allocation according to the network state of the Tor, so as to realize load balancing of each node of the Tor network. However, the method needs to obtain the overall state of the Tor network, and is not beneficial to protecting the privacy and the safety of the user. Researchers have proposed a method for constructing an anti-tracking network by using cross-domain characteristics, which improves the anti-tracking performance of the network by limiting adjacent network nodes to be in different autonomous domains. However, considering the randomness of adding nodes into the anti-tracking network, the situation that the autonomous domain where the network nodes are located is relatively centralized may occur, and it is very difficult to construct a strict cross-domain anti-tracking network. However, with the development of network tracing technology, the anti-tracing performance of the network with the existing P2P topology is greatly reduced by means of P2P topology detection, malicious node penetration and the like. A researcher can conveniently identify direct contact among the nodes of the P2P network, and the tracking of the P2P network is realized by using the PageRank algorithm, and the method can effectively track other network nodes under the condition that certain nodes are known. Researchers put forward that honeypot nodes are implanted into a Tor network, so that traffic in the Tor network is captured, through cooperation of traffic analysis and the honeypot nodes, the identity of an anonymous user can be effectively identified, and a message propagation path of the anonymous user can be tracked.
In the aspect of research on a hybrid topology structure, researchers propose an anti-tracking network based on a Mixnet protocol, wherein the anti-tracking network adopts a hybrid topology structure, strong anonymity is guaranteed by using a fixed-level hybrid node, and the efficiency of network transmission is improved by reducing real-time encryption delay and calculation overhead. However, the fixed-level hybrid nodes are easy to cause targeted traffic monitoring and analysis, and after probe traffic is added, the propagation path of user traffic is easy to track. Some researchers propose an anonymous system, namely, SocialMix, which combines a social network and a Mixnet protocol, and aggregate messages through a trusted social platform, and the identities of two message delivery parties are hidden by using the Mix protocol. However, social platform delivery of messages is blocked, thereby reducing the efficiency of message dissemination. For the anti-tracking network, related researchers propose to utilize a social network analysis method to realize the detection and tracking of the network by analyzing social attributes and incidence relations among nodes. Meanwhile, attacking the Mix node is also an effective way to destroy the anti-tracking network.
Disclosure of Invention
Aiming at the defects of the existing method, the invention discloses an anti-tracking network topology intelligent construction method and system based on a limited Boltzmann machine, so as to improve the robustness of an anti-tracking network and enhance the capability of the anti-tracking network in resisting various network tracking and tracing technologies.
The invention discloses an anti-tracking network topology intelligent construction method based on a limited Boltzmann machine, which comprises the following steps:
(1) the anti-tracking network nodes are all provided with a connection optimization algorithm;
(2) the anti-tracking network node selects a candidate node from the candidate node set of the node, and acquires the network topology parameter and the message forwarding parameter of the candidate node;
(3) taking the two types of parameters obtained in the step (2) as the input of a connection optimization algorithm of the current node, and calculating the connection optimization algorithm based on a limited Boltzmann machine to obtain a corresponding calculation result;
(4) judging the connection state of the current node and the candidate node according to the calculation result obtained in the step (3); if the connection state is disconnected, the current node does not establish connection with the candidate node, and the established connection is disconnected; if the connection state is connection establishment, the current node sends a connection request to the candidate node; at this time, the candidate node also calculates the network topology parameter and the message forwarding parameter of the current node and obtains the corresponding connection state, and when the connection state calculated by the candidate node is also the connection establishment, the two nodes finally establish the connection; otherwise, the candidate node does not establish connection with the current node;
(5) and (5) repeating the steps (2), (3) and (4) until the current node reaches a stable state.
The process of the present invention is further illustrated below. First, some definitions are given, to which the present disclosure relates:
(1) and (6) candidate nodes. The candidate node is a node set which can be selected by the node u in the anti-tracking network and is calculated by using a connection optimization algorithm. C represents a candidate node set, N represents a neighbor node set of the node u, S represents a neighbor node set of the neighbor node of the node u, and C ═ N, S }.
(2) The node is in steady state. For the current node u and the calculable node v thereof, if v belongs to N, the judgment result of calculation through a connection optimization algorithm is connection establishment; and if v belongs to S, the judgment result calculated by the connection optimization algorithm is disconnection. That is, the calculation result of the candidate node with connection is to establish connection, and the calculation result of the candidate node without connection is to disconnect, and the connection state of u is in a stable state.
Whether the anti-tracking network node is connected or disconnected with the corresponding node is judged by a connection optimization algorithm deployed by the node. Firstly, a current node acquires network state parameters of candidate nodes; calculating whether the current node needs to be connected or disconnected according to the network state parameters by using a connection optimization algorithm deployed by the current node; and updating the connection state of the current node according to the calculation result.
A connection optimization algorithm deployed by nodes in the anti-tracking network is an important guarantee for realizing intelligent construction of a topological structure of the network. As shown in fig. 1, the connection optimization algorithm performs calculation according to the node network state parameters, thereby determining the optimal connection state of the node in the current network state, so as to guide the node to update its local topology, and optimize its connection policy. Thus, the selection of input parameters of the restricted boltzmann machine determines the trends and preferences of the nodes for autonomous connection. To ensure the self-maintenance, self-optimization and self-healing capabilities of the anti-tracking network, the criteria for selecting the input parameters herein follow the following principles:
(1) and (4) robustness. The anti-tracking network is an open and distributed platform, nodes are allowed to freely join and leave, and the change of the nodes can change the original topological structure of the network, so that potential threats such as overloaded nodes, cut points and the like are brought to the topological structure of the network. Therefore, the conditions of overload nodes and cut points are considered in the selection of the parameters, the existence of the nodes can be effectively avoided through a connection optimization algorithm, and the robustness of the anti-tracking network is ensured.
(2) And (4) balancing network load. Whether the load of the anti-tracking network is balanced or not is an important index influencing the information transmission performance of the network. Under the condition of no need of intervention of a control party, the anti-tracking network can dynamically adjust the network topology according to the network message propagation characteristics to realize network load balance, so that the network service quality is greatly improved. The parameters are therefore chosen taking into account the nature of the message propagation in the network.
(3) And (4) destruction resistance. Survivability is an important prerequisite to ensure that an anti-tracking network is available. The distribution and connection conditions of the nodes in the network are adjusted through a connection optimization algorithm, and key nodes and cut points in the network are eliminated, so that the network is in a better connection state at all times. Meanwhile, the network can still recover and optimize itself after removing part of nodes in the network, even key nodes.
In view of the above principles, the parameters select the following node information as inputs to the restricted boltzmann machine.
(1) The network topology parameter is a network parameter determined by a network topology structure, and specifically includes the following:
d _ Domain: an autonomous domain in which the candidate node is located;
d _ Degree: the degree of the candidate node;
c _ Domain: the autonomous domain where the current node is located;
c _ Degree: the degree of the current node;
c _ CandidateAmount: the number of candidate nodes of the current node;
c _ NeighborsAvgDegree: average value of degree of neighbor node of current node;
SharingNode: the number of neighbor nodes shared by the current node and the candidate node. If the two nodes are neighbor nodes, the value is 0;
DifferenceOfDegrid is the difference between the degrees of the current node and the candidate node.
(2) The message forwarding parameter, which is a parameter obtained according to the statistics of message forwarding in the network, may change with the situation that the node in the network forwards the message, specifically as follows:
d _ MessageAvg: the average value of the message forwarding amount of the candidate nodes in unit time;
c _ MessageAvg: the average value of the message forwarding amount in unit time of the current node;
c _ neighborsvgmessage: the average value of the forwarding message amount of the neighbor node of the current node in unit time;
DifferenceOfMessage is the difference between the average forwarded message amount of the current node and the candidate node in unit time.
Due to different dimensions of different input parameters, in order to ensure that the evaluation criteria of the input parameters are consistent, normalization processing needs to be performed on the input parameters of the limited Boltzmann machine. The input parameters of the restricted Boltzmann machine are normalized and then restricted in an interval [0, 1 ].
D _ Domain and C _ Domain respectively represent autonomous domains of the candidate node and the current node, all autonomous domains are numbered from 1 to n, and n represents the total number of the autonomous domains. The normalization function of the two input parameters is as follows, where i takes on values from 1 to n:
the data for the D _ Degree, C _ Degree, DifferenceOfDegree, C _ NeighborsAvgDegree, C _ CandidateAmount, SharingNode input parameters may be 0 or any positive integer. DifferenceOfMessage, C _ NeighborsAvgMessage, D _ MessageAvg, and C _ MessageAvg are statistics of the node forwarding messages per unit time. Every time when the node receives the message from other nodes and forwards the message, the information is accumulated to 1, and finally, the average value in unit time is taken as the value of the parameter. Thus, the above parameters may be normalized by the following function, where x represents the above parameters:
the connection optimization algorithm is an algorithm which is realized based on a limited Boltzmann machine and optimizes the connection state of the nodes according to the network state of the nodes. The network nodes calculate the connection states of the related nodes by using a connection optimization algorithm according to the corresponding network state parameters, judge whether to establish or disconnect the connection with the network nodes according to the connection states, and update the connection states, so that the intelligent construction and optimization of the anti-tracking network topology structure are realized.
The output result of the connection optimization algorithm is 0, which indicates that the connection is disconnected, at the moment, the connection is directly disconnected for the nodes which are already connected, and the nodes which are not connected with the nodes which are not connected; an output of 1 indicates that a connection is established, the connection is maintained for nodes that have established a connection, and for nodes that have not established a connection, a connection is requested to be established with them, and then the other party performs a calculation using a connection optimization algorithm, and both parties can finally establish a connection only if the output of the connection optimization algorithm of the other party is also 1. Therefore, the connection optimization algorithm for each node in the anti-tracking network has two calculation states, namely an active calculation state and a passive calculation state. The active calculation state refers to the state that the node actively calculates with the candidate node to update the network connection state; the passive calculation state refers to listening for the connection request and performing calculation to determine whether to establish connection therewith. The specific algorithms in the two calculation states are described below.
(1) forall bi epsilon C// cycle calculation of candidate node of current node u
(2) parameter (getparameter) (bi); v/obtaining network status parameter of node bi
(3) connection ═ rbm (parameter); // calculating the connection status with the node bi by a restricted Boltzmann computer
(4)If connection＝＝0,then
(4.1)If bi∈N,then
(4.2) BreakLink (bi); if the candidate node is a neighbor node, disconnecting it
(4.3) update (u); v/update the network status parameter of node u and the candidate node set
(5)If connection＝＝1,then
(5.1) parameter _ u ═ getparameter (u); v/obtaining network status parameter of current node u
(5.2) RequestLink (bi, parameter _ u); v/send the network status parameter of the current node u to the node bi,
requesting connection establishment with a candidate node
(1) if GetLinkRequest (), the/current node u receives connection requests of other nodes;
(1.1) parameter _ b ═ GetParameter (); v/obtaining network status parameters of requesting connecting node b
(1.2) connection _ b ═ RBM (parameter _ b); // connection to node b by a constrained boltzmann computer
Connection condition
(1.3)If connection_b＝＝0,then
(1.4) RefuseLink (b); if the calculation result is 0, refusing to establish connection with the request node
(1.5)If connection_b＝＝1,then
(1.6) connection (u); if the calculation result is 1, the connection is established with the connection requesting node
(1.7) update (u); v/update the network status parameter of node u and the candidate node set
The intelligent construction and optimization of the anti-tracking network topology structure are realized by the dynamic evolution of the nodes according to the network topology state and the network message forwarding condition. The addition or the exit of the node can cause the network topology to change, the corresponding node can realize the dynamic change of the connection according to the new network state, the connection strategy is adjusted to be optimal, the network self-maintenance topological structure is realized, and the dynamic evolution is called as passive evolution; and the nodes dynamically adjust the network topology structure according to the change of the network message forwarding condition so as to realize the continuous change of the network message transmission path and the load balance of the network, and the dynamic evolution is called as active evolution.
The passive evolution realizes the self-optimization of the anti-tracking network topology structure, improves the robustness of the network, and enables the network to be automatically recovered to a corresponding stable state after the network topology changes. Fig. 2 is a schematic diagram of passive evolution, and illustrates a process of passive evolution and an influence on network robustness by taking a node u as an example. The specific process is as follows:
(1) FIG. 2(a) shows the initial state of the current network;
(2) after the node 1 exits the network, the network topology becomes as shown in fig. 2 (b). For the node u, the network state is also changed, the degree is changed from 3 to 2, and the neighbor nodes and the candidate nodes are also changed.
(3) And after the network state is changed, the node u realizes the update of the connection state through a connection optimization algorithm. As shown in fig. 2(c), node u establishes a connection with its candidate node 4, 7.
(4) After the node u realizes the update of the connection state through the connection optimization algorithm each time, both the network topology and the network state of the node u are changed. Therefore, the node u continues to calculate the candidate node through the connection optimization algorithm until the node u reaches a steady state. As shown in fig. 2(d), after the node establishes connection with the candidate nodes 5 and 6, a stable state is reached, and the passive evolution is finished.
The passive evolution is that after the node detects that the network topology changes, the corresponding network state parameter and the candidate node set are updated, and the candidate node is calculated through the new network state parameter, so that the current node can still automatically recover to a stable state after the network topology changes, and the robustness of the network is improved.
Active evolution is an important mode for ensuring the load balance of the anti-tracking network, avoiding the generation of key nodes and improving the anti-tracking performance. Fig. 3 is a schematic diagram of active evolution, which illustrates active evolution and its influence on network load by taking node u as an example. The specific process is as follows:
(1) fig. 3(a) shows the topology of node u in a steady state. However, the nodes 2, 3, 4, and 5 all forward the message to the node u from the node 1, and although the network topology parameters will not change when the topology state is stable, the message forwarding parameters D _ message avg, C _ message avg, DifferenceOfMessage, and C _ neighborsvgmessage will all change.
(2) With the change of the message forwarding parameters, when the node u performs connection optimization calculation on the candidate node, the output result also changes. Fig. 3(b) shows that the result of the connection optimization calculation performed by the node u on the node 4 is 1. Node u therefore requests a connection to be established with node 4.
(3) The node 4 also performs connection optimization calculation on the node u, and if the output result is still 1, the two sides establish connection. Similarly, when the node u and the node 1 perform calculation, because the message forwarding amount of the node 1 is too large, and the output result of the connection optimization algorithm is 0, the node u and the node 1 are disconnected, as shown in fig. 3 (c). Fig. 3(b) shows that node u is connected to node 4 and disconnected from node 1.
(4) Fig. 3(d) shows a topology structure after the node u actively evolves, which avoids excessive message transmission by the node 1.
In the passive evolution, when the topological structure is changed, in order to ensure the stability of the network system structure, the nodes dynamically change the connection state according to the network topological parameters, and the stability and the robustness of the network structure are maintained. The active evolution is to dynamically change the connection status of the network nodes according to the message forwarding condition and the message forwarding parameters according to the network message forwarding condition, and change the message forwarding rule, thereby avoiding the generation of overload nodes and also avoiding the detection of the topology structure and the message forwarding path by the related network tracking technology.
The passive evolution promotes the network to enter a stable state, and the higher robustness of the network is kept; and the passive evolution changes the network topology according to the message forwarding condition, thereby forming a cycle and enabling the whole network to be in a self-dynamic evolution process. However, the passive evolution and the active evolution are not performed independently, and the evolution process is determined by the input parameters of the connection optimization algorithm. Therefore, the connection optimization algorithm can calculate according to the input parameters and the corresponding weights, not only considering the network topology structure, but also considering the message forwarding condition, and further performing comprehensive judgment, so as to realize the balance between the network topology stability and the network load balance in the dynamic evolution process.
Correspondingly to the method, the invention also provides an anti-tracking network topology intelligent construction system based on the limited Boltzmann machine, which comprises a plurality of anti-tracking network nodes, wherein each anti-tracking network node is provided with a parameter selection module and a connection optimization algorithm module;
the parameter selection module selects a candidate node from a candidate node set of the current node, acquires network topology parameters and message forwarding parameters of the candidate node, and takes the two types of parameters as the input of a connection optimization algorithm module of the current node;
the connection optimization algorithm module carries out calculation based on a limited Boltzmann machine according to the two input parameters and obtains a corresponding calculation result;
judging the connection state of the current node and the candidate node according to the calculation result obtained by the connection optimization algorithm module; if the connection state is disconnected, the current node does not establish connection with the candidate node, and the established connection is disconnected; if the connection state is connection establishment, the current node sends a connection request to the candidate node; at this time, the candidate node also calculates the network topology parameter and the message forwarding parameter of the current node and obtains the corresponding connection state, and when the connection state calculated by the candidate node is also the connection establishment, the two nodes finally establish the connection; otherwise, the candidate node does not establish connection with the current node; and repeating the operation until the current node reaches a stable state.
The intelligent construction and optimization method of the network topology structure improves the tracking resistance of the network to a certain extent. Due to the dynamic evolution characteristic of the network, the topology of each node can be dynamically changed according to the network state, so that the communication topology among the network nodes, the propagation path of the message and the like can be changed along with the change, and the difficulty of network tracing by an adversary is increased. The tracking resistance is mainly shown in the following aspects:
(1) increasing the difficulty of the enemy in detecting the network topology. The network tracking method based on the graph can effectively analyze the topological structure of communication between the network nodes by monitoring the node flow information and the communication relation, thereby carrying out tracking and tracing in a targeted manner. However, for a network with a dynamically evolved topology structure, it is difficult to infer the topology relationship of the nodes after the dynamic evolution by using this method, thereby reducing the network tracing capability.
(2) Increasing the difficulty of tracking network traffic. The traffic information of the enemy monitoring network node can easily track the propagation path of the information. However, the dynamically evolved topology structure changes the path of information transmission, thereby increasing the difficulty of enemy tracking and improving the tracking resistance of the network.
Drawings
FIG. 1 is a frame diagram of an anti-tracking network topology intelligent construction method based on a limited Boltzmann machine, wherein RBM is an abbreviation of the limited Boltzmann machine.
Fig. 2 is a schematic diagram of a passive evolution process of an anti-tracking network.
Fig. 3 is a schematic diagram of an active evolution process of an anti-tracking network.
FIG. 4 is a flow chart of the intelligent construction of an anti-tracking network topology.
FIG. 5 is a schematic diagram of a restricted Boltzmann machine.
Detailed Description
The present invention will be described in detail with reference to fig. 4. For convenience of presentation, the limited boltzmann machine is abbreviated as RBM, and the connection optimization algorithm is abbreviated as CJA algorithm.
The invention provides an anti-tracking network topology intelligent construction method based on a restricted Boltzmann machine, which is specifically described in detail in three aspects of parameter acquisition, weight optimization and topology intelligent construction.
The specific steps of parameter acquisition are as follows:
(1) and the nodes in the anti-tracking network are normalized by the following formula according to the number of the autonomous Domain where the nodes are located to generate the network topology parameter C _ Domain of the nodes. Wherein n in the formula represents the number of autonomous domains, and the autonomous domains are numbered according to 1 to n;
(2) the node counts the Degree of the node, and the normalization is carried out by the following formula to generate a network topology parameter C _ Degrid of the node;
(3) the node counts the message forwarding amount in unit time, and normalization is carried out by using the formula shown in the step (2) to generate a message forwarding parameter C _ MessageAvg of the node;
(4) and (3) requesting the neighbor nodes of the node, acquiring a node list with the span of 2, forming a candidate node set C of the node together with the neighbor nodes of the node, counting the number of elements in the candidate node set C, normalizing by using the formula shown in the step (2), and generating a network topology parameter C _ CandidateAmout. And generating a network topology parameter SharingNode through a node list which is returned by the neighbor node and has the span of 2. If the current node and the candidate node are neighbor nodes, setting SharingNode as 0; otherwise, the value is set to the number of neighbor nodes shared between the current node and the candidate node. For example, the current node has 3 neighbor nodes, and the list of nodes with span 2 received from its neighbor is: {1, 2}, {2, 3}, and {2, 3, 4 }. Then the SharingNode values of the current node and the nodes 1, 2, 3, and 4 are respectively: 1. 3, 2 and 1. SharingNode can also be normalized by the formula shown in step (2).
(5) And (3) the node acquires the network topology parameter C _ Degrid of all the neighbor nodes, calculates the network topology parameter C _ NeighborsAvgDegrid of the current node by the following formula, and normalizes by using the formula in the step (2). Wherein n represents the number of the candidate node set elements of the current node, and Degree (i) represents the value of the network topology parameter C _ Degree of the ith node in the candidate node set;
(6) and (3) the node acquires the message forwarding parameters C _ MessageAvg of all the neighbor nodes, calculates the message forwarding parameter C _ NeighborsAvgMessage of the current node by using the following formula, and normalizes by using the formula in the step (2). Wherein n represents the number of candidate node set elements of the current node, and message (i) represents the value of the message forwarding parameter C _ MessageAvg of the ith node in the candidate node set;
(7) selecting one candidate node v in a candidate node set of any node u in the countermeasure tracking network, and acquiring network topology parameters C _ Domain, C _ Degree, C _ MessageAvg and a message forwarding parameter C _ MessageAvg of the candidate node v, wherein the parameters are respectively used as input parameters D _ Domain, D _ Degree, D _ MessageAvg and D _ Messaavageg of a connection optimization algorithm of the node u;
(8) and (3) calculating the network topology parameter DifferenceOfDegreee according to the network topology parameters C _ Degreee of the node u and the candidate node v, wherein the calculation formula is as follows, and the normalization is performed by using the formula shown in the step (2). Wherein, the Degree (i) represents the value of the network topology parameter C _ Degree of the acquisition node i;
f(x)＝degree(u)-degree(v).
(9) and (3) calculating a message forwarding parameter DifferenceOfMessage according to the network topology parameters C _ MessageAvg of the node u and the candidate node v, wherein the calculation formula is as follows, and the normalization is performed by using the formula shown in the step (2). Wherein, the method represents the value of the message forwarding parameter C _ MessageAvg of the acquisition node i;
f(x)＝message(u)-message(v).
(10) and (3) integrating the network topology parameters and the message forwarding parameters obtained in the steps (1) to (9), and calculating the parameters as the input of the limited Boltzmann machine.
For the RBM weight optimization problem, a genetic algorithm is adopted to optimize the RBM weight, and the specific steps are as follows:
(1) adaptation of genetic algorithmsThe fitness function is shown as follows, wherein α, β and gamma are respectively influence coefficients of each calculation formula, and d are respectivelyavgRespectively representing the degree of the current node and the average degree, num, of the candidate nodesdRepresents the number of nodes in the same autonomous domain as the current node in the candidate nodes, mavgRespectively representing the message forwarding times of the current node and the average message forwarding times of the candidate nodes in unit time;
(2) in the process of weight optimization of the genetic algorithm, crossover and variation are important processes for promoting gradual optimization of the algorithm and finally solving the approximate optimal solution of the problem. The crossing rate and the variation rate are important parameters for controlling the genetic algorithm to carry out crossing and variation operations, and directly influence the running performance of the algorithm. In the weight optimization process, in order to store individuals with high fitness and ensure the diversity of genetic algorithm populations in the operation process, the following formulas are adopted to calculate the cross rate and the variation rate:
wherein λ is1，λ2Is a constant with a value ranging between (0, 1). If the intersection rate is calculated, f in the above formula represents the higher fitness of the two individuals needing the intersection operation. If the mutation rate is calculated, f in the above formula represents the fitness of the number of mutation operations. f. ofmaxRepresenting the maximum fitness in the entire population. f. ofavgRepresenting the average fitness across the population. Therefore, when the fitness is higher, the probability of carrying out cross and mutation operation is low, and premature convergence is effectively avoided; on the contrary, when the fitness is too low, the probability of performing intersection and mutation operation is high, so that the search range of the optimal solution is expanded. When the fitness approaches to the population average value, the intersection and mutation operation is carried out with a relatively stable probability. Is obtainingAfter obtaining the corresponding crossing rate and the variation rate, crossing and varying by a roulette algorithm;
(3) for the crossover operation, two individuals performing the crossover operation are respectively crossed by different gene position pairs, one tends to be crossed evenly between the two father individuals, and the other tends to be crossed to the father individual with high fitness. As used herein, giAnd gjRespectively representing the values of the corresponding gene positions of two individuals subjected to the cross operation. siAnd sjRespectively representing the numerical values of the corresponding gene positions of the two children generated after the crossing of the two father individuals. For children s with balanced crossover between two parentsiThe cross formula is as follows, wherein pcRepresents the crossover rate:
si＝pcgi+(1-pc)gj
defining two intermediate variables delta for the sub-individuals which tend to have high fitness and carry out cross operation on the parent individuals1And Δ2The specific calculation formula is as follows:
based on the two intermediate variables Δ1And Δ2New individuals sjThe calculation formula of (a) is as follows:
sj＝pcΔ1+(1-pc)Δ2
wherein xminAnd xmaxRespectively, represent the minimum and maximum values of the weights for the RBMs used herein to allow variant runs to be performed within reasonable bounds. Therefore, when the fitness of the individual is close to the optimal fitness, the variation interval is small, and conversely, when the fitness of the individual is too low, the variation interval is large. Thus, on one hand, the influence of variant operation on individuals with high fitness is effectively avoided, and on the other hand, the variant operation on individuals with high fitness is effectively avoidedThe aspect can also enlarge the search range through mutation operation, and avoid premature convergence.
(4) The genetic algorithm is used for searching an optimal approximate solution, when the genetic algorithm tends to converge, CJA can quickly converge by using the weight parameters acquired by the genetic algorithm, and the neural network training efficiency is improved. Therefore, when the genetic algorithm is in the solving process, the difference value of the fitness of the individuals obtained by the last two operation results is smaller than the specified threshold value, the genetic algorithm is considered to be converged, and the neural network is reused for the next training of the weight parameters.
In the aspect of intelligent topology construction, a CJA algorithm is utilized to calculate the network topology parameters and the message forwarding parameters of the nodes, so that the connection state of the nodes is judged, and the anti-tracking network can intelligently construct and optimize the network topology according to the network state. The method comprises the following specific steps:
(1) the structure of the RBM neural network adopted in the invention is shown in figure 5, wherein the visible layer of the first RBM comprises 12 nodes, and the hidden layer comprises 6 nodes. The visible layer of the second RBM contains 6 nodes and the hidden layer contains 1 node. The first RBM outputs the result as an input to a visible layer in the second layer RBM. The second RBM takes a hidden layer only containing one node as a classification layer for calculating and judging the node connection state;
(2) the calculation formula for calculating the node values of the hidden layer from the visible layer is as follows. Wherein v isiValue, w, representing the ith node of the visible layerijRepresenting the weight between the ith node of the visible layer and the jth node of the hidden layer, biRepresents the bias of the ith node of the visible layer, n represents the number of nodes in the visible layer, hjA calculated value representing a jth node of the hidden layer;
(3) in the RBM shown in fig. 5, a tanh function is used as an activation function of each node of the hidden layer, where x represents a node value of each node of the hidden layer:
(4) in fig. 5, the output result of the second RBM at the classification level is a binary result, i.e. output 1 indicates establishment of connection and output 0 indicates rejection of connection. The following threshold function is therefore used to achieve the binary output of the final classification layer:
(5) in summary, the two-level RBM neural network in fig. 5 can achieve the binary result of whether to establish a connection from the input parameter to the output by the following function. Wherein b, c are the bias of the first RBM and second RBM visible layer, respectively, wijRepresents the weight corresponding to the ith node of the visible layer and the jth node of the hidden layer in the first RBM, viInput parameter, m, representing a visible layerjRepresenting the weight between the second RBM visible layer node and the hidden layer node.
(6) When the CJA algorithm output result is 0, the current node is not connected with the corresponding candidate node, and if the connection is established, the connection is disconnected;
(7) when the CJA algorithm output result is 1, the current node sends a connection request to the corresponding candidate node and attaches the network topology parameter and the message forwarding parameter related to the current node, the corresponding candidate node carries out calculation again according to the received parameter, and judgment on whether connection is established or not is made according to the calculation result. If the CJA algorithm calculation result of the corresponding candidate node is 0, connection with the node requesting connection is not established, and on the contrary, if the calculation result is 1, the node requesting connection is responded and connection is established;
(8) after the current node changes its connection state according to the calculation result of the CJA algorithm, it needs to update its candidate node set, network topology parameters and message forwarding parameters according to the new connection state. Then, reselecting the nodes in the candidate node set for calculation to judge whether the connection state needs to be continuously updated or optimized;
(9) after the current node is calculated with all the candidate nodes in the candidate node set, the topology of the current node is not changed any more, and then the current node enters a stable state. When the candidate node with the span of 2 is calculated, the output result of the CJA algorithm is 0;
(10) after the current node enters a stable state, the current node can continuously monitor the network topology state and count the message forwarding parameters. If the network state of the current node changes, if the neighbor node of the current node exits the network or a new node establishes connection with the current node, the candidate node set and the network topology parameters of the current node need to be updated again, and the candidate node is selected to be calculated so as to judge whether the connection state needs to be updated to adapt to the new network state after the network topology state changes; if the message forwarding parameter of the current node is changed greatly in unit time, if the current node receives a large amount of messages, the message forwarding parameter of the current node needs to be updated, and a corresponding candidate node is selected according to the new message forwarding parameter for calculation so as to judge whether the connection state needs to be dynamically changed or not, and change the local topology of the current node, so that the network load is balanced.
The invention provides an anti-tracking network topology intelligent construction system based on a limited Boltzmann machine, which comprises a plurality of anti-tracking network nodes, wherein each anti-tracking network node is provided with a parameter selection module and a connection optimization algorithm module;
the parameter selection module selects a candidate node from a candidate node set of the current node, acquires network topology parameters and message forwarding parameters of the candidate node, and takes the two types of parameters as the input of a connection optimization algorithm module of the current node;
the connection optimization algorithm module carries out calculation based on a limited Boltzmann machine according to the two input parameters and obtains a corresponding calculation result;
judging the connection state of the current node and the candidate node according to the calculation result obtained by the connection optimization algorithm module; if the connection state is disconnected, the current node does not establish connection with the candidate node, and the established connection is disconnected; if the connection state is connection establishment, the current node sends a connection request to the candidate node; at this time, the candidate node also calculates the network topology parameter and the message forwarding parameter of the current node and obtains the corresponding connection state, and when the connection state calculated by the candidate node is also the connection establishment, the two nodes finally establish the connection; otherwise, the candidate node does not establish connection with the current node; and repeating the operation until the current node reaches a stable state.
The above embodiments are only intended to illustrate the technical solution of the present invention and not to limit the same, and a person skilled in the art can modify the technical solution of the present invention or substitute the same without departing from the principle and scope of the present invention, and the scope of the present invention should be determined by the claims.
Claims (8)
1. An anti-tracking network topology intelligent construction method based on a limited Boltzmann machine comprises the following steps:
1) the anti-tracking network nodes are all provided with a connection optimization algorithm;
2) the anti-tracking network node selects a candidate node from the candidate node set of the node, and acquires the network topology parameter and the message forwarding parameter of the candidate node; the network topology parameters include: the method comprises the following steps that an autonomous Domain D _ Domain where a candidate node is located, a Degree D _ Degree of the candidate node, an autonomous Domain C _ Domain where a current node is located, a Degree C _ Degree of the current node, the number of the candidate nodes of the current node C _ CandidataAmount, an average value C _ NeighborsAvgDegree of the degrees of neighbor nodes of the current node, the number SharingNode of the neighbor nodes shared by the current node and the candidate node, and a difference DifferenceOfDegree of the degrees of the current node and the candidate node; the message forwarding parameters include: the average value D _ MessageAvg of the message volume forwarded in unit time of the candidate node, the average value C _ MessageAvg of the message volume forwarded in unit time of the current node, the average value C _ NeighborsAvgMessage of the message volume forwarded in unit time of the neighbor node of the current node, and the difference DifferenceOfMessage of the message volume forwarded in unit time of the current node and the candidate node;
3) taking the two types of parameters obtained in the step 2) as the input of a connection optimization algorithm of the current node, and calculating the connection optimization algorithm based on a limited Boltzmann machine to obtain a corresponding calculation result; the calculation process comprises the following steps: the current node performs connection judgment on the calculation results of the two types of parameters obtained in the step 2) according to the limited Boltzmann machine, and when the calculation result is 0, the current node is disconnected or not connected with the corresponding candidate node; when the calculation result is 1, requesting to establish connection with the corresponding candidate node; judging the calculation results of the two types of parameters obtained in the step 2) by using a limited Boltzmann machine corresponding to the candidate node, if the calculation result of the candidate node is 1, establishing connection with the corresponding node, and if not, rejecting the connection request of the corresponding node;
4) judging the connection state of the current node and the candidate node according to the calculation result obtained in the step 3); if the connection state is disconnected, the current node does not establish connection with the candidate node, and the established connection is disconnected; if the connection state is connection establishment, the current node sends a connection request to the candidate node; at this time, the candidate node also calculates the network topology parameter and the message forwarding parameter of the current node and obtains the corresponding connection state, and when the connection state calculated by the candidate node is also the connection establishment, the two nodes finally establish the connection; otherwise, the candidate node does not establish connection with the current node;
5) and repeating the steps 2), 3) and 4) until the current node reaches a stable state.
2. The method according to claim 1, characterized in that the connection optimization algorithms deployed by the anti-tracing network nodes each have two computation states, namely an active computation state and a passive computation state; the active calculation state refers to the state that the node actively calculates with the candidate node to update the network connection state; the passive calculation state refers to listening for the connection request and performing calculation to determine whether to establish connection therewith.
3. The method of claim 1, wherein the parameters D _ Domain, C _ Domain are normalized using the following formula, wherein i represents the number of autonomous domains in which the node is located, the autonomous domains are numbered from 1 to n, and n represents the total number of autonomous domains:
and the parameters C _ Degree, D _ Degree, C _ CandidataAmount, C _ NeighborsAvgDegree, SharingNode, DifferenceOfDegree, D _ MessageAvg, C _ NeighborsAvgMessage, DifferenceOfMessage are normalized by the following formulas, wherein x represents the parameter value for performing the normalization operation:
4. the method of claim 1, wherein intelligent construction and optimization of the network topology is achieved by calculating network topology parameters and message forwarding parameters of the nodes, that is, the anti-tracking network can dynamically evolve according to different network states to intelligently optimize the topology structure to adapt to different network states.
5. The method of claim 4, characterized in that after the network topology changes due to the addition or the exit of the anti-tracking network node, the corresponding node realizes the dynamic change of the connection according to the new network state, adjusts the connection strategy to reach the optimum, and realizes the network autonomous maintenance topology structure, and this kind of dynamic evolution is called passive evolution; the self-optimization of the anti-tracking network topology structure is realized through passive evolution, the robustness of the network is improved, and the anti-tracking network can be automatically recovered to a corresponding stable state after the network topology changes.
6. The method of claim 5, wherein the nodes dynamically adjust the network topology according to the change of the network message forwarding condition to realize the continuous change of the network message transmission path and the load balance of the network, and such dynamic evolution is called active evolution; the anti-tracking network load balance is ensured through active evolution, key nodes are avoided, and the anti-tracking performance is improved.
7. The method as claimed in claim 5 or 6, wherein the passive evolution and the active evolution are not performed independently, and the evolution process is determined by input parameters of a connection optimization algorithm, the connection optimization algorithm performs calculation according to the input parameters and corresponding weights, and the network topology structure and the message forwarding condition are considered, so as to perform comprehensive judgment, and realize balance between network topology stability and network load balance in the dynamic evolution process of the anti-tracking network; the difficulty of enemy to network topology detection and network flow tracking is increased through the dynamic evolution characteristic of the anti-tracking network, and the anti-tracking capability of the network is improved.
8. An anti-tracking network topology intelligent construction system based on a restricted Boltzmann machine and adopting the method of any one of claims 1-7, characterized by comprising a plurality of anti-tracking network nodes, wherein each anti-tracking network node is provided with a parameter selection module and a connection optimization algorithm module; the system performs the following operations:
A. the parameter selection module selects a candidate node from a candidate node set of the current node, acquires network topology parameters and message forwarding parameters of the candidate node, and takes the two types of parameters as the input of a connection optimization algorithm module of the current node;
B. the connection optimization algorithm module carries out calculation based on a limited Boltzmann machine according to the two input parameters and obtains a corresponding calculation result;
C. judging the connection state of the current node and the candidate node according to the calculation result obtained by the connection optimization algorithm module; if the connection state is disconnected, the current node does not establish connection with the candidate node, and the established connection is disconnected; if the connection state is connection establishment, the current node sends a connection request to the candidate node; at this time, the candidate node also calculates the network topology parameter and the message forwarding parameter of the current node and obtains the corresponding connection state, and when the connection state calculated by the candidate node is also the connection establishment, the two nodes finally establish the connection; otherwise, the candidate node does not establish connection with the current node;
the above operation A, B, C is repeated until the current node reaches a steady state.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910163503.3A CN110061961B (en) | 2019-03-05 | 2019-03-05 | Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910163503.3A CN110061961B (en) | 2019-03-05 | 2019-03-05 | Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110061961A CN110061961A (en) | 2019-07-26 |
| CN110061961B true CN110061961B (en) | 2020-08-25 |
Family
ID=67316583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910163503.3A Expired - Fee Related CN110061961B (en) | 2019-03-05 | 2019-03-05 | Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110061961B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111314336B (en) * | 2020-02-11 | 2021-03-23 | 中国科学院信息工程研究所 | Dynamic transmission path construction method and system for anti-tracking network |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103281743A (en) * | 2013-06-18 | 2013-09-04 | 清华大学 | Source position tracing resistant privacy protection method based on route extension |
| CN106507426A (en) * | 2016-11-23 | 2017-03-15 | 西安电子科技大学 | A kind of new energy consumption sensor network topological evolution method in a balanced way |
| CN106600041A (en) * | 2016-12-02 | 2017-04-26 | 天津大学 | Restricted Boltzmann machine-based photovoltaic power generation short-term power probability prediction method |
| CN107528824A (en) * | 2017-07-03 | 2017-12-29 | 中山大学 | A kind of depth belief network intrusion detection method based on two-dimensionses rarefaction |
| CN108108475A (en) * | 2018-01-03 | 2018-06-01 | 华南理工大学 | A kind of Time Series Forecasting Methods that Boltzmann machine is limited based on depth |
| CN108270608A (en) * | 2017-01-04 | 2018-07-10 | 中国科学院声学研究所 | A kind of foundation of link prediction model and link prediction method |
| CN108848068A (en) * | 2018-05-29 | 2018-11-20 | 上海海事大学 | Based on deepness belief network-Support Vector data description APT attack detection method |
| CN109034391A (en) * | 2018-08-17 | 2018-12-18 | 王玲 | The multi-source heterogeneous information RBM network integration framework and fusion method of automatic Pilot |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2554981A (en) * | 2016-08-16 | 2018-04-18 | British Telecomm | Configuration parameters for virtual machines |
-
2019
- 2019-03-05 CN CN201910163503.3A patent/CN110061961B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103281743A (en) * | 2013-06-18 | 2013-09-04 | 清华大学 | Source position tracing resistant privacy protection method based on route extension |
| CN106507426A (en) * | 2016-11-23 | 2017-03-15 | 西安电子科技大学 | A kind of new energy consumption sensor network topological evolution method in a balanced way |
| CN106600041A (en) * | 2016-12-02 | 2017-04-26 | 天津大学 | Restricted Boltzmann machine-based photovoltaic power generation short-term power probability prediction method |
| CN108270608A (en) * | 2017-01-04 | 2018-07-10 | 中国科学院声学研究所 | A kind of foundation of link prediction model and link prediction method |
| CN107528824A (en) * | 2017-07-03 | 2017-12-29 | 中山大学 | A kind of depth belief network intrusion detection method based on two-dimensionses rarefaction |
| CN108108475A (en) * | 2018-01-03 | 2018-06-01 | 华南理工大学 | A kind of Time Series Forecasting Methods that Boltzmann machine is limited based on depth |
| CN108848068A (en) * | 2018-05-29 | 2018-11-20 | 上海海事大学 | Based on deepness belief network-Support Vector data description APT attack detection method |
| CN109034391A (en) * | 2018-08-17 | 2018-12-18 | 王玲 | The multi-source heterogeneous information RBM network integration framework and fusion method of automatic Pilot |
Non-Patent Citations (2)
| Title |
|---|
| Training Restricted Blotzmann Machine Using Gradient Fixing Based Algorithm;Li Fei,Gao Xiaoguang;《电子学报》;20181231;全文 * |
| 基于遗传算法的RBM优化设计;刘凯，张立民，孙永威;《微电子学与计算机》;20150702;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110061961A (en) | 2019-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Buchegger et al. | The effect of rumor spreading in reputation systems for mobile ad-hoc networks | |
| Manavi | Defense mechanisms against distributed denial of service attacks: A survey | |
| Farahani | Black hole attack detection using K‐nearest neighbor algorithm and reputation calculation in mobile ad hoc networks | |
| Hu et al. | Trust based secure and energy efficient routing protocol for wireless sensor networks | |
| Seyfollahi et al. | A review of intrusion detection systems in RPL routing protocol based on machine learning for internet of things applications | |
| Nogueira et al. | A security management architecture for supporting routing services on WANETs | |
| Zhu et al. | Deployment and exploitation of deceptive honeybots in social networks | |
| Sangeetha et al. | Efficient intrusion detection of malicious node using Bayesian hybrid detection in MANET | |
| Gali et al. | Multi-Context Trust Aware Routing For Internet of Things. | |
| CN110061961B (en) | Anti-tracking network topology intelligent construction method and system based on limited Boltzmann machine | |
| CN113364810B (en) | Link flooding attack detection and defense system and method | |
| Deb et al. | TIDS: trust-based intrusion detection system for wireless ad-hoc networks | |
| CN114553458A (en) | Method for establishing and dynamically maintaining credible group in power Internet of things environment | |
| Thanuja et al. | Unethical network attack detection and prevention using fuzzy based decision system in mobile ad-hoc networks | |
| Ahmed et al. | Trust management for IoT security: taxonomy and future research directions | |
| Alajeely et al. | Reputation based malicious node detection in OppNets | |
| Maheshwar et al. | Black hole effect analysis and prevention through IDS in MANET environment | |
| Tetarave et al. | A routing table poisoning model for peer-to-peer (P2P) botnets | |
| Burange et al. | Trust based secured Routing System for low power networks | |
| Nowak et al. | Cognitive routing for improvement of IoT security | |
| Bharathisindhu et al. | An improved model based on genetic algorithm for detecting intrusion in mobile ad hoc network | |
| Zhou et al. | Defense mechanism of interest flooding attack based on deep reinforcement learning | |
| Backes et al. | Anonymity and censorship resistance in unstructured overlay networks | |
| Ghandar et al. | Performance analysis of observation based cooperation enforcement in ad hoc networks | |
| Yin | Application of Active Learning Algorithm in Mobile Ad Hoc Network Intrusion Detection |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200825 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |