CN111083106A - Detection method and detection system for attack robot in multi-robot network - Google Patents
Detection method and detection system for attack robot in multi-robot network Download PDFInfo
- Publication number
- CN111083106A CN111083106A CN201911099762.0A CN201911099762A CN111083106A CN 111083106 A CN111083106 A CN 111083106A CN 201911099762 A CN201911099762 A CN 201911099762A CN 111083106 A CN111083106 A CN 111083106A
- Authority
- CN
- China
- Prior art keywords
- robot
- adjacent
- robots
- period
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- 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
- H04L63/08—Network architectures or network communication protocols for network security for authentication of entities
- H04L63/0807—Network architectures or network communication protocols for network security for authentication of entities using tickets, e.g. Kerberos
-
- 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
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
- H04L63/1416—Event detection, e.g. attack signature detection
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a detection method and a detection system for an attack robot in a multi-robot network, wherein the method comprises the following steps: the method comprises the steps that wireless signals from adjacent robots are reflected by a plurality of labels placed on a target robot, the target robot firstly constructs a unique signal characteristic matrix for each adjacent robot based on extracted rebound wireless signals, then the distance between every two characteristic matrices is effectively measured, then a matrix containing the similarity probability between every two adjacent robots is generated based on the measured distance, finally, the adjacent robots at the same position or on the same path have similar signal characteristics, and legal robots and malicious attackers in the adjacent robots are judged based on the signal space characteristics according to the similarity matrices so as to detect false robots counterfeited by Sybil attackers in the current network. The invention can realize accurate Sybil attack detection even on a lightweight robot platform only provided with a single antenna, and has simple structure and strong practicability.
Description
Technical Field
The invention belongs to the field of robot network Sybil attack detection, and particularly relates to a method and a system for detecting an attacking robot in a multi-robot network.
Background
The rapid development of wireless technology has greatly facilitated effective collaboration between miniaturized, agile robotic teams, enabling their widespread application in many real-world tasks, including city surveillance, performing robotic formation, and search and rescue efforts in disaster scenarios. While the openness of the wireless medium allows the multi-robot community to perform efficient and fast collaborative tasks, it also simultaneously poses a threat to multi-robot networks from malicious network attacks. In a multi-robot network, one particularly harmful form of attack is a witch attack. In particular, witch attackers gain unfair advantage in the overall network by forging many false identities (e.g., IDs), thereby easily breaking the basic trust assumption in robot collaboration. For example, by forging a set of false IDs, a witch robot can easily obtain valuable communication bandwidth resources from other robots. In addition, in the path planning task, unreal spatial position information can be easily transmitted in the whole robot network, and further, the problems of serious collision and congestion among other robots are caused.
However, due to the temporal, dynamic, and miniaturized nature of the robotic platform, detection of witch attacks in multi-robot networks remains a challenging problem. The traditional key encryption method needs to previously assume that all network nodes are completely trusted, but the assumption is difficult to implement in a robot network with temporary characteristics. In addition, recent research results identify spatial uniqueness of each network node by using wireless signal physical layer information such as Received Signal Strength (RSSI), Channel State Information (CSI), and angle of arrival (AoA), however, the RSSI or CSI based methods not only require cooperation among multiple antennas or receiving ends, but also require all nodes to be in a stationary or semi-stationary state, and although AoA features can be used for detection under dynamic channels, these methods require a large array of antennas to obtain these fine-grained signal features. Therefore, none of the existing methods is suitable for miniaturized robots with limited load capacity and hardware resources.
Disclosure of Invention
The invention provides a detection method and a detection system for attacking robots in a multi-robot network, which are used for solving the technical problem that the existing miniaturized robot with limited load capacity and hardware resources is difficult to apply due to the requirement of more complex hardware facilities and implementation conditions in the detection of Sybil attacks in the existing multi-robot network.
The technical scheme for solving the technical problems is as follows: a method for detecting an attack robot in a multi-robot network comprises the following steps:
a target robot in the multi-robot network sequentially and directly receives a wireless signal of each adjacent robot through a single antenna of the target robot according to a preset adjacent robot time sequence, and meanwhile, in a time period when the single antenna directly receives the wireless signal of any adjacent robot, a plurality of labels arranged on the target robot sequentially receive the wireless signal of the adjacent robot according to a preset working time sequence and rebound to the single antenna;
denoising a rebound wireless signal of each label received by a single antenna in any working period, solving an amplitude mean value of a rebound point, constructing a multipath amplitude mean value vector corresponding to each adjacent robot in each received period based on the amplitude mean values corresponding to the labels, and constructing a signal feature matrix of the adjacent robot by all the multipath amplitude mean value vectors of the adjacent robot;
and calculating and determining the adjacent robots belonging to the attacking robot through the similarity probability based on the distance between every two signal feature matrixes.
The invention has the beneficial effects that: the invention sets a plurality of labels on a target robot and only needs a single antenna, the single antenna of the target robot directly receives the wireless signal of a certain adjacent robot, and simultaneously, the wireless signal of the adjacent robot rebounded by the labels is indirectly received by the labels, after a preset time period, the target robot can extract the rebounded wireless signal of each label in each working time period by denoising according to the different frequency bands of the directly received wireless signal and the indirectly received wireless signal from all the received wireless signals, the invention actively changes the multipath propagation of the wireless signal by using the labels, can ensure that a small robot only with the single antenna can obtain the wireless signal characteristic with information source space information, avoids a large and heavy antenna array system, and the method is further based on the rebounded wireless signals of all the labels received in all the time periods of the wireless signal of each adjacent robot, and constructing a unique signal characteristic matrix of the adjacent robot, and further judging whether a legal robot and a Sybil attacker in the adjacent robot are the attacking robot according to the similar probability based on the distance between every two signal characteristic matrices corresponding to every two adjacent robots. The invention authenticates the real identity information of the adjacent robot based on a plurality of labels, can ensure that the target robot with a single antenna has the capability of effectively resisting network attack, has simple structure, and can be applied to a miniaturized robot platform only with a single antenna.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, each of the tags is a passive backscatter tag.
The invention has the further beneficial effects that: the passive backscatter tag is light and convenient, is suitable for a miniaturized robot, and ensures excellent performances such as flexibility of the robot.
Further, the method for extracting the bounce wireless signal of any working time period of each tag from the bounce wireless signal received by the single antenna within the preset time period comprises the following steps:
carrying out moving average on the rebounded wireless signals received by the single antenna within a preset time period;
and according to the adjacent robot time sequence, performing correlation calculation on the rebound wireless signals corresponding to each adjacent robot after the sliding average and the preset working time sequence, and extracting the rebound wireless signals of any working time period of each label received by the single antenna.
The invention has the further beneficial effects that: the moving average operation can effectively remove noise, and each working period of each tag can be reliably demodulated from the rebound wireless signal based on the preset working time sequence of a plurality of tags.
Further, the determining, by the similarity probability, the neighboring robot belonging to the attacking robot is specifically:
constructing a distance matrix based on the distance between every two signal characteristic matrixes;
calculating the similarity probability between every two adjacent robots by adopting a logistic regression method based on the distance matrix to obtain a similarity probability matrix;
and based on the similarity probability matrix, judging the similarity, if the similarity probability between two adjacent robots is greater than a threshold value, using the two adjacent robots as attacking robots, and otherwise, determining the adjacent robots belonging to the attacking robots.
The invention has the further beneficial effects that: by using logistic regression and similarity judgment, the identity authentication with low computation complexity, high speed and high accuracy can be realized.
Further, denoising the rebound wireless signal of each tag received by the single antenna in any working period and solving an amplitude average value of a rebound point, specifically:
taking a first average value of the amplitudes of all rebound sampling points and a second average value of the amplitudes of all non-rebound sampling points in the rebound wireless signal of any working period of each label, and subtracting the second average value from the first average value to obtain the amplitude average value of the rebound point of the label in the working period;
and forming a multipath amplitude mean value vector of the adjacent robot in each received period by the amplitude mean value of all the labels of each adjacent robot in each received period.
The invention has the further beneficial effects that: by averaging the amplitudes of the rebounding points and the non-rebounding points and differencing the average values, the influence caused by noise can be effectively removed.
Further, before constructing the signal feature matrix of the neighboring robot, the method further includes:
and regularizing each multipath amplitude mean value vector L-2 to obtain a new multipath amplitude mean value vector.
The invention has the further beneficial effects that: and performing L-2 regularization on each multipath amplitude mean value vector to resist the change of an attacker to the transmission power of the multipath amplitude mean value vector and improve the detection precision.
Further, the preset adjacent robot time sequence is divided into a plurality of time sequence cycles, and each time sequence cycle comprises the sequence of each adjacent robot and a received time period;
calculating the distance between every two signal feature matrices, specifically:
and respectively calculating the cosine distance between the multipath amplitude mean value vectors in the same time sequence period between every two signal feature matrixes.
The invention has the further beneficial effects that: similarity information between two adjacent robots within a period of time is obtained, so that finer-grained characteristics are provided, and attacker detection accuracy is improved.
Further, the multipath amplitude mean value vector distance between the ith adjacent robot and the jth adjacent robot in the ith time sequence period And d isij≠dji,filIs the multipath amplitude mean value vector, f, of the ith adjacent robot in the ith time sequence periodjlThe multipath amplitude mean vector of the jth adjacent robot in the ith time sequence period is shown, and L is the number of the time sequence periods.
The invention has the further beneficial effects that: based on the modified cosine distance, the angular difference between two feature vectors from different locations will increase, while the angular difference between feature vectors from similar locations will remain small, allowing attack detection with high accuracy.
The invention also provides a system for detecting the attacking robot in the multi-robot network, wherein the system is a target robot which is provided with an antenna, a plurality of labels and a control processor in the multi-robot network, the control processor adopts any one of the above methods for detecting the attacking robot in the multi-robot network, and the single antenna and the labels are controlled to collect rebounding wireless signals and detect the rebounding wireless signals to obtain the attacking robot of the current robot.
The invention has the beneficial effects that: the invention introduces an attack robot detection system in a multi-robot network, namely a robot, which is provided with an antenna, a plurality of labels and a control processor, particularly, the control processor can control one single antenna and a plurality of labels to collect and rebound wireless signals of adjacent robots and detect attackers, and the detection is realized by executing the attack robot detection method in the multi-robot network.
The invention also provides a storage medium, wherein the storage medium is stored with instructions, and when the instructions are read by a computer, the computer is enabled to execute the detection method of the attack robot in any one of the multi-robot networks.
Drawings
Fig. 1 is a flowchart of a method for detecting an attacking robot in a multi-robot network according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a wireless signal of a legal neighboring robot and two illegal neighboring robots, which is received by a single antenna after the wireless signal is bounced by two backscattering tags according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a construction of a multipath amplitude mean value based on a backscattering label according to an embodiment of the present invention;
fig. 4 is a flowchart of estimating similarity probability between two adjacent robots based on logistic regression according to an embodiment of the present invention;
fig. 5 is a schematic view of a detection flow of an attacking robot in a multi-robot network based on multipath signal changes according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
Example one
An attack robot detection method 100 in a multi-robot network, as shown in fig. 1, includes:
120, denoising the rebound wireless signals of each label received by a single antenna in any working period, solving the amplitude mean value of the rebound point, constructing a multipath amplitude mean value vector corresponding to each adjacent robot in each received period based on the amplitude mean values corresponding to a plurality of labels, and constructing a signal feature matrix of the adjacent robot by all multipath amplitude mean value vectors of the adjacent robot;
and step 130, calculating and determining the adjacent robots belonging to the attacking robot through the similarity probability based on the distance between every two signal feature matrixes.
It should be noted that, a plurality of labels that set up on the target robot work according to predetermineeing the work chronogenesis in proper order, specifically do: according to the self-preset bounce time sequence, when a wireless signal of a certain adjacent robot is received and bounces to a single antenna, the tag bounces and does not bounce alternately in the working process, and the alternate process is the bounce time sequence and represents the bounce and non-bounce time length. The preset working time sequence in step 110 is nested in the received time period of the wireless signal of each adjacent robot, and the bounce time sequence of each tag is nested in the working time period of the tag. Each neighboring robot is a robot other than the target robot in the determined robot network.
The "multipath" in the "multipath amplitude mean value vector" represents different multiple bounce paths corresponding to multiple tags, and since the multiple tags sequentially operate according to a preset operating time sequence, and each tag operates when receiving signals from a single antenna, the signals are signals of different paths received sequentially, so that the vector formed by the multipath amplitude mean values is called a multipath amplitude mean value vector.
In addition, in step 120, a signal feature matrix of each neighboring robot is constructed by all multipath amplitude mean vectors of the neighboring robot in a preset research period, which is determined according to practice.
The invention uses the labels to actively change the multipath propagation of the wireless signals, so that a small robot only with a single antenna can obtain the wireless signal characteristics with information source space information, a large and heavy antenna array system is avoided, and the method is further based on the rebounding wireless signals of all the labels in all the time periods when the wireless signals of each adjacent robot are received, the method is based on a plurality of labels, real identity information of the adjacent robot is authenticated, and the target robot with the single antenna can have the capability of resisting network attack. Therefore, the present embodiment is a multi-robot network attack detection system based on multipath signal variation, and the attack detection system can be applied to a miniaturized robot platform with only a single antenna.
Preferably, each tag is a passive backscatter tag.
The passive backscatter tag is light and convenient, is suitable for a miniaturized robot, and ensures excellent performances such as flexibility of the robot. In addition, the backscattering tag can provide fine-grained wireless signal physical layer information, and the information contains information of relative positions of a source and a sink.
Backscatter tags propagate their own bit information by constantly absorbing and bouncing the radio signals around them. When the target robot and the adjacent robot normally communicate, the absorption and the rebound of the surrounding signals by the backscattering labels cause the multipath signals between the receiving antenna and the adjacent robot to change correspondingly, and the multipath changes are highly correlated with the spatial positions of the receiving antenna and the adjacent robot. Therefore, the backscattering label can be used for actively changing the propagation of multipath signals, extracting the multipath changes related to the position information, and then carrying out light-weight authentication on the spatial uniqueness of the source based on the change characteristics.
For example, as shown in fig. 2, two backscatter tags are placed on a single-antenna target robot and receive wireless signals from a legal close robot and two dummy robots (both the legal close robot and the two dummy robots are close robots of the target robot), and the amplitudes of signals bounced by the two tags are extracted, and experiments show that the bounced signals of the two dummy robots are almost the same. And for a legal robot and a certain false robot, the rebounding signals of the legal robot and the certain false robot are greatly different, and through the observation, the attack detection in the multi-robot network can be carried out by extracting the physical characteristics in the rebounding signals of the backscattering labels.
Preferably, before performing processing such as denoising on the bounce wireless signal of each tag in any working period, the bounce wireless signal of each tag in any working period needs to be extracted from the bounce wireless signal received by the single antenna in a preset period, and specifically, the extraction method includes:
carrying out moving average on the rebounded wireless signals received by the single antenna within a preset time period;
and according to the adjacent robot time sequence, performing correlation calculation on the rebound wireless signals corresponding to each adjacent robot after the sliding average and the preset working time sequence, and extracting the rebound wireless signals of any working time period of each label received by the single antenna.
As shown in fig. 3, the moving average operation can effectively remove noise to reliably demodulate bit information from the received signal (see clear bit information), and then correlate the moving average bouncing radio signal s (n) with the known bit information i (n) of all tags as follows:
where c (n) is the correlation result, M is the length of i (n), and n and M both represent the time.
According to the above formula, c (n) has the maximum value when s (n) and i (n) are completely overlapped, so that n corresponding to the maximum value of c (n) can be found as the starting time t of the reboundstart. In addition, since the length of the rebounded signal is M, the time t at which the rebounding ends can be obtained accordinglyend. At the time of obtaining tstartAnd tendIn this case, the bounce wireless signal B of each tag in any operation period can be accurately separated from all the bounce wireless signals (B ═ B)1,b2,…,bK),biIs the fraction of the i-th tag bounces and K is the number of tags.
Preferably, the determining, by the similarity probability, the neighboring robot belonging to the attacking robot is specifically: constructing a distance matrix based on the distance between every two signal characteristic matrixes; calculating the similarity probability between every two adjacent robots by adopting a logistic regression method based on the distance matrix to obtain a similarity probability matrix; and on the basis of the similarity probability matrix, adopting a dichotomy, if the similarity probability between two adjacent robots is greater than a threshold value, enabling the two adjacent robots to attack the robots, otherwise, determining the adjacent robots belonging to the attacking robots.
Preferably, the denoising and calculating the amplitude average value of the bounce point of the denoised bounce wireless signal received by the single antenna in any working period of each tag specifically includes:
taking a first average value of the amplitudes of all rebound sampling points and a second average value of the amplitudes of all non-rebound sampling points in the rebound wireless signal of any working period of each label, and subtracting the second average value from the first average value to obtain the amplitude average value of the rebound point of the label in the working period;
then the multipath amplitude mean value vector of the adjacent robot in each received period is formed by the amplitude mean value of all the labels of each adjacent robot in each received period.
It should be noted that the amplitude mean value obtained by the denoising method is regarded as the actual rebound amplitude, so the multipath amplitude mean value vector may be referred to as the multipath actual rebound amplitude vector.
Since all objects in the environment of the target robot will bounce the wireless signal, the bounce wireless signal of each tag in each working period is further extracted from the bounce wireless signal to remove noise. A rebound wireless signal b corresponding to the ith label in a certain period of time when the wireless signal of a certain nearby robot is receivedi(N) assuming that there are N each1And N0Individual bounce and non-bounce sample points (tag i bounce sequence, with N bounce states per bounce state)1Each bounce sample point having N in non-bounce state0One sample point). The actual bounce magnitude from the ith tag can then be expressed as:
when the wireless signal of the current adjacent robot is received, the K labels correspond to K actual rebounding amplitudes p, and a multipath amplitude mean vector p of the adjacent robot in the received period is obtained (p ═ p)1,p2,…,pK). In view of piEqual to or greater than 0, so the multipath amplitude mean vector p is a non-negative vector in the feature space.
In order to solve the problem that signal fluctuation caused by the moving state of the robot causes that a single multipath amplitude mean vector p cannot accurately represent the spatial uniqueness of a signal source, for each adjacent robot, the method periodically extracts a multipath amplitude mean vector p (namely the preset adjacent robot time sequence is divided into a plurality of time sequence cycles, each time sequence cycle comprises the sequence of each adjacent robot and a received time period), and a series of multipath amplitude mean vectors p are taken to form a signal characteristic matrix F to represent the motion track of the adjacent robot in a period of time. In general, a signal characteristic matrix P ═ P1;p2;…;pL],piAnd a multipath amplitude mean value vector p representing the ith clock period, wherein L represents the number of clock periods.
Preferably, before the signal feature matrix of the neighboring robot is constructed, the method 100 further includes: and regularizing each multipath amplitude mean value vector L-2 to obtain a new multipath amplitude mean value vector.
After the multipath amplitude mean vector p is obtained, L-2 regularization is carried out on p to resist the change of the transmission power of an attacker. Specifically, p is divided by its L-2 norm to obtain a regularized multipath magnitude mean vector f.
Then a signal characteristic matrix P is converted to F ═ F1;f2;…;fL],fiAnd expressing the regularized multipath amplitude mean value vector f of the ith period, wherein L expresses the number of periods.
Preferably, the preset time sequence of the adjacent robots is divided into a plurality of time sequence cycles, each time sequence cycle comprises the sequence of each adjacent robot and a received time period, the received time lengths of the adjacent robots are the same, and the sequence of the robots in each time sequence cycle is the same; calculating the distance between every two signal feature matrixes specifically as follows:
and respectively calculating the cosine distance between the multipath amplitude mean value vectors in the same time sequence period between every two signal feature matrixes.
Preferably, the multipath amplitude mean value vector distance between the ith adjacent robot and the jth adjacent robot in the ith time sequence period And d isij≠dji,filIs the multipath amplitude mean value vector, f, of the ith adjacent robot in the ith time sequence periodjlThe multipath amplitude mean vector of the jth adjacent robot in the ith time sequence period is shown, and L is the number of the time sequence periods.
For example, now that there are N neighboring robots in a robot network communicating with a target robot, their signal feature matrices can be represented as a set { F }1,F2,…,FN}. Based on these feature matrices, the distance between two matrices can be calculated and a distance matrix D is obtained:
wherein d isijIs a matrix FiAnd matrix FjThe distance between them. Because one feature matrix comprises L multipath amplitude mean value vectors, the distance between the multipath amplitude mean value vectors corresponding to the same allowable period in every two matrixes is calculated, and an L-dimensional distance vector d is obtainedij=(d1,d2,…,dL)。
Wherein, the matrix FiAnd matrix FjDistance d in the l-th timing cyclelCan be expressed as a modified cosine distance Is noteworthy because So there is distadcos(fil,fjl)≠distadcos(fjl,fil) I.e. dij≠dji。
Based on the distance matrix D obtained in the above way, the method can estimate the similarity of every two adjacent robots, thereby identifying the false robot forged by the Sybil attacker. Specifically, for the distance d between two adjacent robots i and jijIt is desired to obtain a similarity probability of sij=P(i=j|dij). Here, sijTrends 0 indicate two different nearby robots, while trends 1 indicate two fake robots forged by the same attacker. To achieve this, the conditional probability P (. cndot. cndot.) is learned using a logistic regression method (LR). Based on logistic regression, the likelihood probability sij=g(w·dij+ b), where w and b represent the weight and offset parameters, respectively. As shown in fig. 4, these two parameters together determine the decision boundary in the feature space. The function g (-) is a sigmoid function and is expressed asLR is a generalized linear model, and the result after weighting z ═ w · dij+ b maps between 0 and 1, which is well suited for estimating the likelihood probability between two robots. In addition, LR is a low computational complexity algorithm because the parameters w and b that need to be learned are both linear. Removing deviceBesides, the LR also utilizes the sigmoid function to eliminate the problem of asymmetric data distribution, and can alleviate the influence of abnormal points. These advantages of LR greatly eliminate the effects of robot movement and environmental changes.
In the training phase, Maximum Weighted Likelihood Estimation (MWLE) is used to estimate the parameters w and b. In particular, in MWLE, the goal is to maximize the weight likelihood function L (w, b):
wherein z isn=w·dn+ b. Where N is the number of training samples, dnIs a distance vector sample, ynE {0,1} is the class v of the samplenIs the weight of the sample. In this example, y n1 denotes the distance vector between two dummy robots, y n0 denotes two different robots. In addition, to further mitigate the effects of sample imbalance, the sample weight v is weightednSet to be inversely proportional to the number of samples of the same class.
After all distance vectors are input into the LR model, a similarity probability matrix S is obtained:
for two robots, two distance vectors dijAnd djiAre different, so that there are two corresponding similar probabilities sijAnd sjiAnd not the same. Therefore, the similarity probability matrix S is asymmetric.
Based on the similarity matrix S, the basis that two adjacent robots i and j are forged by the same attacker is SijNot less than sigma and sji≧ σ, where σ ∈ (0,1) is the decision threshold. Considering sijAnd sjiIs two probability values, σ may be set to 0.5.
The method is applied to detecting Sybil attack behaviors in a multi-robot network, so that a single target robot can find a malicious adjacent robot with multiple identities, and as shown in fig. 5, the method comprises the following steps: the method comprises the steps that wireless signals from adjacent robots are reflected by a plurality of backscattering labels placed on a target robot, the target robot firstly constructs a unique signal characteristic matrix for each adjacent robot based on extracted rebounding wireless signals, then the distance between every two characteristic matrices is effectively measured, then a matrix containing the similarity probability between every two adjacent robots is generated based on the measured distance, finally, the adjacent robots at the same position or on the same path have similar signal characteristics, and legal robots and malicious attackers in the adjacent robots are judged according to the similar matrices based on the signal space characteristics so as to detect false robots counterfeited by Wuyi attackers in the current network. By the invention, even on a lightweight robot platform only provided with a single antenna, accurate Sybil attack detection can be realized.
Example two
A detection system for an attacking robot in a multi-robot network is a target robot with an antenna, a plurality of labels and a control processor, wherein the control processor adopts any one of the detection methods for the attacking robot in the multi-robot network, and controls a single antenna and the labels to collect rebounding wireless signals and detect the rebounding wireless signals to obtain the attacking robot of the current robot.
The related technical solution is the same as the first embodiment, and is not described herein again.
Example two
A storage medium, wherein the storage medium stores instructions, and when the instructions are read by a computer, the computer is caused to execute the attack robot detection method in the multi-robot network according to the first embodiment.
The related technical solution is the same as the first embodiment, and is not described herein again.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A method for detecting an attack robot in a multi-robot network is characterized by comprising the following steps:
a target robot in the multi-robot network sequentially and directly receives a wireless signal of each adjacent robot through a single antenna of the target robot according to a preset adjacent robot time sequence, and meanwhile, in a time period when the single antenna directly receives the wireless signal of any adjacent robot, a plurality of labels arranged on the target robot sequentially receive the wireless signal of the adjacent robot according to a preset working time sequence and rebound to the single antenna;
denoising a rebound wireless signal of each label received by a single antenna in any working period, solving an amplitude mean value of a rebound point, constructing a multipath amplitude mean value vector corresponding to each adjacent robot in each received period based on the amplitude mean values corresponding to the labels, and constructing a signal feature matrix of the adjacent robot by all the multipath amplitude mean value vectors of the adjacent robot;
and calculating and determining the adjacent robots belonging to the attacking robot through the similarity probability based on the distance between every two signal feature matrixes.
2. The method as claimed in claim 1, wherein each of the tags is a passive backscatter tag.
3. The method for detecting the attacking robot in the multi-robot network as claimed in claim 1, wherein the method for extracting the rebounding wireless signal of any working time period of each tag from the rebounding wireless signals received by the single antenna within the preset time period comprises:
carrying out moving average on the rebounded wireless signals received by the single antenna within a preset time period;
and according to the adjacent robot time sequence, performing correlation calculation on the rebound wireless signals corresponding to each adjacent robot after the sliding average and the preset working time sequence, and extracting the rebound wireless signals of any working time period of each label received by the single antenna.
4. The method for detecting the attacking robot in the multi-robot network as claimed in claim 1, wherein the determining of the neighboring robot belonging to the attacking robot through the similarity probability specifically comprises:
constructing a distance matrix based on the distance between every two signal characteristic matrixes;
calculating the similarity probability between every two adjacent robots by adopting a logistic regression method based on the distance matrix to obtain a similarity probability matrix;
and based on the similarity probability matrix, through similarity judgment, if the similarity probability between two adjacent robots is greater than a threshold value, the two adjacent robots are attacking robots, otherwise, the two adjacent robots are not attacking robots, so as to determine the adjacent robots belonging to the attacking robots.
5. The method for detecting the attacking robot in the multi-robot network as claimed in any one of claims 1 to 4, wherein the denoising and averaging of the rebounded wireless signals received by the single antenna in any working period of each tag is performed to obtain the average value of the amplitudes of the rebounded points, specifically:
taking a first average value of the amplitudes of all rebound sampling points and a second average value of the amplitudes of all non-rebound sampling points in the rebound wireless signal of any working period of each label, and subtracting the second average value from the first average value to obtain the amplitude average value of the rebound point of the label in the working period;
and forming a multipath amplitude mean value vector of the adjacent robot in each received period by the amplitude mean value of all the labels of each adjacent robot in each received period.
6. The method for detecting the attacking robot in the multi-robot network as claimed in any one of claims 1 to 4, wherein before said constructing the signal feature matrix of the neighboring robot, said method further comprises:
and regularizing each multipath amplitude mean value vector L-2 to obtain a new multipath amplitude mean value vector.
7. The method for detecting the attacking robot in the multi-robot network as claimed in any one of claims 1 to 4, wherein said preset timing sequence of the adjacent robots is divided into a plurality of timing cycles, each timing cycle comprising the sequence of the adjacent robots and a received time period;
calculating the distance between every two signal feature matrices, specifically:
and respectively calculating the cosine distance between the multipath amplitude mean value vectors in the same time sequence period between every two signal feature matrixes.
8. The method as claimed in claim 7, wherein the multipath amplitude mean vector distance between the ith neighboring robot and the jth neighboring robot in the ith timing cycle is determined by the multi-robot network attack robot detection method And d isij≠dji,filIs the multipath amplitude mean value vector, f, of the ith adjacent robot in the ith time sequence periodjlThe multipath amplitude mean vector of the jth adjacent robot in the ith time sequence period is shown, and L is the number of the time sequence periods.
9. An attack robot detection system in a multi-robot network, characterized in that the system is a target robot provided with an antenna, a plurality of tags and a control processor in the multi-robot network, wherein the control processor adopts the attack robot detection method in the multi-robot network according to any one of claims 1 to 8, and controls a single antenna and the tags to collect rebounding wireless signals and detect to obtain the attack robot of the current robot.
10. A storage medium having stored therein instructions, which when read by a computer, cause the computer to execute a method for detecting an attacking robot in a multi-robot network according to any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911099762.0A CN111083106B (en) | 2019-11-12 | 2019-11-12 | Detection method and detection system for attack robot in multi-robot network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911099762.0A CN111083106B (en) | 2019-11-12 | 2019-11-12 | Detection method and detection system for attack robot in multi-robot network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111083106A true CN111083106A (en) | 2020-04-28 |
CN111083106B CN111083106B (en) | 2020-12-08 |
Family
ID=70310792
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911099762.0A Active CN111083106B (en) | 2019-11-12 | 2019-11-12 | Detection method and detection system for attack robot in multi-robot network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111083106B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112702110A (en) * | 2020-12-31 | 2021-04-23 | 华中科技大学 | Method, device and system for simultaneous communication and positioning of multiple backscatter tags |
CN116886446A (en) * | 2023-09-06 | 2023-10-13 | 北京安天网络安全技术有限公司 | Automatic attack detection method, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9094816B2 (en) * | 2006-05-16 | 2015-07-28 | RedSky Technologies, Inc. | Method and system for an emergency location information service (E-LIS) from unmanned aerial vehicles (UAV) |
CN108353081A (en) * | 2015-09-28 | 2018-07-31 | 13部门有限公司 | Unmanned plane intrusion detection and confrontation |
CN109218744A (en) * | 2018-10-17 | 2019-01-15 | 华中科技大学 | A kind of adaptive UAV Video of bit rate based on DRL spreads transmission method |
CN110073301A (en) * | 2017-08-02 | 2019-07-30 | 强力物联网投资组合2016有限公司 | The detection method and system under data collection environment in industrial Internet of Things with large data sets |
CN110166265A (en) * | 2018-02-11 | 2019-08-23 | 陕西爱尚物联科技有限公司 | A kind of method and its module of network control |
US20190339688A1 (en) * | 2016-05-09 | 2019-11-07 | Strong Force Iot Portfolio 2016, Llc | Methods and systems for data collection, learning, and streaming of machine signals for analytics and maintenance using the industrial internet of things |
-
2019
- 2019-11-12 CN CN201911099762.0A patent/CN111083106B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9094816B2 (en) * | 2006-05-16 | 2015-07-28 | RedSky Technologies, Inc. | Method and system for an emergency location information service (E-LIS) from unmanned aerial vehicles (UAV) |
CN108353081A (en) * | 2015-09-28 | 2018-07-31 | 13部门有限公司 | Unmanned plane intrusion detection and confrontation |
US20190339688A1 (en) * | 2016-05-09 | 2019-11-07 | Strong Force Iot Portfolio 2016, Llc | Methods and systems for data collection, learning, and streaming of machine signals for analytics and maintenance using the industrial internet of things |
CN110073301A (en) * | 2017-08-02 | 2019-07-30 | 强力物联网投资组合2016有限公司 | The detection method and system under data collection environment in industrial Internet of Things with large data sets |
CN110166265A (en) * | 2018-02-11 | 2019-08-23 | 陕西爱尚物联科技有限公司 | A kind of method and its module of network control |
CN109218744A (en) * | 2018-10-17 | 2019-01-15 | 华中科技大学 | A kind of adaptive UAV Video of bit rate based on DRL spreads transmission method |
Non-Patent Citations (1)
Title |
---|
张红强: "《基于简化虚拟受力模型的群机器人自组织协同围捕研究》", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112702110A (en) * | 2020-12-31 | 2021-04-23 | 华中科技大学 | Method, device and system for simultaneous communication and positioning of multiple backscatter tags |
CN112702110B (en) * | 2020-12-31 | 2022-02-15 | 华中科技大学 | Method, device and system for simultaneous communication and positioning of multiple backscatter tags |
CN116886446A (en) * | 2023-09-06 | 2023-10-13 | 北京安天网络安全技术有限公司 | Automatic attack detection method, electronic equipment and storage medium |
CN116886446B (en) * | 2023-09-06 | 2023-11-24 | 北京安天网络安全技术有限公司 | Automatic attack detection method, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN111083106B (en) | 2020-12-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20220094710A1 (en) | Detection of cyber attacks targeting avionics systems | |
CN104320845B (en) | A kind of primary user's localization method based on sensor and quantum intelligence computation | |
CN110049057B (en) | Sensor network event trigger information fusion method under false data injection attack | |
CN111083106B (en) | Detection method and detection system for attack robot in multi-robot network | |
de Lima Pinto et al. | A machine learning approach for detecting spoofing attacks in wireless sensor networks | |
Sun et al. | BERT-ADLOC: A secure crowdsourced indoor localization system based on BLE fingerprints | |
Li et al. | A transfer double deep Q network based DDoS detection method for internet of vehicles | |
Rodić et al. | Privacy leakage of LoRaWAN smart parking occupancy sensors | |
Shi et al. | Towards environment-independent human activity recognition using deep learning and enhanced CSI | |
Jajamovich et al. | Joint multitarget tracking and sensor localization in collaborative sensor networks | |
Khalid et al. | Physical layer authentication in line-of-sight underwater acoustic sensor networks | |
Koroniotis et al. | A deep learning-based penetration testing framework for vulnerability identification in internet of things environments | |
Yang et al. | Determining the number of attackers and localizing multiple adversaries in wireless spoofing attacks | |
Yang et al. | Jointly beam stealing attackers detection and localization without training: an image processing viewpoint | |
Huang et al. | Lightweight sybil-resilient multi-robot networks by multipath manipulation | |
Pawar et al. | Detection of blackhole and wormhole attacks in WSN enabled by optimal feature selection using self-adaptive multi-verse optimiser with deep learning | |
Renganathan et al. | Spoof resilient coordination in distributed and robust robotic networks | |
Rashid et al. | Cyber attacks detection from smart city applications using artificial neural network | |
Shi et al. | WiFi-based activity recognition using activity filter and enhanced correlation with deep learning | |
Zhang et al. | Poison neural network-based mmWave beam selection and detoxification with machine unlearning | |
Muthu Mariappan et al. | A novel location pinpointed anti-jammer with knowledged estimated localizer for secured data transmission in mobile wireless sensor network | |
Zhao et al. | Backdoor Attacks Against Deep Learning-Based Massive MIMO Localization | |
US20220272122A1 (en) | Autonomous vehicle control attack detection and countermeasures | |
KP | Delta Ruled First Order Iterative Deep Neural Learning for Sybil and Wormhole Attacks Detection in Healthcare Wireless Sensor Network | |
Asaad et al. | Fingerprinting Based Positioning Techniques Using Machine Learning Algorithms Principles, Approaches and Challenges |
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 |