A Multidimensional Vehicle Trust Evaluation Method for the Internet of Vehicles

Abstract

The present invention belongs to the technical field of vehicle networking security, and discloses a multi-dimensional vehicle trust evaluation method for vehicle networking. The evaluation method mainly consists of two parts: in the vehicle-to-vehicle trust module, the evaluating vehicle analyzes the context and message quality of the target vehicle, and combines the recommendations of neighboring vehicles to establish the vehicle-to-vehicle trust value of the target vehicle, and decides whether to receive or discard the messages sent by it, which can effectively reduce the spread of malicious messages in the network; in the vehicle global reputation module, the RSU filters malicious feedback, calculates the feedback trust of the target vehicle using an asymmetric forgetting factor, can resist malicious slander attacks and impose heavier penalties on illegal behaviors, and the fog server calculates the global reputation value after considering its historical behavior and then decides whether to add it to the blacklist. The present invention can not only measure the reliability of vehicle nodes, but also measure the reliability of vehicle messages, and at the same time has a certain robustness against malicious attacks existing in the network.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle networking security, and more particularly to a reputation evaluation scheme based on multi-dimensional trust measurement in a vehicle networking environment. Background Art

[0002] With the continuous maturity of wireless communication technology, Vehicular Ad Hoc Networks (VANETs) have reached a new height. By providing traffic safety information and various entertainment service information, it has significantly improved traffic safety and efficiency, brought a good driving experience to drivers, and become an essential part of the intelligent transportation system, attracting extensive attention and research from people in all walks of life. However, due to the openness of wireless communication, the security issues of VANETs cannot be ignored, especially the trust security issues.

[0003] VANETs provide various application services based on the collaborative data and information exchange between vehicles and between vehicles and roadside infrastructure. Vehicles can sense the surrounding environment through equipped sensors, collect relevant information, and communicate with other vehicles or roadside units (RSUs) (also known as V2V communication and V2I communication) with the help of on-board units (OBUs), informing surrounding vehicles or RSUs of their own position, speed, sensed events, etc. Adjacent vehicles can take corresponding measures based on these safety messages. For example, a vehicle that receives a traffic jam message can reduce its own driving speed to prevent collisions. At the same time, there are also security risks. If the received safety message is unreliable, such as a vehicle broadcasting false messages or not broadcasting sensed events to save resources, it may cause other vehicles to make wrong decisions and lead to serious traffic accidents. It can be seen that effective information distribution is an important factor to ensure the successful implementation of VANETs applications. Before taking any action based on the received message, a vehicle should check the credibility of the message. Therefore, a reputation evaluation mechanism is needed to solve the trust security problems of VANETs, analyze the reliability of vehicle nodes and vehicle messages, and achieve trust management.

[0004] Currently, researchers at home and abroad have proposed different trust management solutions for trust and security issues in VANETs. From the perspective of the subject, they can be divided into three categories: entity-oriented, data-oriented, and hybrid models. The main difference between the first two lies in whether to revoke malicious nodes or illegal messages. The hybrid model comprehensively considers the reliability of entities and data. From the perspective of architecture composition, they can be divided into centralized and distributed types. A centralized trust management system requires a trust center to uniformly perform trust calculations, which usually results in a large delay. However, the vehicle networking environment is very sensitive to delay, especially when processing emergency messages related to security. In some distributed trust management systems, reliable network edge computing resources in infrastructure such as RSU are rarely utilized. Usually, only the resources of vehicle nodes are considered for trust calculation. However, due to the extremely limited resources of vehicle nodes, they are easily attacked. Summary of the Invention

[0005] In order to solve the trust problems of vehicle nodes and messages in the untrusted environment of vehicle networking in the prior art, the present invention proposes a reputation evaluation scheme based on multi-dimensional trust measurement. Using fog computing technology, through the vehicle-to-vehicle trust module, the vehicle analyzes the context and message quality of the target vehicle, combines the recommendations of neighboring vehicles to establish a vehicle-to-vehicle trust value for the target vehicle, and decides to receive or discard the messages sent by it, which can effectively reduce the spread of malicious messages in the network; through the vehicle global reputation module, the RSU filters malicious feedback, calculates the feedback trust of the target vehicle using an asymmetric forgetting factor, can resist malicious slander attacks and impose heavier penalties on illegal behaviors; the fog server calculates the global reputation value after considering its historical behavior and then decides whether to add it to the blacklist, which can reduce the number of malicious nodes in the network.

[0006] To achieve the above object, the present invention is implemented through the following technical solutions:

[0007] The present invention is a multi-dimensional vehicle trust evaluation method for vehicle networking, including the following steps:

[0008] Step 1) The trusted center TA is responsible for initializing the entire reputation system and sending the latest vehicle global reputation scores to each fog server serving as a local trusted center. If there is a new vehicle user joining the vehicular ad hoc network, TA is responsible for its registration work.

[0009] Step 2) The vehicle nodes in the vehicular ad hoc network periodically broadcast beacon messages, mainly including: context information such as vehicle ID, speed, location, current time, etc.; in addition, when a traffic event occurs, the vehicle nodes near the event location will sense it first, and then the nodes will broadcast the sensed event. The vehicle nodes far from the event location can learn about the event by receiving the message. Malicious nodes in the network will launch attacks, such as tampering with the message content, making malicious feedback, etc.

[0010] Step 3) The vehicle that receives the message calculates the trust between vehicles for the vehicle that sends the message. Introduce a small time interval, called a time slot, and calculate the trust value in each time slot. The vehicle that performs the trust calculation is called the evaluating vehicle V i , and the vehicle to be evaluated is called the target vehicle V j ; V i Calculate the direct trust DT of V j based on the context relevance and message quality, and then calculate the recommended trust RT based on the recommendation scores of other neighbor vehicles for the target vehicle i,j . Finally, combine the historical interaction status of V i,j to calculate the latest vehicle-to-vehicle trust value TrV j . i,j .

[0011] Step 4) V i Makes a message decision based on the calculated vehicle-to-vehicle trust value. If the value of TrV i,j is less than the set threshold, it means that V i does not trust the event message reported by V j , then discards the message sent by V j , otherwise receives or forwards the message; finally, feedback the satisfaction of the interaction with V j to the RSU

[0012] Step 5) The RSU collects the feedback from other vehicles on V j . First, calculate the aggregated feedback score and filter malicious feedback using the overall difference, then convert the filtered feedback score into a pair of positive and negative feedbacks, and calculate its feedback trust FT j using an asymmetric forgetting factor. Finally, send the feedback trust on V j to the local fog server

[0013] Step 6) The fog server combines the feedback trusts of multiple RSUs on V j with its historical reputation value to calculate the latest global reputation value of V j . If the global reputation value of V j is lower than the threshold, it is identified as a malicious vehicle and added to the blacklist. Upload the latest vehicle global reputation value and the blacklist to the cloud for storage, and then send the blacklist to the RSUs within the jurisdiction. The RSUs broadcast the blacklist to the vehicles within the communication range

[0014] Among them, the specific content of step 1) is as follows

[0015] Step 11) The trusted center TA sends the latest vehicle node reputation value to the corresponding fog server for localizing and evaluating the global reputation value of the vehicle

[0016] Step 12) The newly added vehicle user needs to register with TA using their real identity. TA assigns a uniquely identifiable identity ID to the user and initializes the vehicle's reputation value, and then sends the vehicle information to the corresponding local fog server.

[0017] The specific steps of step 2) are as follows:

[0018] Step 21) Vehicle nodes periodically broadcast beacon messages to inform other surrounding vehicles of context information such as vehicle ID, speed, location, and current time. When a vehicle node perceives a traffic event, it integrates the perceived event into a data message and broadcasts it. Vehicle nodes located far from the event occurrence location can learn about the event by receiving the message.

[0019] Step 22) Malicious vehicle nodes will launch attacks, and three attack modes are considered in the present invention: ① False information attack: On the one hand, the attacker may broadcast non-existent event messages, and on the other hand, may tamper with the received message content and then forward it, injecting error messages into the network. ② Malicious slander attack: The attacker provides false feedback about other vehicle nodes. ③ Switching attack: The attacker will switch between malicious behavior and honest behavior, and adopt a random mode for attacks. For false information attacks: By making decisions on the received messages through step 41), vehicles can discard false messages, reducing the malicious impact caused by the spread of false messages in the network. For malicious slander attacks: First, calculate the aggregated feedback score through step 51), and then calculate the overall difference through step 52) to filter malicious feedback, thereby reducing the adverse effects caused by malicious slander. For switching attacks: Through the asymmetric forgetting factor in step 53), the malicious behavior of the vehicle is more severely punished. When the vehicle switches from honest behavior to malicious behavior, its reputation value is quickly reduced, thereby reducing the impact caused by switching attacks.

[0020] The specific steps of step 3) are as follows:

[0021] Step 31) Evaluate vehicle V i Extract its context information from the beacon message of the target vehicle V j and process the context information into an n-dimensional normalized vector vector i 、vector j ,V i Calculate the context correlation degree of V j :

[0022]

[0023] where respectively represent the k-th value in the normalized vector, and n represents the vector (vector i 、vectorj )'s dimension, where i represents the evaluated vehicle V i , and j represents the target vehicle V j .

[0024] Step 32) Since the messages in the direct interaction process can all be decrypted and analyzed by the recipient, V i can determine the legitimacy of the data messages sent by V j . V i Counts the number of correct messages legal received from V j during this time slot i,j and the number of incorrect messages mal i,j , and calculates the message quality of V j :

[0025]

[0026] where the variable parameter tends to 1 as the number of correct messages increases, and the target vehicle can improve its message quality by sending more correct messages.

[0027] Step 33) Uses the context relevance as the weight of the message quality to adjust the accuracy of the message quality, and uses the Gompertz function to calculate the direct trust of V i in V j :

[0028]

[0029] where b and c are function parameters, b controls the displacement of the output along the x-axis, c adjusts the growth rate of the function, and the output range of the function is between 0 and 1.

[0030] Step 34) V i will give a recommendation score rs i,j for V j according to DT i,j . The higher the score, the better the performance of V j , and the more worthy of recommendation. If V i is a malicious vehicle, it may tamper with the recommendation score, give a lower or higher score, and then share it with the surrounding vehicles, thus affecting the trust evaluation of other vehicles for V j .

[0031] Step 35) V i obtains the recommendation scores rs i,k of the neighbor vehicle V j for V k,j , and selects TrV i,kRecommendation of neighbor vehicles greater than the threshold, i.e., trustworthy neighbors. Initially, the suggestions of all neighbors can be adopted. After trust is established later, only the recommendations of trusted neighbors are adopted. Denote the set of neighbor vehicles as Let V i For the neighbor vehicle V i,k The direct trust DT i,k As the weight, calculate the recommended trust:

[0032]

[0033] Step 36) V i According to the direct trust and recommended trust obtained in this time slot, and considering the historical interaction behavior of V j Calculate the latest in-vehicle trust value regarding V j :

[0034]

[0035] Where Represents the historical in-vehicle trust value with V j , λ is between 0 and 1, which is the proportion it occupies. Since more weight should be given to the recent vehicle behavior, λ should be greater than 0.5, and n is the number of interactions between the two vehicles.

[0036] The specific content of the said step 4) is as follows:

[0037] Step 41) V i Make a message decision according to TrV i,j . If the value of TrV i,j is less than the set threshold, it means that V i does not trust the event message reported by V j , then discard the message sent by V j , otherwise receive or forward the message.

[0038] Step 42) V i Give the satisfaction feedback score fs i,j for V j according to TrV i,j , and send the feedback to the RSU. The higher the feedback score, the higher the satisfaction of V i with V j . If V i is a malicious vehicle, it may tamper with the feedback score and then send it to the RSU.

[0039] The specific content of the said step 5) is as follows:

[0040] Step 51) Denote the set of vehicles that give feedback to V j as The RSU obtains the historical reputation value of the feedback vehicle from the fog server Calculate regarding V j The aggregated feedback score of:

[0041]

[0042] Then calculate V x Regarding V j The feedback difference value of

[0043]

[0044] Step 52) The communication between vehicles is multi - to - multi. V x In addition to giving feedback on V j It will also give feedback to other vehicles it has interacted with. Malicious vehicles in the network may initiate a switch attack. The RSU needs to consider the overall feedback difference of V x And then decide whether to adopt the feedback of V x The set of vehicles that have been given feedback by V x Is denoted as E x The total number of vehicles in it is n x Calculate the overall feedback difference value of V x :

[0045]

[0046] If Diff x Is greater than the set threshold, then do not adopt the feedback sent by V x Delete it from T j

[0047] Step 53) For The RSU converts its feedback value fs x,j Into a pair of positive and negative feedback, denoted as <α x ,β x >, where α x Is the positive feedback representing the satisfaction degree, that is, α x = fs x,j ; β x Is the negative feedback representing the dissatisfaction degree, that is, β x = 1 - fs x,j . Add a forgetting factor so that the weight of the most recent feedback is greater than that of the old feedback. Use the historical reputation value of the vehicle As the weight of the feedback, then the positive feedback α x And the negative feedback β x Are:

[0048]

[0049] Among them, ζ1 and ζ2 are asymmetric forgetting factors, where ζ1 < ζ2, so that malicious behaviors of vehicles are punished more severely. t is the current time, and t x represents the time when the feedback is sent. Positive and negative feedback calculations are used to calculate V j Feedback trust:

[0050]

[0051] Then the feedback trust of V j is sent to the local fog server.

[0052] Step 61) The fog server combines the feedback trusts of multiple RSUs for V j with its historical reputation value to calculate the latest global reputation value of V j :

[0053]

[0054] Among them, λ is the same as in the vehicle-to-vehicle trust, between 0 and 1. Taking λ greater than 0.5 assigns more weight to the latest feedback trust. AVG(FT j ) represents the average value of the feedback trusts of multiple RSUs for V j .

[0055] Step 62) If the reputation value of V j is lower than the threshold, it is identified as a malicious vehicle and added to the blacklist. The latest global reputation value of the vehicle and the blacklist are uploaded to the cloud for storage, and then the blacklist is sent to the RSUs within the jurisdiction. The RSUs broadcast the blacklist to the vehicles within the communication range.

[0056] The beneficial effects of the present invention are:

[0057] (1) The present invention uses multi-dimensional trust metrics to solve security problems in the vehicle networking environment, implementing a hybrid trust evaluation scheme. On the one hand, it can measure the authenticity of vehicle messages so that vehicles can decide whether to receive or discard the received messages. On the other hand, it can measure the reliability of vehicle nodes to determine whether they are malicious nodes.

[0058] (2) The present invention uses a differential filtering algorithm to effectively reduce the adverse effects brought by malicious feedback. By using asymmetric forgetting factors and increasing the punishment for malicious behaviors, the reputation value is easy to lose and difficult to obtain, which can effectively resist malicious attack modes such as malicious slander and switch attacks.

[0059] (3) The present invention adopts a distributed architecture to avoid large communication overhead and delay generated by the centralized reputation scheme. The local fog server is used to assist the RSU in updating the reputation score of the vehicle. The fog server can monitor the RSU to prevent the harm caused by the invasion of the RSU. Brief Description of the Drawings

[0060] Figure 1 It is a flowchart of the evaluation method of the present invention.

[0061] Figure 2 It is the actual scenario applied by the evaluation method of the present invention.

[0062] Figure 3 It is the specific flowchart of trust evaluation based on multi-dimensional trust measurement. Detailed Implementation Manner

[0063] The embodiments of the present invention will be disclosed below with reference to the drawings. For the sake of clarity, many practical details will be described together in the following description. However, it should be understood that these practical details are not used to limit the present invention. That is to say, in some embodiments of the present invention, these practical details are not necessary.

[0064] Such as Figure 2 The actual scenario applied by the evaluation scheme shown. In the vehicular ad hoc network environment, vehicle nodes can sense the surrounding environment through the equipped sensors, collect relevant information, and inform surrounding vehicles or RSU of their own positions, speeds, sensed events, etc. through V2V communication and V2I communication. Adjacent vehicles can take corresponding measures according to these safety messages. For example, a vehicle that receives a traffic collision message can reduce its own driving speed to prevent traffic jams. At the same time, there are also security risks. Due to the high openness of wireless communication, vehicle entities in VANETs are vulnerable to various types of attacks. For example, malicious attackers broadcast false messages, or do not broadcast sensed events to save resources, which may cause other vehicles to make wrong decisions and lead to serious traffic accidents. It can be seen that effective information distribution is an important factor to ensure the successful implementation of VANETs applications. Before taking any action based on the received message, a vehicle should check the credibility of the message. In addition, attackers may also use switch attacks and malicious slander to affect the credibility of vehicles. Therefore, a trust mechanism is needed to solve the trust problem in the vehicle network, analyze the reliability of vehicle nodes and vehicle messages, and achieve trust management.

[0065] Figure 3This is the specific flowchart of the trust evaluation of the present invention, which mainly consists of two parts: workshop trust and vehicle global reputation. In the workshop trust evaluation module, the evaluating vehicle first obtains the context information of the target vehicle from the beacon message, compares the context relevance between the two, and analyzes its message quality. Vehicles in a similar context environment can perceive more common data, have a more comprehensive understanding of each other, and will be more accurate in measuring message quality. The direct trust DT is calculated through the Gompertz function using the context relevance Con and the message quality Mq. Then, the evaluating vehicle selects the recommended scores rs of the trusted neighbor vehicles for the target vehicle and aggregates them to obtain the recommended trust RT. Finally, considering the historical interaction situation with the target vehicle, the current workshop trust value TrV is updated, and the latest workshop trust value is used to determine whether the event message reported by the target vehicle is credible. If it is worthy of trust, the event message is received or forwarded, otherwise the message is discarded, so as to reduce the malicious impact brought by false messages.

[0066] In the vehicle global reputation module, after the RSU collects the feedback of other vehicles on the target vehicle within its jurisdiction, to avoid slander attacks by malicious vehicles, it first calculates the aggregated feedback score IFS and filters the malicious feedback using the overall difference value Diff. Subsequently, the filtered feedback score fs is converted into a pair of positive and negative feedback <α,β>, and then its feedback trust FT is calculated. An asymmetric time forgetting factor (ζ1, ζ2) is added to the feedback to impose a heavier penalty on the malicious behavior of the vehicle, making the vehicle reputation value easy to lose and difficult to obtain. When a malicious vehicle launches a switch attack, its reputation value can be quickly reduced. After the fog server obtains the feedback trust of the RSU for the target vehicle, it combines its historical reputation value RepV last to update the final global reputation value RepV, and uses the latest reputation value to measure the reliability of the vehicle node. If the vehicle node is unreliable, it is added to the blacklist, and then the blacklist is broadcast to other vehicle nodes through the RSU, so as to reduce the number of malicious nodes in the network.

[0067] Specifically, the multi-dimensional vehicle trust evaluation method for the vehicle networking includes the following steps:

[0068] Step 1) The trusted center TA is responsible for the initialization of the entire reputation system, and sends the vehicle global reputation score calculated last time to the corresponding fog server for localizing the evaluation of the vehicle global reputation value. If there is a newly added vehicle user, it registers with the TA using its real identity. The TA assigns it a unique identifiable identity ID and initializes the reputation value, and then sends the vehicle information to the corresponding local fog server.

[0069] Step 2) The vehicle nodes in the vehicular ad hoc network periodically broadcast beacon messages, mainly including context information such as vehicle ID, speed, location, and current time. In addition, when a traffic event occurs, the vehicle nodes near the event location will sense it first, and then the nodes will broadcast the sensed event. The vehicle nodes farther away from the event location can learn about the event by receiving the messages. Malicious nodes in the network will launch attacks. In the present invention, three attack modes are considered: false information attack, malicious slander attack, and switch attack.

[0070] Step 3) The vehicles that receive the messages calculate the inter-vehicle trust of the vehicles that send the messages. The vehicle that conducts the trust calculation is called the evaluation vehicle V i , and the vehicle to be evaluated is called the target vehicle V j . V i calculates the direct trust DT j of V i,j based on the context relevance and message quality, and then calculates the recommended trust RT i,j based on the recommendation scores of other neighbor vehicles for the target vehicle. Finally, the latest inter-vehicle trust value TrV j is calculated by combining the historical interaction status of V i,j . When the vehicles that receive the messages calculate the trust value between vehicles for the vehicles that send the messages, a method of calculating the trust value regularly is adopted, introducing a time interval, i.e., a time slot. The trust value is calculated in each time slot. That is to say, when calculating the trust value, it is not calculated every time a message is received, but calculated regularly. For example, it is calculated every 30 s, and this interval is called the time slot.

[0071] Specifically, it includes the following steps:

[0072] Step 31) The evaluation vehicle V i extracts its context information from the beacon message of the target vehicle V j , processes the context information into an n-dimensional normalized vector vector i , vector j , and the evaluation vehicle V i calculates the context relevance degree of the target vehicle V j :

[0073]

[0074] where respectively represent the k-th value in the normalized vector, n represents the dimension of the vector (vector i , vector j ), i represents the evaluation vehicle V i , and j represents the target vehicle V j .

[0075] Step 32) Since the messages in the direct interaction process can all be decrypted and analyzed by the receiver, vehicle V is evaluated i to determine whether the data message sent by the target vehicle V j is legal. Vehicle V for evaluation i counts the number legal of correct messages received from the target vehicle V during this time slot j and the number mal of incorrect messages i,j and calculates the message quality of the target vehicle V i,j : j

[0076]

[0077] where the variable parameter tends to 1 as the number of correct messages increases. The target vehicle can improve its message quality by sending more correct messages.

[0078] Step 33) Use the context relevance as the weight of the message quality to adjust the accuracy of the message quality, and use the Gompertz function to calculate the direct trust of vehicle V for evaluation i in the target vehicle V j :

[0079]

[0080] where b and c are function parameters. b controls the displacement of the output along the x-axis, and c adjusts the growth rate of the function. The output range of the function is between 0 and 1.

[0081] Step 34) Vehicle V for evaluation i will give a recommendation score rs i,j for the target vehicle V j . The higher the score, the better the performance of the target vehicle V i,j , and the more worthy of recommendation. If vehicle V for evaluation j is a malicious vehicle, it may tamper with the recommendation score, give a lower or higher score, and then share it with the surrounding vehicles, thus affecting the trust evaluation of other vehicles in the target vehicle V i . j

[0082] Step 35) Vehicle V for evaluation i obtains the recommendation score rs i,k of the neighbor vehicle V j for the target vehicle V k,j , and selects TrV i,k ​​Recommendation of neighbor vehicles greater than the threshold, i.e., trustworthy neighbors. Initially, the suggestions of all neighbors can be adopted. After establishing trust later, only the recommendations of trusted neighbors are adopted. Denote the set of neighbor vehicles as Denote the evaluation vehicle V i For neighbor vehicle V i,k The direct trust DT i,k As the weight, calculate the recommended trust:

[0083]

[0084] Step 36) Evaluation vehicle V i According to the direct trust and recommended trust obtained in this time slot, and considering the historical interaction behavior of the target vehicle V j Calculate the latest vehicle-to-vehicle trust value regarding the target vehicle V j :

[0085]

[0086] Where Represents the historical vehicle-to-vehicle trust value with the target vehicle V j , λ is between 0 and 1 and is the proportion it occupies. Since more weight should be given to the recent vehicle behavior, λ should be greater than 0.5, and n is the number of interactions between the two vehicles.

[0087] Step 4) Evaluation vehicle V i On the one hand, make a message decision according to TrV i,j . If the value of TrV i,j is less than the set threshold, it means that the evaluation vehicle V i does not trust the event message reported by the target vehicle V j , then discard the message sent by the target vehicle V j . Otherwise, receive or forward the message. On the other hand, the evaluation vehicle V i gives a satisfaction feedback score fs i,j to the target vehicle V j according to TrV i,j , and send the feedback to the RSU. The higher the feedback score, the higher the satisfaction of the evaluation vehicle V i with the target vehicle V j . If the evaluation vehicle V i is a malicious vehicle, it may tamper with the feedback score and then send it to the RSU.

[0088] Step 5) The RSU collects the feedback of other vehicles on the target vehicle V j . First, calculate the aggregated feedback score, filter the malicious feedback using the overall difference, and then convert the filtered feedback score into a pair of positive and negative feedbacks, and calculate its feedback trust FT using the asymmetric forgetting factorj , finally, the feedback trust of the target vehicle V j is sent to the local fog server.

[0089] The specific steps of step 5) are as follows:

[0090] Step 51): Denote the set of vehicles that give feedback to the target vehicle V j as The RSU obtains the historical reputation values of the feedback vehicles from the fog server Calculate the aggregated feedback score regarding the target vehicle V j :

[0091]

[0092] Then calculate the feedback difference value of the feedback vehicle V x regarding the target vehicle V j

[0093]

[0094] Step 52): The communication between vehicles is multi - to - multi. The feedback vehicle V x in addition to giving feedback to the target vehicle V j , will also give feedback to other vehicles it has interacted with. Malicious vehicles in the network may initiate switch attacks. The RSU needs to consider the overall feedback difference of the feedback vehicle V x and then decide whether to adopt the feedback of the feedback vehicle V x . Denote the set of vehicles that have been feedback by the feedback vehicle V x as E x The total number of vehicles in it is n x , calculate the overall feedback difference value of the feedback vehicle V x :

[0095]

[0096] If Diff x is greater than the set threshold, then do not adopt the feedback sent by V x , and delete it from T j .

[0097] Step 53): For The RSU converts its feedback value fs x,j into a pair of positive and negative feedback, denoted as <α x , β x , where α x is the positive feedback representing the satisfaction degree, that is, α x = fs x,j ; β xNegative feedback represents dissatisfaction, i.e., β x = 1 - fs x,j . Adding a forgetting factor makes the weight of recent feedback greater than that of old feedback. Taking the historical reputation value of the vehicle as the weight of feedback, the positive feedback α x and negative feedback β x are as follows:

[0098]

[0099] where ζ1 and ζ2 are asymmetric forgetting factors, ζ1 < ζ2, so that malicious behaviors of the vehicle are punished more severely. t is the current time, and t x represents the time when the feedback is sent. Calculate the feedback trust for V j using positive and negative feedback:

[0100]

[0101] Then send the feedback trust for the target vehicle V j to the local fog server.

[0102] Step 6) The fog server combines the feedback trusts from multiple RSUs for the target vehicle V j with its historical reputation value to calculate the latest global reputation value of the target vehicle V j :

[0103]

[0104] where λ is the same as in the in-vehicle trust, between 0 and 1. Taking λ greater than 0.5 gives more weight to the latest feedback trust. AVG(FT j ) represents the average of the feedback trusts from multiple RSUs for the target vehicle V j . If the global reputation value of the target vehicle V j is lower than the threshold, it is identified as a malicious vehicle and added to the blacklist. Upload the latest global reputation value of the vehicle and the blacklist to the cloud for storage, and then send the blacklist to the RSUs within the jurisdiction. The RSUs broadcast the blacklist to the vehicles within the communication range.

[0105] The reputation scheme of the present invention specifically includes four measurement dimensions: direct trust, referral trust, feedback trust, and historical trust. Utilizing the diversity of the trust measurement dimensions, a hybrid trust evaluation scheme is implemented, which can not only measure the reliability of vehicle nodes but also measure the reliability of vehicle messages, and at the same time has a certain robustness against malicious attacks in the network.

[0106] The above are only the embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A multi-dimensional vehicle trust evaluation method for the vehicle networking, characterized in that: Establish a reputation system in the vehicular ad hoc network. The multi-dimensional vehicle trust evaluation method is implemented by calculating the reputation of vehicles through the reputation system, and specifically includes the following steps: Step 1: The trusted center TA is responsible for the initialization of the reputation system. The trusted center TA sends the vehicle global reputation score calculated in the most recent time to each fog server acting as a local trusted center, and the trusted center TA registers the vehicles newly joined to the vehicular ad hoc network; Step 2: The vehicle nodes in the vehicular ad hoc network periodically broadcast beacon messages. When a traffic event occurs, the vehicle nodes located near the event occurrence location will first sense the traffic event, and then the vehicle nodes located near the event occurrence location will broadcast the sensed event. The vehicle nodes located farther away from the event occurrence location learn about the traffic event by receiving the messages. Among them, the beacon message is the context information of the vehicle node, including the vehicle ID, speed, location, and current time; Step 3: The vehicle that receives the message calculates the trust value between vehicles for the vehicle that sends the message. The vehicle performing the trust calculation is called the evaluating vehicle V i , and the vehicle to be evaluated is called the target vehicle V j . The evaluating vehicle V i calculates the direct trust DT j of the target vehicle V i,j based on the context relevance and message quality, and then calculates the recommended trust RT j based on the recommendation score of the neighboring vehicles for the target vehicle V i,j . Finally, the latest inter-vehicle trust value TrV j is calculated by combining the historical interaction status of the target vehicle V i,j ; Step 4: Evaluate vehicle V i Based on the calculated in-vehicle trust value TrV i,j Make a message decision. If the in-vehicle trust value TrV i,j is less than the set threshold, it means that the evaluated vehicle V i does not trust the event message reported by the target vehicle V j and then discards the message sent by the target vehicle V j Otherwise, receive or forward the message, and finally feedback the satisfaction of the interaction with the target vehicle V j to the roadside unit RSU; Step 5: The roadside unit (RSU) collects feedback on the target vehicle V j , first calculates the aggregated feedback score, filters malicious feedback using the overall difference, then converts the filtered feedback score into a pair of positive and negative feedbacks, and calculates the feedback trust FT of the target vehicle V j using an asymmetric forgetting factor j . Finally, the feedback trust of the target vehicle V j is sent to the local fog server; Step 6: The fog server sends multiple roadside units RSU to the target vehicle V j Feedback Trust FT j and the target vehicle V j The target vehicle V is calculated by combining the historical reputation value j The latest global reputation value, if the target vehicle V j If the global reputation value of a vehicle is lower than the threshold, it is identified as a malicious vehicle and put into the blacklist. The latest global reputation value and blacklist are uploaded to the cloud for storage, and then the blacklist is sent to the roadside unit RSU within the jurisdiction, and the roadside unit RSU broadcasts the blacklist to vehicles within the communication range.

2. The multi-dimensional vehicle trust evaluation method for the Internet of Vehicles according to claim 1, wherein: The specific steps of Step 3 include the following steps: Step 3-1: Evaluate vehicle V i Extract the context information of the target vehicle V j from the beacon message of the target vehicle V j and process the context information into an n-dimensional normalized vector vector i 、vector j ,Evaluate vehicle V i Calculate the context relevance degree to the target vehicle V j : Among them respectively represent the k-th value in the normalized vector, n represents the dimension of the normalized vectors vector i and vector j , i represents the evaluation vehicle V i , j represents the target vehicle V j ; Step 3-2: When the vehicle receiving the message calculates the inter-vehicle trust value of the vehicle sending the message, a method of calculating the trust value regularly is adopted, introducing a time interval, i.e., a time slot, and calculating the trust value in each time slot. Since the messages in the direct interaction process can be decrypted and analyzed by the receiver, vehicle V i can determine the legitimacy of the data message sent by the target vehicle V j . Vehicle V i counts the number of correct messages legal received from the target vehicle V j during this time slot i,j and the number of incorrect messages mal i,j , and calculates the message quality of the target vehicle V j : Among them, variable parameters Tends to 1 as the number of correct messages increases, and the target vehicle V j Improves its own message quality by sending more correct messages; Step 3-2: Use the context relevance as the weight of the message quality of the target vehicle V j to adjust the accuracy of the message quality, and use the Gompertz function to calculate the evaluation of the direct trust of vehicle V i for the target vehicle V j : Where b and c are function parameters. b controls the displacement of the output along the x-axis, and c adjusts the growth rate of the function. The output range of the function is between 0 and 1; Step 3-4: Evaluate vehicle V i Based on DT i,j Give the recommended score rs j for the target vehicle V i,j , the higher the score, the better the performance of the target vehicle V j , and the more worthy of recommendation. If the evaluated vehicle V i is a malicious vehicle, it will tamper with the recommended score, give a lower or higher score, and then share it with surrounding vehicles, thus affecting the trust evaluation of surrounding vehicles for the target vehicle V j ; Step 3-5: Evaluate vehicle V i Obtain neighboring vehicle V i,k For the target vehicle V j The recommended score rs k,j , and select TrV from it i,k Recommendations of neighboring vehicles with a TrV greater than the threshold, that is, trustworthy neighbors. Initially, adopt the suggestions of all neighbors. After establishing trust later, only adopt the recommendations of trusted neighbors. Denote the set of neighboring vehicles as Take the evaluated vehicle V i For the neighboring vehicle V i,k The direct trust DT i,k As the weight, calculate the recommended trust: Step 3-6: Evaluate vehicle V i Based on the direct trust and recommended trust obtained in this time slot, and taking into account the historical interaction behavior of the target vehicle V j calculate the latest in-vehicle trust value for the target vehicle V j as follows: Among them represents the historical in-vehicle trust value of the target vehicle V j where λ is between 0 and 1 and is the proportion it occupies. Since more weight should be given to the recent vehicle behavior, λ is greater than 0.5, and n is the number of interactions between the two vehicles.

3. The multi-dimensional vehicle trust evaluation method for the Internet of Vehicles according to claim 2, characterized in that: The specific steps of Step 4 are as follows: Step 4-1: Evaluate vehicle V i Based on the latest workshop trust value TrV i,j Make a message decision. If the value of the latest workshop trust value TrV i,j is less than the set threshold, it means that the evaluated vehicle V i does not trust the event message reported by the target vehicle V j and then discard the message sent by the target vehicle V j Otherwise, receive or forward the message; Step 4-2: Evaluate vehicle V i According to the latest workshop trust value TrV i,j Give the satisfaction feedback score fs j for the target vehicle V i,j , and send the feedback to the roadside unit RSU. The higher the feedback score, the higher the satisfaction of the evaluating vehicle V i with the target vehicle V j . If the evaluating vehicle V i is a malicious vehicle, it will tamper with the feedback score and then send it to the roadside unit RSU.

4. A multi-dimensional vehicle trust evaluation method for the Internet of Vehicles according to claim 3, characterized in that: The specific steps of Step 5 include the following steps: Step 5-1: For the target vehicle V j The set of vehicles that give feedback is denoted as The roadside unit RSU obtains the historical reputation value of the feedback vehicle V x from the fog server Calculate the aggregated feedback score for the target vehicle V j : Then calculate the feedback vehicle V x with respect to the target vehicle V j for the feedback difference value Step 5-2: The communication between vehicles is many-to-many. The feedback vehicle V x In addition to generating feedback for the target vehicle V j It will also give feedback to the vehicles that have interacted. Malicious vehicles in the network will initiate switch attacks. The roadside unit RSU needs to consider the overall feedback difference of the feedback vehicle V x and then decide whether to adopt the feedback of the feedback vehicle V x The vehicles that have been given feedback by V x The set is denoted as E x The total number of vehicles in it is n x , calculate the overall feedback difference value of the feedback vehicle V x : If Diff x is greater than the set threshold, the feedback sent by the feedback vehicle V x will not be adopted, and it will be deleted from T j ; Step 5-3: For the roadside unit RSU converts its feedback value fs x,j into a pair of positive and negative feedbacks, denoted as <α x , β x . Among them, α x is the positive feedback representing the satisfaction degree, that is, α x = fs x,j ; β x is the negative feedback representing the dissatisfaction degree, that is, β x = 1 - fs x,j . By adding a forgetting factor, the recent feedback has a greater weight than the old feedback. Taking the historical reputation value of the vehicle as the weight of the feedback, the positive feedback α x and the negative feedback β x are: where ζ1 and ζ2 are asymmetric forgetting factors, ζ1 < ζ2, such that malicious behavior of the vehicle is more severely punished, t is the current time, and t x represents the time when the feedback is sent, and positive and negative feedback calculations are used to target vehicle V j Feedback trust: Then send the feedback trust for the target vehicle V j to the local fog server.

5. The multi-dimensional vehicle trust evaluation method for the vehicle networking according to claim 4, characterized in that: The specific steps of Step 6 include the following steps: Step 6-1: The fog server combines the feedback trust of multiple roadside units (RSUs) for the target vehicle V j and the historical reputation value RepV of the target vehicle V j to calculate the latest global reputation value of the target vehicle V j last : j ​ RepV j = λRepV j last +(1 - λ)AVG(FT j ) where λ is between 0 and 1, and λ greater than 0.5 is taken to assign more weight to the latest feedback trust, and AVG(FT j ) represents the average value of the feedback trust of multiple RSUs for the target vehicle V j ; Step 6-2: If the target vehicle V j has a reputation value lower than the threshold, it is identified as a malicious vehicle and added to the blacklist. The latest global reputation value and the blacklist are uploaded to the cloud for storage. Subsequently, the blacklist is sent to the roadside units RSU within the jurisdiction, and the roadside units RSU broadcast the blacklist to the vehicles within the communication range.

6. A multi-dimensional vehicle trust evaluation method for the Internet of Vehicles according to any one of claims 1-5, characterized in that: The specific steps of Step 1 include the following steps: Step 1-1: The trusted center TA sends the reputation value of the vehicle node calculated in the most recent time to the corresponding fog server for localizing the evaluation of the vehicle's global reputation value; Step 1-2: The newly joined vehicle user needs to register with the trusted center TA using its real identity. TA assigns a uniquely identifiable identity ID to it and initializes the reputation value of the vehicle, and then sends the vehicle information to the corresponding local fog server.

Images