Vehicle crowdsourcing perception task privacy protection method and device

Abstract

The vehicle crowdsourcing perception task privacy protection method provided by the application comprises the following steps: a task requester sends encrypted content C containing public information and a preset location confusion parameter and a task demand parameter to a service provider, so that the task can be safely transmitted to all vehicles based on the service provider and the encrypted information; secondly, the vehicle side filters candidate vehicles based on the same preset function and the own parameter information of the target vehicle, and performs problem optimization modeling under Markov decision based on the candidate vehicles, and solves the selected vehicles by using the problem optimization model, thereby improving the distribution efficiency and the privacy protection of the crowdsourcing perception task. Since the encryption and decryption processes are based on cryptography calculation, compared with the blockchain technology, the calculation cost is greatly saved, and under the action of the location confusion parameter in the method, the service provider can accurately solve the problem optimization model according to the relative position of the candidate vehicle, thereby improving the overall utility of the perception task.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of Internet of Vehicles security, and particularly relates to a vehicle crowdsourcing perception task privacy protection method and device. BACKGROUND

[0002] The core idea of vehicle crowdsensing is to outsource perception tasks to independent vehicles, which can collect data by carrying various sensors. For example, by equipping embedded cameras and radar sensors, vehicles can assist in achieving various perception tasks. For example: a prominent example is that the prior art (M.Y. Alam, A. Nandi, and A. Kumar, Crowdsourcing from the True crowd: Device, vehicle, road-surface and driving independent road profiling from smartphone sensors, Pervasive and Mobile Computing, vol. 61, pp. 101103-101123, 2020) proposes to use sensor data carried by vehicles to construct a 3D map by aggregating video information. Compared with the traditional perception task execution mode, the vehicle crowdsensing mechanism no longer relies on specific task executors, but relies on independent vehicles, thereby effectively reducing the cost of the entire task. In such a mechanism, each perception task is initiated by a requester and organized and managed by a third-party service provider. Vehicle crowdsensing technology has wide application prospects in practical applications. For example, in the field of smart cities, vehicle crowdsensing can be used to monitor traffic flow, road conditions and parking conditions in real time, thereby helping city planners better optimize traffic management. In the logistics industry, through vehicle crowdsensing, real-time monitoring and tracking of the transportation process of goods can be achieved, improving the efficiency and visual management of logistics transportation. In addition, in the field of environmental monitoring and disaster warning, vehicle crowdsensing can also play an important role, such as collecting environmental data through vehicle-mounted sensors to achieve real-time monitoring and early warning of natural disasters, air quality, etc. In summary, vehicle crowdsensing has wide application prospects and important application value in the fields of smart cities, logistics transportation and environmental monitoring.

[0003] In the application of vehicle crowd sensing, in order to maximize the effectiveness of crowd sensing tasks, vehicles need to upload their location information and submit it to service providers for task assignment. However, this process carries serious privacy risks. When there are potential attackers or untrusted service providers, these attackers or service providers may illegally obtain and leak this sensitive data. Furthermore, this sensitive data may be combined with the vehicle's identity, leading to further personal privacy leaks. Therefore, an innovative, efficient, and privacy-preserving crowd sensing mechanism is urgently needed to maximize data privacy protection and avoid directly exposing data to service providers.

[0004] The prior art also proposes several privacy protection schemes for crowdsourcing and crowdsensing problems. For example: technical scheme one (J. Zhang, F. Yang, and Z. Ma, A Decentralized Location Privacy Preserving Spatial Crowdsourcing for Internet of Vehicles, IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 4, pp. 2299-2313, 2020) proposes a decentralized location privacy protection scheme based on blockchain technology. This scheme one realizes the secure management of vehicle owner data and the multi-level privacy of work sites, but in actual scenarios, scheme one may be limited by the high implementation cost of blockchain. Technical scheme two (Y. Jiang, K. Zhang, and Y. Qian, Preserving Location Privacy and Accurate Task Allocation in Edge-assisted Mobile Crowdsensing, in IEEE Wireless Communications and Networking Conference. IEEE, 2022, pp. 704-70) proposes an edge-assisted mobile crowdsensing protocol that cooperatively protects location privacy by deploying homomorphic encryption in task allocation. However, the calculation and communication overhead in the encryption algorithm of scheme two introduces relatively high costs. Technical scheme three (L. Wang, D. Zhang, and D. Yang, Sparse Mobile Crowdsensing With Differential and Distortion Location Privacy, IEEE Transactions on Information Forensics and Security, vol. 15, pp. 2735-2749, 2020) proposes a privacy strategy that combines differential privacy and obfuscation privacy for location mixing, but scheme three may result in inaccurate vehicle location information due to the introduction of additional noise, although it achieves good privacy protection, it also reduces the overall utility of the sensing task.

[0005] Therefore, the existing vehicle crowdsensing privacy protection schemes have the problems of high implementation cost or low overall utility for sensing tasks. SUMMARY

[0006] In order to solve the above problems existing in the prior art, the application provides a vehicle crowdsourcing perception task privacy protection method and equipment.

[0007] The technical problem to be solved by the application is solved by the following technical scheme:

[0008] In a first aspect, the application provides a vehicle crowdsourcing perception task privacy protection method, comprising:

[0009] The task requester converts the preset task demand parameter into a bit sequence limit, and concatenates the bit sequence limit based on a preset function to obtain a symmetric key K s ;

[0010] The task requester encrypts the preset location confusion parameter using the symmetric key K s , obtains encrypted content C, and sends the encrypted content C, the bit sequence limit and public information to the service provider;

[0011] The service provider broadcasts the encrypted content C, the bit sequence limit and the preset function to the target vehicle;

[0012] The target vehicle generates a matching key that satisfies the bit sequence limit according to its own parameter information and the preset function Based on the matching key Decrypts the encrypted content C to obtain a decryption result, and obtains a candidate vehicle according to the decryption result; the decryption result contains the preset location confusion parameter;

[0013] The candidate vehicle confuses its coordinate data according to the preset location confusion parameter to obtain confused location information, and sends the confused location information to the service provider;

[0014] The service provider performs optimization modeling processing based on the public information and the confused location information to generate a problem optimization model, optimizes and solves the problem optimization model to obtain a selected vehicle, and uses a proxy re-encryption strategy and the public information to allocate the task privacy to the selected vehicle; the problem optimization model is constructed based on a Markov decision.

[0015] Optionally, the task demand parameter includes a longitude range, a latitude range and a vehicle credit range.

[0016] Optionally, the preset task demand parameter is converted into a bit sequence limit, and the bit sequence limit is concatenated based on a preset function to obtain a symmetric key K s , comprising:

[0017] The task demand parameter is converted into a bit sequence limit according to a binary forcing algorithm;

[0018] The bit sequence limit is concatenated based on a hash function to obtain a symmetric key Ks .

[0019] Optionally, a symmetric key K s The preset position confusion parameter is encrypted to obtain encrypted content C, including:

[0020] The AES encryption algorithm and the symmetric key K s The preset position confusion parameter is encrypted to obtain encrypted content C.

[0021] Optionally, the self parameter information of the target vehicle includes: coordinates of the target vehicle and a reputation value of the target vehicle.

[0022] Optionally, the matching key The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result, including:

[0023] The matching key The encrypted content C is decrypted to obtain a decryption result.

[0024] The matching key The corresponding target vehicle is taken as a candidate vehicle.

[0025] Optionally, the public information includes: data collection demand Perception area Budget vehicle number And completion time

[0026] The problem optimization model is represented as:

[0027]

[0028] Wherein, The problem optimization model is represented as: U n The utility of the candidate vehicle n is represented as: θ n The selection factor of the candidate vehicle n is represented as: N c The total number of candidate vehicles is represented as: s.t.1 represents a constraint condition 1, s.t.2 represents a constraint condition 2, s.t.3 represents a constraint condition 3, d represents the distance between any candidate vehicles, and D min The distance threshold between selected vehicles is represented as: Any.

[0029] Optionally, the utility U n The utility of the candidate vehicle n is represented as:

[0030]

[0031] Wherein, U0 represents basic utility, -Δ represents the punishment of completing the task failure, t n represents the time required for the candidate vehicle n to complete the task;

[0032]

[0033] Wherein, p n represents the data collection speed of the candidate vehicle n;

[0034] The distance threshold D between the selected vehicles min represents:

[0035]

[0036] Wherein, R represents the distance from the center of the hexagon to the vertex under the honeycomb conjecture.

[0037] Optionally, the selected vehicle is obtained by optimizing and solving the problem optimization model, comprising:

[0038] The selected vehicle is obtained by optimizing and solving the optimization model based on deep reinforcement learning.

[0039] In a second aspect, the present application provides a vehicle crowd-sourced perception task privacy protection device, comprising: a processor, a storage medium and a bus, the storage medium stores machine readable instructions executable by the processor, when the vehicle crowd-sourced perception task privacy protection device is running, the processor and the storage medium communicate through the bus, and the processor executes the machine readable instructions to perform the steps of the vehicle crowd-sourced perception task privacy protection method as described above.

[0040] The present application provides a vehicle crowd-sourced perception task privacy protection method and device. One vehicle crowd-sourced perception task privacy protection method comprises: a task requester converts a preset task demand parameter into a bit sequence limit, and obtains a symmetric key K s based on a preset function by cascading processing the bit sequence limit; the task requester encrypts a preset location confusion parameter using the symmetric key K s to obtain encrypted content C; the task requester sends the encrypted content C, the bit sequence limit and public information to a service provider; the service provider broadcasts the encrypted content C, the bit sequence limit and the preset function to a target vehicle; the target vehicle generates a matching key K based on the matching key K The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result; the decryption result contains a preset location confusion parameter; the candidate vehicle performs confusion processing on own coordinate data according to the preset location confusion parameter to obtain confusion location information; the confusion location information is sent to a service provider; the service provider performs optimization modeling processing based on public information and the confusion location information to generate a problem optimization model; the problem optimization model is solved to obtain a selected vehicle; the proxy re-encryption strategy and the public information are used to perform task privacy allocation for the selected vehicle; and the problem optimization model is constructed based on a Markov decision. In the present application, the task requester sends the public information and the encrypted content C containing the task demand parameter and the preset location confusion parameter to the service provider, so that the task can be safely transmitted to all vehicles based on the service provider and the encrypted information; secondly, the vehicle side performs screening of the candidate vehicle based on the same preset function and the own parameter information of the target vehicle, and performs problem optimization modeling under the Markov decision based on the candidate vehicle, and uses the problem optimization model to solve the selected vehicle, thereby improving the allocation efficiency and the privacy protection of the crowdsourcing perception task. Since the encryption and decryption processes of the present application scheme are based on cryptographic calculations, compared with the blockchain technology, the calculation cost is greatly saved, and under the action of the location confusion parameter in the present application method, the service provider cannot obtain the absolute location information of the candidate vehicle, but can accurately solve the problem optimization model according to the relative position of the candidate vehicle, not only protecting the location privacy of the vehicle but also avoiding the noise processing of the vehicle location, thereby improving the overall utility of the perception task.

[0041] The present application will be further described in detail below in combination with the drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 A flowchart of a vehicle crowdsourcing perception task privacy protection method provided by an embodiment of the present application;

[0043] Figure 2 A vehicle crowdsourcing model block diagram provided by an embodiment of the present application;

[0044] Figure 3 A flowchart of a privacy protection task publishing process provided by an embodiment of the present application;

[0045] Figure 4 A diagram for solving a problem optimization model based on a deep Q network provided by an embodiment of the present application;

[0046] Figure 5 A relationship diagram between task allocation utility and budget vehicle quantity based on comparison between the present application method and other methods for different task scales provided by an embodiment of the present application;

[0047] ​Figure 6 A relationship diagram of the utility of task allocation and the number of budget vehicles under different vehicle reputation ranges provided for an embodiment of the present application

[0048] Figure 7 A structural schematic diagram of a vehicle crowdsourcing perception task privacy protection device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0049] The present application will be further described in detail below in conjunction with specific embodiments, but the embodiments of the present application are not limited thereto.

[0050] In order to realize the privacy of crowdsourcing perception tasks while reducing the computing cost and improving the overall utility of perception tasks, an embodiment of the present application provides a vehicle crowdsourcing perception task privacy protection method. Figure 1 A flowchart of a vehicle crowdsourcing perception task privacy protection method provided for an embodiment of the present application, as shown in Figure 1 , includes:

[0051] S101, the task requester converts the preset task demand parameter into a bit sequence limit, and obtains a symmetric key K based on a preset function by cascading processing the bit sequence limit s .

[0052] Figure 2 A vehicle crowdsourcing model block diagram provided for an embodiment of the present application, the present embodiment relates to three entities: a task requester, a service provider and a vehicle.

[0053] Task requester: in the present embodiment, the task requester refers to various institutions or organizations, such as companies, government departments or urban planning agencies, which need to collect perception data to improve map data or urban traffic planning. These task requesters usually have limited resources and budget, so they entrust the perception task to the service provider so that the service provider can execute the task.

[0054] Service provider: in the present embodiment, the service provider refers to an organization or platform with strong computing and storage resources, as well as high-speed data transmission capability. They are responsible for processing perception tasks and allocating tasks according to the performance, geographical location and other factors of vehicles. Once receiving the sensor report of the vehicle, the service provider returns the report to the task requester and issues rewards to the vehicle on behalf of the task requester. As a third-party organization, the service provider needs to consider privacy issues to ensure that the processing and transmission of data comply with relevant laws and regulations and privacy protection requirements.

[0055] ​Vehicles: In this embodiment, vehicles are considered mobile platforms with communication and storage capabilities. They are typically equipped with onboard cameras and can participate in the data collection, processing, and storage of perception tasks. When several vehicles are willing to perform a task, they compete with each other, and vehicles with better performance and a more suitable location are more likely to be selected to perform the task.

[0056] like Figure 2 As shown, the overall implementation can be divided into the following steps, in the order of 1-8. First, the task requester determines the target area and task parameters for crowd sensing. Then, the service provider broadcasts the task to the vehicle. The vehicle assumes the task requirements and conducts task competition to report vehicle attributes. The service provider selects the vehicle as a candidate vehicle and assigns the task. After the perception process, the perception data is uploaded by the vehicle and aggregated by the service provider. Once the sensing data is received, the task requester will send a reward to the vehicle. The entire crowd sensing can be divided into three stages: task requirements and task release (1-2) represent the task release stage, while task competition, vehicle selection and task allocation (3-5) represent the task allocation stage, and uploading reports, aggregating reports and issuing rewards (6-8) represent the data aggregation and reward allocation stage. The embodiment of the present invention mainly focuses on the task release and task allocation stages, and data aggregation and reward allocation are performed using existing technologies.

[0057] Optionally, the mission requirement parameters include: longitude range, latitude range, and vehicle credibility range.

[0058] Optionally, S101 may specifically include: converting the task requirement parameters into bit sequence constraints according to the binary forcing algorithm; performing cascade processing on the bit sequence constraints based on the hash function to obtain the symmetric key K s .

[0059] In the embodiment of the present invention, the bit sequence restriction includes: latitude range bit restriction acc x , longitude range bit limit acc y And the vehicle credit range bit limit acc r .

[0060] For the longitude range, latitude range and vehicle reputation range, the embodiment of the present invention uses the binary addition method to convert the task requirement parameters into the number of bits, which is expressed as the number of latitude bits bit x , number of longitude bits bit y and vehicle reputation bit number bit r Specifically, we recursively calculate the bit x The next digit of the digit, repeatedly divides a new interval into two equal parts, and finally, uses the hash function to cascade bit x , bit y and bit rto calculate the symmetric key K s . The generated K s includes the task requirements, which can be used to filter available vehicles.

[0061] Table 1 gives an example of encoding using different vehicle reputation ranges r (such as 78-83), for example: for reputation r1 = 78, since the reputation distribution range is 0 to 100, [L r , U r ] = [0, 100], L r represents the lower bound of the vehicle reputation approximation, and U r represents the upper bound of the vehicle reputation approximation. In the bisection, since (0+100) / 2 = 50, and 78 is greater than 50, bit r = bit r ||1, the first bit sequence is 1, and L r = 50, and U r is unchanged, and so on. Since acc r = 3, Table 1 shows that the first 3 bits are the same, but the 4th bit is different. Therefore, only vehicles with reputation in the range can generate the correct key. The same applies to the calculation of latitude and longitude, so finally only vehicles with reputation in the range and in the task area can decrypt.

[0062] Table 1. Bit sequence restriction example using vehicle reputation range r (78-83) for encoding

[0063] Binary bit sequence 1 2 3 4 L 0 50 75 75 Mid 50 75 87.5 81.25 U 100 100 100 87.5 [r1 = 78] 1 1 0 0 [r2 = 83] 1 1 0 1

[0064] S102, the task requester encrypts the preset location obfuscation parameter using the symmetric key K s to obtain encrypted content C; sends the encrypted content C, the bit sequence restriction, and the public information to the service provider.

[0065] Optionally, the symmetric key K s is used to encrypt the preset location obfuscation parameter to obtain encrypted content C, including:

[0066] The AES encryption algorithm and the symmetric key K s are used to encrypt the preset location obfuscation parameter to obtain encrypted content C.

[0067] S103, the service provider broadcasts the encrypted content C, the bit sequence restriction, and the preset function to the target vehicle.

[0068] S104, the target vehicle generates a matching key that satisfies the bit sequence restriction according to its own parameter information and the preset function based on the matching key The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result.

[0069] Optionally, the self parameter information of the target vehicle includes: coordinates of the target vehicle and a reputation value of the target vehicle.

[0070] Optionally, the matching key is obtained based on the matching key The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result, comprising:

[0071] The matching key is obtained based on the matching key The encrypted content C is decrypted to obtain a decryption result.

[0072] The matching key that can correctly decrypt the encrypted content C in the decryption result is obtained The corresponding target vehicle is taken as the candidate vehicle.

[0073] S105, the candidate vehicle performs confusion processing on the self coordinate data based on the preset position confusion parameter, to obtain confusion position information; and sends the confusion position information to the service provider.

[0074] Figure 3 The flowchart of the privacy protection task publishing process provided by the embodiment of the application is shown in FIG. 1. Figure 3 As shown in FIG. 1, the task requester (party) first calculates latitude range bit limit acc x , longitude range bit limit acc y and vehicle reputation range bit limit acc r based on the task demand parameter under the action of the binary search algorithm. Then, based on acc x , acc y , acc r , a preset hash function KeyGen(·) is used to obtain a symmetric key K s . The encrypted content C is calculated by using the AES encryption algorithm based on K s and preset position confusion parameters (m x , m y ) to realize the protection of the subsequent vehicle position. Finally, KeyGen(·), acc x , acc y , acc r , the encrypted content C and public information are sent to the service provider (party).

[0075] Service provider: the service provider propagates part of the obtained information, that is, C, acc x , acc y , acc rKeyGen(·) sends to possible vehicles (target vehicles).

[0076] The target vehicles generate corresponding accurate keys based on KeyGen(·) and their coordinates and reputation values Since K s is generated according to the area and reputation range of the required task, only the vehicles within the area and meeting the reputation requirement can calculate and correctly decrypt C. The vehicles capable of correctly decrypting C are taken as candidate vehicles. The decrypted C contains preset position confusion parameters (m x ,m y ), and then a preset confusion strategy is applied to confuse the coordinates of the absolute positions of the candidate vehicles. For example, for any candidate vehicle absolute position coordinates (x, y), the relative position coordinates (x i ,y i ) can be calculated, x i =(x+m x )mod360-180 and y i =(y+m y )mod180-90.

[0077] Service provider: Since the position coordinates of the vehicles are confused, it is difficult for the service provider to obtain the absolute positions of all the candidate vehicles, but the accurate relative positions are known. Then, the optimal candidate vehicle is selected by the service provider, and a vehicle list is sent to the task requester.

[0078] Task requester, service provider and vehicle: a proxy re-encryption PRE strategy is used to publish the task.

[0079] S106, the service provider performs optimization modeling processing based on public information and confused position information, generates a problem optimization model, performs optimization solving on the problem optimization model to obtain a selected vehicle, and performs task privacy allocation for the selected vehicle by using a proxy re-encryption strategy and the public information.

[0080] Optionally, the public information includes: data collection demand sensing area budget vehicle quantity and completion time

[0081] Based on the public information, each task can be represented as a tuple vehicle quantity is usually limited by a budget.

[0082] The problem optimization model is represented as:

[0083]

[0084] in, Represents the problem optimization model, U n represents the utility of candidate vehicle n, θ n represents the selection factor of candidate vehicle n, N c represents the total number of candidate vehicles, st1 represents constraint 1, st2 represents constraint 2, st3 represents constraint 3, d represents the distance between any candidate vehicles, D min represents the distance threshold between selected vehicles, Indicates any.

[0085] In the embodiment of the present invention, θ n As a selection factor, when vehicle n is selected, θ n =1, otherwise θ n = 0. The value of d can be calculated based on the relative position coordinates of the vehicles and the Euclidean distance.

[0086] Optionally, the utility U of candidate vehicle n n Expressed as:

[0087]

[0088] Among them, U0 represents the basic utility, -Δ represents the penalty for failure to complete the task, and t n represents the time required for candidate vehicle n to complete the task;

[0089]

[0090] Among them, ρ n represents the data collection speed of candidate vehicle n;

[0091] The distance threshold D between selected vehicles min Expressed as:

[0092]

[0093] Where R represents the distance from the center to the vertex of the hexagon under the honeycomb conjecture.

[0094] Optionally, optimizing and solving the problem optimization model to obtain the selected vehicle includes:

[0095] The optimization model is optimized and solved based on deep reinforcement learning to obtain the selected vehicle.

[0096] To simplify the solution process, the optimization problem can be expressed as a Markov decision process (MDP), and the system state space, action space, and reward function are defined. To solve the problem using deep reinforcement learning, the state space s(t) at time t is first defined as:

[0097]

[0098] where P n (t) represents the ambiguous position of candidate vehicle n at time t, p n (t) represents the corresponding data collection speed of candidate vehicle n at time t. represents the data collection requirement under s(t), represents the perception area under s(t), represents the budget vehicle number under s(t), represents the completion time under s(t).

[0099] In addition, the action space a(t) is also defined to make vehicle selection at each step, which is represented as:

[0100]

[0101] where θ n (t) = 1 represents that the candidate vehicle is selected. After performing the action a(t), the overall task utility is optimized by the immediate reward r(t), which is represented as:

[0102]

[0103] where U n is the utility of candidate vehicle n. The problem optimization model is constructed based on Markov decision.

[0104] A deep Q network (DQN) is used to develop a suitable decision strategy. In the traditional Q learning method, a Q table is used to store each possible state-action combination, and the Q values of these combinations are updated in each iteration. However, when the state and action space become very large, the Q table becomes very large, and a large amount of memory is required to store all these combinations. With the increase of state and action space, the size of the Q table presents an exponential growth. In order to overcome this problem, DQN is a deep learning-based reinforcement learning algorithm that can use a deep neural network to approximate the Q table. Figure 4 The schematic diagram of the problem optimization model based on deep Q network (DQN) provided by the embodiment of the present application for solving is shown in Fig. 1. Figure 4As shown in the figure, the DQN agent has two Q networks and an experience buffer. The current Q network has a network structure Q and parameters Θ. It receives the state of the environment as input and outputs the Q value of each action. The target Q network is also designed to stabilize the training process. It has the same network structure Q but different parameters Θ′. The experience buffer is used to store the training experience tuples (s(t), a(t), r(t), s(t+1)) obtained by the interaction between the agent and the environment. After a certain number of iterations, a small batch of experience is randomly sampled from the experience replay buffer and used to update the target Q network. For the state-action pair (s, a), the loss function is defined as:

[0105] L(Θ)=E[(Q target -Q(s,a,Θ)) 2 ];

[0106] Q target =r+γmax a′ Q(s′,a′,Θ′), where r is the reward and γ is the discount factor. Q(s′,a′,Θ′) represents the target Q-network, s′, a′, and Θ′ represent the state, action, and parameters of the target Q-network, respectively. Q(s,a,Θ) represents the current Q-network, where s, a, and Θ represent the state, action, and parameters of the current Q-network, respectively. The Q-network parameters Θ are updated using gradient descent.

[0107] In order to verify the performance of the privacy protection method for the vehicle crowdsourcing perception task provided by the embodiment of the present invention, the following simulation experiments were also conducted:

[0108] like Figures 5 to 6 As shown in Figure 1, in order to evaluate the overall task utility, the proposed privacy protection method for vehicle crowdsourcing perception tasks was experimentally evaluated, and the results were simulated on the corresponding devices using PyTorch and Python 3.7. In the task allocation process, two perception areas ranging from 100 to 10,000 square kilometers were first defined. Contains 100 and 1000 vehicles respectively. Data collection requirements for crowdsourcing perception tasks Defined in [1.5, 3.5]Gb, completion time Within {0.2, 0.5, 0.7} hours. Then, the proposed method is compared with existing task allocation algorithms, such as Soft Actor Critic (SAC), greedy algorithm and random algorithm.

[0109] (1) SAC: For each task, the actor network learns a policy for selecting a vehicle, while the critic network evaluates how well the actor performs a specific action.

[0110] (2) Greedy algorithm: For each task, the service provider always prioritizes the vehicle with the highest task execution efficiency and calculates the task utility accordingly.

[0111] (3) Randomized algorithm: For each task, the service provider randomly selects a vehicle for task offloading and calculates the task utility.

[0112] Figure 5 The utility of task allocation and the number of budgeted vehicles based on the method of the present invention and other methods based on different task scales provided by the embodiment of the present invention are compared. relationship diagram. Figure 5 (a) shows the task allocation utility of 100 vehicles in 100 square kilometers, while Figure 5 Figure (b) shows the task allocation utility of 1,000 vehicles in 10,000 square kilometers.

[0113] like Figure 5 As shown, the allocation method proposed in the embodiment of the present invention is for all The highest utility is achieved, while the random selection algorithm is the worst. This is because the optimization goal of the method of the present invention is designed for long-term rewards, while the greedy algorithm only focuses on the current reward. The SAC algorithm is not suitable for discrete reinforcement learning problems such as task allocation, resulting in a low utility for task allocation. In addition, Figure 5 In (a), when the required number of vehicles is 14, the utility decreases significantly because the vehicle distribution does not meet the constraints.

[0114] Figure 6 The utility of task allocation and the budgeted number of vehicles under different vehicle reputation ranges provided by the embodiment of the present invention relationship diagram. Figure 6 Figure 2 shows that when the reputation range is set to (55, 65), the proposed method achieves the highest task utility, especially when the number of vehicles is greater than 9. This is because a wider reputation range can contribute more available candidates, so the DQN model can further optimize the overall task utility. However, a wider reputation range also brings higher costs to the task requester, so there is a trade-off between overall task utility and reputation range.

[0115] In this embodiment of the present invention, a reputation-based symmetric key generation algorithm is proposed for publishing privacy-preserving tasks. The DQN model is then used to optimize the overall task utility of fuzzy location data. Simulation results demonstrate that this method can effectively improve the overall utility of task allocation without leaking sensitive data to service providers.

[0116] The vehicle crowdsourcing perception task privacy protection method provided by the embodiment of the application not only protects the privacy of the task requester and the vehicle, ensures the safe transmission and processing of sensitive information, and avoids the potential key leakage risk. Through the privacy protection task allocation protocol, the vehicle can safely obtain the task content on the premise of protecting the personal privacy, thereby improving the security of information transmission. Secondly, the task allocation algorithm based on deep reinforcement learning can not only optimize the task allocation utility, but also improve the overall system performance while protecting the privacy of the task performer, thereby realizing the dual goals of privacy protection and allocation utility optimization.

[0117] The embodiment of the application provides a vehicle crowdsourcing perception task privacy protection method, which comprises the following steps: a task requester converts preset task demand parameters into bit sequence restrictions, and obtains a symmetric key K s by performing cascade processing on the bit sequence restrictions based on a preset function; the task requester encrypts preset location confusion parameters by using the symmetric key K s to obtain encrypted content C; the task requester sends the encrypted content C, the bit sequence restrictions and public information to a service provider; the service provider broadcasts the encrypted content C, the bit sequence restrictions and the preset function to a target vehicle; and the target vehicle generates a matching key K that meets the bit sequence restrictions according to self parameter information and the preset function. The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result; the decryption result contains a preset location confusion parameter; the candidate vehicle performs confusion processing on own coordinate data according to the preset location confusion parameter to obtain confusion location information; the confusion location information is sent to a service provider; the service provider performs optimization modeling processing based on public information and the confusion location information to generate a problem optimization model; the problem optimization model is solved to obtain a selected vehicle; the proxy re-encryption strategy and the public information are used to perform task privacy allocation for the selected vehicle; and the problem optimization model is constructed based on a Markov decision. In the embodiment of the present application, the task requester sends the public information and the encrypted content C containing the task demand parameter and the preset location confusion parameter to the service provider, so that the task can be safely transmitted to all vehicles based on the service provider and the encrypted information; secondly, the vehicle side performs screening of the candidate vehicle based on the same preset function and the own parameter information of the target vehicle, and performs problem optimization modeling under the Markov decision based on the candidate vehicle, and uses the problem optimization model to solve the selected vehicle, thereby improving the allocation efficiency and the privacy protection of the crowdsourcing perception task. Since the encryption and decryption processes of the present application scheme are based on cryptographic calculations, compared with the blockchain technology, the calculation cost is greatly saved, and under the action of the location confusion parameter of the present application method, the service provider cannot obtain the absolute location information of the candidate vehicle, but can accurately solve the problem optimization model according to the relative position of the candidate vehicle, not only protecting the location privacy of the vehicle but also avoiding the noise processing of the vehicle location, thereby improving the overall utility of the perception task.

[0118] The method provided by the embodiment of the present application can be applied to an electronic device. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc., and the embodiment of the present application is not limited thereto.

[0119] Based on the same inventive concept, the embodiment of the present application further provides a vehicle crowdsourcing perception task privacy protection device. Figure 7 The structure diagram of the vehicle crowdsourcing perception task privacy protection device provided by the embodiment of the present application includes a processor 710, a storage medium 720 and a bus 730. The storage medium 720 stores machine readable instructions executable by the processor 710. When the vehicle crowdsourcing perception task privacy protection device is running, the processor 710 communicates with the storage medium 720 through the bus 730. The processor 710 executes the machine readable instructions to perform the steps of the above method embodiment. The specific implementation manner and technical effects are similar, and will not be repeated here.

[0120] The storage medium can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk storage. Optionally, the storage medium can also be at least one storage device located remotely from the aforementioned processor.

[0121] The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc. It can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic component, a discrete hardware component.

[0122] It should be noted that the terms "first", "second", and the like, are used to distinguish similar objects, and do not necessarily have to be used to describe a particular order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all the embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application.

[0123] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. The illustrative description of the above terms in the present specification does not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described can be combined in any appropriate manner in one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in the present specification.

[0124] Although the present application has been described in connection with certain embodiments, persons skilled in the art will understand and appreciate that many modifications can be made to the described embodiments and these modifications can be made without departing from the scope of the application in its broader aspects. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude a plurality, and the

[0125] The above description is further detailed in connection with specific preferred embodiments of the application, and it is not to be construed that the specific implementation of the application is limited to these descriptions. For those skilled in the art, without departing from the concept of the application, a number of simple deductions or substitutions can be made, and all of these should be considered as falling within the scope of protection of the application.

Claims

1. A privacy protection method for a vehicle crowdsourcing perception task, characterized in that, The method comprises the following steps: The task requester converts preset task demand parameters into a bit sequence limit, and concatenates the bit sequence limit based on a preset function to obtain a symmetric key K s ; The task requester uses the symmetric key K s The preset location confusion parameter is encrypted to obtain encrypted content C; the encrypted content C, the bit sequence restriction, and public information are sent to a service provider; The service provider broadcasts the encrypted content C, the bit sequence limit and the preset function to the target vehicle; The target vehicle generates a matching key satisfying the bit sequence restriction according to self parameter information and the preset function Based on the matching key Decrypt the encrypted content C to obtain a decryption result, and obtain a candidate vehicle according to the decryption result; the decryption result contains the preset position confusion parameter The candidate vehicle performs obfuscation processing on the coordinate data of the candidate vehicle according to the preset position obfuscation parameter to obtain obfuscated position information; The obfuscated position information is sent to the service provider; The service provider performs optimization modeling processing based on the public information and the obfuscated position information to generate a problem optimization model; The problem optimization model is solved to obtain a selected vehicle; The proxy re-encryption strategy and the public information are used to perform task privacy allocation for the selected vehicle; The problem optimization model is constructed based on a Markov decision. 2.The method of claim 1, wherein, The task demand parameter comprises a longitude range, a latitude range and a vehicle reputation range. 3.The method of claim 1, wherein, The preset task requirement parameter is converted into a bit sequence limit, and the bit sequence limit is concatenated based on a preset function to obtain a symmetric key K s comprising: The task demand parameter is converted into the bit sequence limit according to a binary forcing algorithm; concatenating the bit sequence restriction based on a hash function to obtain the symmetric key K s . 4.The method of claim 1, wherein, The use of the symmetric key K s encrypting the preset location confusion parameter to obtain encrypted content C, comprising: An AES encryption algorithm and the symmetric key K are used s The preset location confusion parameter is encrypted to obtain the encrypted content C. 5.The method of claim 1, wherein, The self parameter information of the target vehicle comprises a coordinate of the target vehicle and a reputation value of the target vehicle. 6.The method of claim 1, wherein, The matching key is based on the encrypted content C The encrypted content C is decrypted to obtain a decryption result, and a candidate vehicle is obtained according to the decryption result. based on the matching key decrypting the encrypted content C to obtain a decryption result; The matching key in the decryption result that can correctly decrypt the encrypted content C The corresponding target vehicle is used as the candidate vehicle. 7.The method of claim 1, wherein, The common information includes: data collection requirements Awareness area Budget vehicle number And completion time The problem optimization model is represented as: wherein, represents a problem optimization model, U n represents the utility of candidate vehicle n, θ n represents the selection factor of candidate vehicle n, N c represents the total number of candidate vehicles, s.t. 1 represents constraint condition 1, s.t. 2 represents constraint condition 2, s.t. 3 represents constraint condition 3, d represents the distance between any candidate vehicles, D min represents the distance threshold between selected vehicles, represents any. 8.The method of claim 7, wherein, The utility U of the candidate vehicle n n is represented as: where U0represents a basic utility, -Δ represents a penalty for failing to complete the task, t n represents the time required for candidate vehicle n to complete the task; wherein p n denotes the data collection speed of the candidate vehicle n; The inter-vehicle distance threshold value D min is represented as: Wherein, R represents the distance from the center of the hexagon to the vertex under the hexagonal conjecture theorem. 9.The method of claim 1, wherein, The problem optimization model is solved to obtain a selected vehicle, comprising: The problem optimization model is solved based on deep reinforcement learning to obtain the selected vehicle.

10. A vehicle crowdsourcing perception task privacy protection device, characterized in that, The method comprises the following steps: A processor, a storage medium and a bus, the storage medium stores machine readable instructions executable by the processor, when the vehicle crowdsourcing perception task privacy protection device is running, the processor and the storage medium communicate through the bus, and the processor executes the machine readable instructions to execute the steps of the vehicle crowdsourcing perception task privacy protection method in any one of claims 1-9.

Images