VANET cooperative downloading method based on multi-dimensional screening and encrypted reverse auction

Abstract

The present invention provides a VANET cooperative download method based on multi-dimensional screening and encrypted reverse auctions, belonging to the field of incentive mechanisms for vehicle-mounted ad hoc networks. The present invention adds a network factor screening mechanism based on information entropy to the cooperative vehicle selection stage, innovatively integrating relevant knowledge such as information entropy and network latency. This method can screen out highly efficient cooperative vehicles with good network performance, improve the overall transaction quality and success rate of the vehicle-mounted ad hoc network system, and enhance the efficiency, security, and fairness of cooperative downloads in the vehicle-mounted ad hoc network. The present invention integrates a "sealed auction" through the design of a supplementary and optimized reverse auction mechanism, encrypting the entire transaction process and enhancing the security and fairness of the transaction. Furthermore, a data pricing limitation strategy is added to the data bidding stage of the reverse auction model, limiting data price fluctuations to a certain range and maximizing transaction stability and predictability.

Description

Technical Field

[0001] The present invention belongs to the field of vehicle-mounted ad hoc network incentive mechanism, and in particular relates to a VANET cooperative downloading method based on multi-dimensional screening and encrypted reverse auction. Background Art

[0002] Vehicular Ad Hoc Network (VANET), a cornerstone of Intelligent Transportation Systems (ITS), is an innovative communication technology that cleverly integrates with fixed infrastructure to enable autonomous networking and information exchange between vehicles. In a VANET, each vehicle node equipped with wireless communication equipment can function as both a router and a host, executing user-facing applications while also running routing protocols responsible for data forwarding and route maintenance. This flexible, self-organizing network enables efficient and stable information transmission even at high speeds or in environments with limited network coverage. Collaborative downloading within VANETs, ​​a core application of VANETs, ​​aims to address the download challenges associated with fluctuating network conditions and unstable signals while vehicles are in motion. Through direct communication and resource sharing between vehicles, collaborative downloading significantly improves the efficiency and speed of data downloads. When a vehicle needs to download a large amount of data, it can send a download request to surrounding vehicles. Vehicles receiving the request that possess the required data or are currently downloading the same data can act as seed nodes and transmit data blocks to the requesting vehicle. This distributed downloading approach not only reduces the download burden on individual vehicles but also accelerates the entire download process through parallel transmission. However, not all vehicles are ideal partners for collaborative downloads in VANETs. Because factors such as network conditions, download speeds, and driving routes vary among vehicles, screening of partner vehicles is crucial. The goal of selecting partner vehicles is to ensure efficient and stable downloads and avoid interruptions or speed drops due to network differences between vehicles. Scientific screening methods, such as those based on fuzzy mathematics or repeated game theory, can comprehensively evaluate multiple dimensions, such as vehicle network conditions, download capabilities, and driving trajectories, to select the most suitable partner. Partner vehicle selection not only impacts download efficiency but also has a profound impact on resource utilization and user experience. First, selected partner vehicles typically possess good network conditions and download speeds, significantly improving overall download efficiency and shortening download times. Second, a sound screening mechanism can avoid wasting network resources and ensure optimal resource allocation and utilization. Furthermore, efficient download speeds and a stable download process can significantly improve user satisfaction and experience, strengthening their trust and reliance on connected vehicle technology.

[0003] More importantly, the development of cooperative routing in VANETs and its cooperative vehicle selection mechanisms is of great significance for promoting the advancement of IoV technology and the construction of intelligent transportation systems. On the one hand, as a key application of IoV technology, the continuous optimization and improvement of cooperative routing will promote the overall advancement and popularization of IoV technology. On the other hand, intelligent transportation systems require efficient and stable data transmission as a foundation, and cooperative routing in VANETs and its selection mechanisms are one of the key technologies to achieve this goal. Therefore, in-depth research and development of cooperative routing in VANETs and its selection mechanisms are of great significance for building a more intelligent, efficient, and safe transportation system.

[0004] Regarding cooperative vehicle selection, Zhu Wenlong proposed a method for forming vehicle clusters in his paper "Research on Inter-Vehicle Data Distribution Based on Network Coding," applying network coding to the data distribution process. Clusters are formed by vehicles located close to each other on the road. Within the cluster, vehicles collaborate on downloads, utilizing the widely available mobile 2G / 3G network as a control channel. Vehicle-to-Infrastructure (V2I) communications within the cluster are achieved through roadside access points, achieving high data download rates. This improves the average download rate and shortens download times. Zeng Cheng proposed an efficient incentive method based on infinite repeated games in "Research on cooperative downloading problems in the Internet of Vehicles". This scheme screens the cooperative willingness of cooperative vehicles through autonomous judgment, and proves that as long as a certain discount factor is met, its incentive system can meet the interests of most users, and they will have enough motivation to participate in cooperative downloading, thus achieving the efficiency of the incentive system and the autonomous judgment and screening of the discount factor of cooperative willingness; Lai Chengzhe, Chen Yao, Guo Qili and others proposed a cooperative downloading incentive scheme based on two-layer games in the Internet of Vehicles in "Cooperative Downloading Incentive Scheme Based on Game Theory in the Internet of Vehicles", using the reverse auction mechanism to select cooperative downloading vehicles, and the evolutionary game solved the problem of vehicle participation in data forwarding; however, the method of forming vehicle clusters proposed by Zhu Wenlong in "Research on Data Distribution between Vehicles Based on Network Coding" is applied to the geographic-based While cooperative vehicle selection schemes based on location and direction of travel can achieve high efficiency and minimize network overhead, they lack a screening step for cooperative vehicles based on network factors (such as network latency and packet loss rate). Zeng Cheng, in "Research on Cooperative Downloading Problems in Internet of Vehicles," proposed an efficient incentive method based on infinitely repeated games, but this method also lacks a mechanism for filtering out bad vehicles based on network factors. Lai Chengzhe, Chen Yao, Guo Qili, et al., in "Game Theory-Based Cooperative Downloading Incentive Scheme in Internet of Vehicles," proposed a cooperative downloading incentive scheme based on a two-layer game. This scheme, however, only eliminates malicious nodes based on reputation, eliminating superficial risks. It fails to perform a secondary filtering of cooperative vehicles based on network factors, potentially leading to poor transaction quality caused by poor network factors. In real life, if vehicles fail to consider the impact of network factors on vehicle transactions when selecting cooperative vehicles, inter-vehicle communication may be affected by poor network performance, resulting in reduced communication quality, including increased latency, increased packet loss, and slower data transmission speeds. This is reflected in the fact that when the driver's vehicle conducts data transactions with the cooperating vehicle, a series of problems may occur, such as data loss, slow data transmission, and delayed transaction time; secondly, if the cooperative task requires real-time communication or relies on vehicles with high network performance for collaboration, then choosing a vehicle with poor network performance may reduce the efficiency of collaboration, delay the task, and affect the quality of task completion and real-time benefits.Furthermore, vehicles with poor network performance may face a higher risk of data loss and tampering, especially in scenarios involving the transmission of sensitive information, where this phenomenon is even more pessimistic.

[0005] In terms of data pricing, Wang Xiaodan proposed a routing decision based on the OBCDI algorithm in the study of traffic offloading based on intimacy and cooperative downloading in VANET. That is, the cooperative vehicles are sorted according to intimacy (vehicle spacing, activity value, relative speed, relative connection time), ensuring that the requesting vehicle has better transmission quality. Zhao Bin proposed a grid resource management model based on reverse auction technology (RAT) in the study of grid resource management model based on reverse auction technology (RAT), which also used the reverse auction mechanism and introduced it into the network. Integrating this into grid resource management significantly reduces task costs for grid users. Lai Chengzhe, Chen Yao, Guo Qili, and others proposed a method combining a perfect reverse auction (PRA) and a random reverse auction (RRA) in the cooperative vehicle selection phase in "A Game-Theory-Based Cooperative Download Incentive Scheme in the Internet of Vehicles." This approach allows decision-makers (RSUs) to obtain as much and accurate data as possible. This unique combination of a perfect reverse auction (PRA) and a random reverse auction (RRA) (PRA determines a relatively accurate transaction price, while RRA introduces a random factor) increases transaction flexibility, variability, and interest. However, Wang Xiaodan's research on traffic offloading based on intimacy and cooperative downloading in VANETs, ​​which uses the OBCDI algorithm for routing decisions, only collects bids during the reverse auction process during the candidate vehicle selection process, without imposing specific price limits on the bids for cooperative vehicle data (e.g., setting a floor price and a ceiling price). Zhao Bin's research on a grid resource management model based on reverse auction technology (RAT) in "Research on a Grid Resource Management Model Based on Reverse Auction Technology (RAT)" also failed to set a reasonable price range for data during the reverse auction. Lai Chengzhe, Chen Yao, Guo Qili, et al.'s "A Game-Theory-Based Cooperative Download Incentive Scheme in the Internet of Vehicles" proposed a method combining a perfect reverse auction (PRA) and a randomized reverse auction (RRA) for the cooperative vehicle selection phase. However, due to the high randomness of transaction prices, this approach may increase the risk of price volatility. This is particularly true in situations of high transaction volume, information explosion, or unstable market conditions, which can lead to drastic price fluctuations, impacting transaction stability and predictability. In such situations, setting a floor price, a ceiling price, and secondary price limits for transaction data to ensure that it fluctuates within a reasonable and predictable range, ensuring that both parties are not negatively impacted, and enhancing transaction stability, is crucial.

[0006] In terms of data pricing, Cai Wansheng, Song Xi, Gao Wenpeng, and others proposed in "Reverse Auction-Based Incentive Mechanism for Electric Vehicle Energy Trading" that they introduced a multi-user negotiation game framework and subsequently developed an optimal pricing payment strategy for electric vehicles, which can significantly reduce the costs of energy aggregators and incentivize vehicle discharge. Wang Xiaodan proposed in "Research on Traffic Offloading Based on Intimacy and Cooperative Download in VANET" that they used a reverse auction model to incentivize vehicle cooperative download. Zhao Bin proposed in "Research on a Grid Resource Management Model Based on Reverse Auction Technology (RAT)" that they established a fair grid resource trading platform to allow users to fairly compete for the grid resources they need, which is conducive to fair user incentives. Lai Chengzhe, Chen Yao, Guo Qili, and others proposed in "Game Theory-Based Cooperative Download Incentive Scheme in the Internet of Vehicles" that an incentive mechanism based on game theory achieves a positive cycle of overall vehicle system motivation. However, in their paper "Reverse Auction-Based Incentive Mechanism for Electric Vehicle Energy Trading," Cai Wansheng, Song Xi, Gao Wenpeng, and others proposed an optimal pricing and payment strategy for electric vehicles. However, during the reverse auction participant selection process, after the aggregator announces the task to electric vehicles, the bids submitted by the electric vehicles are not specially encrypted, lacking data security and potentially posing the risk of backroom dealings. Wang Xiaodan's paper "Research on Traffic Offloading Based on Affinity and Cooperative Downloading in VANET" utilizes a reverse auction model, but similarly lacks encryption of data bids during the bidding process between candidate vehicles. Zhao Bin's paper "Research on a Grid Resource Management Model Based on Reverse Auction Technology (RAT)" proposes a fair grid resource trading platform, but the grid service providers' bids also lack data protection and encryption during the transaction process. Lai Chengzhe, Chen Yao, Guo Qili, and others' paper "Game Theory-Based Cooperative Downloading Incentive Scheme in the Internet of Vehicles" proposes a game-theory-based incentive mechanism that similarly fails to encrypt and hide the bids of partner vehicles during the data auction phase of partner vehicle selection. In summary, the aforementioned literature achieves positive incentives for the system by utilizing the reverse auction mechanism, but fails to consider the encryption process of data bidding. This not only results in privacy data leakage and information insecurity, but also leads to unfair and subjective auction competition, undoubtedly posing a significant challenge to fair competition and vehicle incentives. For example, in real life, malicious nodes may engage in shady operations and unfair competition after bid information is leaked, resulting in malicious node vehicles becoming the winning set and participating in the transaction. During the transaction, malicious behavior may cause drivers to receive malicious information or receive adverse inducements. Furthermore, near the end of the auction, if the corresponding competitor of the cooperating vehicle learns the current cooperating vehicle's bid, the competitor may adjust its bidding strategy based on the current cooperating vehicle's bid. This may be another level of "knowing oneself and knowing the enemy." However, this not only harms the interests of other vehicles that bid normally, but also leads to unfair and unreasonable auctions.After the auction process is over, if the bidding information of the vehicle that failed the auction is leaked, its competitors can infer its bidding ability and economic level, which is undoubtedly a huge challenge to the privacy protection of the vehicle that failed the auction.

[0007] In summary, the defects of the existing technology are mainly reflected in:

[0008] (1) The impact of network factors on vehicle data transactions was not considered during the selection process for cooperative vehicles, and there was no mechanism to screen out vehicles with bad network factors;

[0009] (2) When using the reverse auction mechanism, there is no restriction on the transaction price, which is highly random and may increase the risk of price fluctuations;

[0010] (3) In the transaction process of the reverse auction mechanism, there is a lack of encryption process for data bidding, and lack of data security. Summary of the Invention

[0011] In response to the shortcomings of the existing technology, the present invention provides a VANET cooperative downloading method based on multi-dimensional screening and encrypted reverse auction, which solves the problem that the cooperative vehicle screening mechanism based on the reverse auction mechanism in the cooperative vehicle selection stage does not take into account the impact of network factors on vehicle data transactions, resulting in decreased communication quality and increased data security risks. It also solves the problem of information asymmetry and high price fluctuation risks caused by the imperfect reverse auction mechanism in the cooperative vehicle selection stage, and solves the problem of potential backroom operations and information leakage risks caused by the failure to encrypt the quotations of cooperative vehicles.

[0012] The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction includes the following steps:

[0013] Step 1: Deploy roadside units along the vehicle route and deploy onboard units on each vehicle;

[0014] Step 2: The roadside unit registers and authenticates the vehicle, recording the reputation value of the verified vehicle. Then, the onboard unit samples the three network performance indicators of the verified vehicle n times within a period of time t and reports them to the roadside unit, i.e., obtaining n data values ​​for each network performance indicator.

[0015] The three network performance indicators include network delay D, packet loss rate L and occupied network throughput T;

[0016] Step 3: Preliminary screening of vehicles based on their reputation and network performance indicators to obtain candidate cooperative vehicles;

[0017] Step 3.1: The roadside unit uses a reputation-based incentive mechanism to evaluate the reputation of vehicles and excludes vehicles with reputation values ​​below the threshold;

[0018] Step 3.2: Further screen the vehicles based on network performance indicators to obtain candidate cooperative vehicles;

[0019] Step 3.2.1: Determine the event type;

[0020] Specifically, the maximum and minimum values ​​of the three network performance indicators, namely, network delay D, packet loss rate L, and occupied network throughput T, are determined respectively, and the number of divided numerical intervals is set. Based on the maximum and minimum values ​​of the three network performance indicators and the number of divided numerical intervals, the three network performance indicators are divided into a number of numerical intervals, each numerical interval being an event type;

[0021] Step 3.2.2: Define that when the data value of a certain network performance indicator of a vehicle is within the numerical range of a certain event type, it means that an event of the event type has occurred on the vehicle;

[0022] Step 3.2.3: Based on the data values ​​of the three network performance indicators of each vehicle obtained within a period of time t, calculate the number of events of each event type that occurred for the vehicle;

[0023] Step 3.2.4: Calculate the probability distribution of each event type for each vehicle;

[0024] Step 3.2.5: Based on the probability distribution of each event type, determine the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle within a period of time t;

[0025] Step 3.2.6: Calculate the information content I(l) of each event based on the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle;

[0026] Step 3.2.7: Based on the information volume of all events for each vehicle, calculate the information entropy of network delay D, packet loss rate L, and occupied network throughput T respectively, and sum them to obtain the total information entropy of the vehicle;

[0027] Step 3.2.8: Calculate the average network delay D, the average packet loss rate L, and the average occupied network throughput T for each vehicle;

[0028] Step 3.2.9: Set a set of thresholds and compare each vehicle's total information entropy, average network delay D, average packet loss rate L, and average occupied network throughput T with the set thresholds to select candidate cooperative vehicles; the thresholds include a total information entropy threshold, a network delay threshold, a packet loss rate threshold, and an occupied network throughput threshold;

[0029] The comparison and screening process is specifically as follows: determining whether the total information entropy, average value of network delay D, average value of packet loss rate L, and average value of occupied network throughput T of the vehicle are lower than corresponding thresholds. If so, the vehicle is qualified and the vehicle number is recorded as a candidate for cooperation. If so, the vehicle is unqualified.

[0030] Step 4: The requesting vehicle sends a download request to the roadside unit. The roadside unit broadcasts the download request. After receiving the download request, the candidate cooperative vehicle generates a quote. The download request includes the download data type and budget.

[0031] Step 5: Encrypt the quote generated by the candidate cooperative vehicle, generate a blind signature and zero-knowledge proof, and submit it to the roadside unit;

[0032] Step 5.1: The candidate cooperation vehicle Vi generates a private key private_key and a public key public_key, where i is the number of the candidate cooperation vehicle;

[0033] Step 5.2: Blind the quote Pi according to the public key public_key to obtain the blinded information;

[0034] Step 5.3: Sign the blinded information using the private key private_key to generate a blind signature σi;

[0035] Step 5.4: The candidate cooperative vehicle Vi generates a zero-knowledge proof;

[0036] Specifically, vehicle Vi selects a random value r, calculates the challenge value ci and the response value s, and then uses the zero-knowledge proof protocol to generate a zero-knowledge proof Ti = (ci, s);

[0037] Step 5.5: The candidate cooperative vehicle Vi submits the blind signature and zero-knowledge proof to the roadside unit;

[0038] Step 6: The roadside unit unblinds the blind signature and verifies the validity of the unblinded signature and zero-knowledge proof. Vehicles that fail the verification are no longer considered as candidate cooperative vehicles; the unblinded signature includes the bid.

[0039] Step 6.1: The roadside unit unblinds the blind signature of the candidate cooperative vehicle to obtain the unblinded signature;

[0040] Step 6.2: Verify the unblinded signature of the candidate vehicle. If the verification is successful, proceed to step 6.3. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, the vehicle is no longer a candidate vehicle.

[0041] Step 6.3: Verify the zero-knowledge proof of the selected cooperative vehicle. If the verification is successful, proceed to step 7. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, it will no longer be considered as a candidate cooperative vehicle.

[0042] Step 7: Set the reserve price, soft ceiling price and ceiling price for the reverse auction;

[0043] Step 7.1: The roadside unit uses the linear regression model to estimate the reserve price of the reverse auction;

[0044] Step 7.1.1: Collect N historical transaction data of the requested vehicle It includes the number of transactions q, transaction time ti, transaction conditions c, transaction floor price p and cost cost within a period of time;

[0045] Step 7.1.2: Collect historical transaction data of the requested vehicle Calculate the information entropy of transaction quantity, transaction time and transaction conditions respectively;

[0046] Step 7.1.3: Determine the weights of the entropy of transaction quantity, the entropy of transaction time, and the entropy of transaction conditions, and calculate the total entropy.

[0047] entropy=w q *entropy_q+w ti *entropy_ti+w c *entropy_c

[0048] Among them, w q ,w ti ,w c are the weights of the information entropy of transaction quantity, the information entropy of transaction time, and the information entropy of transaction conditions, respectively. entropy_q is the information entropy of transaction quantity, entropy_ti is the information entropy of transaction time, and entropy_c is the information entropy of transaction conditions.

[0049] Step 7.1.4: Normalize the transaction quantity, transaction time, and transaction conditions in the historical transaction data to obtain the standardized transaction quantity, transaction time, and transaction conditions;

[0050] Step 7.1.5: Build a linear regression model;

[0051] Y=β0+β1X1+β2X2+……+β N X N +ε

[0052] Where X is the independent variable, β is the linear regression model parameter, i.e. the regression coefficient, ε is the random error term, and Y is the predicted value of the reserve price;

[0053] Step 7.1.6: Construct a training sample set Each sample in the training sample set includes the standardized transaction quantity, the standardized transaction time, the standardized transaction conditions, the total information entropy, the cost cost and the transaction floor price p;

[0054] Step 7.1.7: Fit the constructed linear regression model using the training sample set to obtain a fitted linear regression model;

[0055] Specifically: Use a linear regression model to fit the standardized transaction quantity, transaction time, transaction conditions, total information entropy, cost, and transaction floor price p, and determine the estimated values ​​of the optimal linear regression model parameters by minimizing the residual sum of squares;

[0056] Step 7.1.8: Use the fitted linear regression model to estimate the reserve price and obtain the reserve price for the reverse auction;

[0057] Step 7.2: Set a soft ceiling price and a ceiling price based on the reverse auction reserve price;

[0058] The soft maximum price is:

[0059] price soft_max =price min *(α*w1+b min *w2)

[0060] Among them, price soft_max Is the soft maximum price set, price min is the reserve price of the reverse auction, α is the expected margin of the requested vehicle, b min is the minimum margin, w1 and w2 are the expected margin α and the minimum margin b of the requesting vehicle respectively min The weight value of

[0061] The stated maximum price:

[0062] price limit_max =price min *(b max *w2+μ*w3)

[0063] Among them, price limit_max is the maximum price set, b max is the maximum margin, μ is the supply and demand relationship in the market, and w3 is the weight value of the supply and demand relationship in the market;

[0064] Step 8: The roadside unit screens candidate vehicles based on the reserve price, soft ceiling price, and ceiling price of the reverse auction. It then uses a combination of a perfect reverse auction and a random group reverse auction to screen candidate cooperative vehicles again and obtain the winning vehicle set.

[0065] Step 8.1: Determine whether any candidate partner vehicle's bid is between the reserve price and the soft maximum price. If so, select one or more vehicles with the lowest bid as the winner and proceed to Step 8.3. Otherwise, proceed to Step 8.2.

[0066] Step 8.2: When the bids of the candidate vehicles are above the soft ceiling price, the requesting vehicle will receive a prompt asking whether to raise the price cap to continue the transaction. If it chooses not to, the transaction ends. Otherwise, the process continues to determine whether the bids of the candidate vehicles are below the ceiling price. If so, the vehicle or vehicles with the lowest bids are selected as the winners and step 8.3 is executed. If all bids exceed the ceiling price, the transaction ends.

[0067] Step 8.3: Use the perfect reverse auction mechanism to select the winning vehicles;

[0068] Step 8.4: Use the random group reverse auction mechanism to further screen the winning vehicles and obtain the final winning vehicle set;

[0069] Step 8.4.1: The roadside unit randomly divides the winners into two groups, with each group allocated half of the roadside unit's budget;

[0070] Step 8.4.2: Apply the perfect reverse auction mechanism to each group, introduce a random factor into the auction process, and determine the winning vehicle of each group after the auction ends;

[0071] Step 8.4.3: Combine the two sets of winning vehicles from the auction to form the final set of winning vehicles;

[0072] Step 9: The roadside unit sends a notification to the vehicles in the winning set. After the winning vehicle confirms its intention to cooperate, it prepares to start data downloading.

[0073] Step 10: The winning vehicle downloads the data from the roadside unit and passes the data packet to the requesting vehicle via V2V communication.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1. The existing cooperative download incentive scheme based on game theory in the Internet of Vehicles usually only considers the exclusion of bad nodes and the pursuit of maximizing the number of participating vehicles in the cooperative vehicle selection stage, but ignores the screening of network factors and fails to exclude vehicles with poor network status and high transaction risks. The present invention adds a network factor screening mechanism based on information entropy in the cooperative vehicle selection stage, innovatively integrates relevant knowledge such as information entropy and network delay, and forms a network factor screening supplementary optimization process for cooperative vehicle selection. It can screen out high-efficiency cooperative vehicles with good network performance, and eliminate potential problems of poor transaction quality (increased delay, increased packet loss rate, slower data transmission speed, data loss and tampering risk) caused by poor network factors as much as possible, thereby improving the transaction quality and success rate of the overall in-vehicle ad hoc network system, and improving the efficiency, security and fairness of cooperative downloads in the in-vehicle ad hoc network.

[0076] 2. The reverse auction model used in existing game-theory-based cooperative download incentive schemes in the Internet of Vehicles is difficult to cope with more complex and unstable traffic environments. This invention designs a supplementary and optimized reverse auction mechanism that integrates the concept of "sealed auctions", encrypts the entire transaction process, and enhances the security and fairness of transactions. In addition, a data pricing limitation strategy is added to the data bidding stage of the reverse auction model to limit data price fluctuations to a certain range, thereby maximizing transaction stability and predictability. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 Flowchart of a VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction in an embodiment of the present invention;

[0078] Figure 2 It is a box plot of information entropy and network indicators in an embodiment of the present invention;

[0079] Among them, (a) is the box plot of information entropy and network indicators before screening, and (b) is the box plot of information entropy and network indicators after screening;

[0080] Figure 3 Schematic diagram of the average and standard deviation of vehicle information entropy and network indicators after screening in an embodiment of the present invention;

[0081] Figure 4 This is a graph showing the changing trend of the mean square error of the three models as the number of transactions increases in an embodiment of the present invention;

[0082] Figure 5 This is a graph showing the changing trend of the mean absolute error of the three models as the number of transactions increases in an embodiment of the present invention;

[0083] Figure 6 Graph showing the changing trends of the error coefficients of the three models as the number of transactions increases in an embodiment of the present invention;

[0084] Figure 7 This is a simulation diagram of vehicle quotation in an embodiment of the present invention. DETAILED DESCRIPTION

[0085] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0086] VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction, such as Figure 1 As shown, the following steps are included:

[0087] Step 1: Deploy roadside units (RSUs) along the vehicle routes and onboard units (OBUs) on each vehicle to ensure that the vehicle can access the Internet of Vehicles system;

[0088] Step 2: The roadside unit (RSU) registers and authenticates the vehicle, recording the reputation value of the verified vehicle. Then, the onboard unit (OBU) samples the three network performance indicators of the verified vehicle n times within a period of time t and reports them to the roadside unit (RSU). That is, n data values ​​are obtained for each network performance indicator.

[0089] The three network performance indicators include network delay D, packet loss rate L and occupied network throughput T;

[0090] Step 3: Preliminary screening of vehicles based on their reputation and network performance indicators to obtain candidate cooperative vehicles;

[0091] Step 3.1: The roadside unit (RSU) uses a reputation-based incentive mechanism to evaluate the reputation of vehicles and excludes vehicles with reputation values ​​below the threshold;

[0092] Step 3.2: Further screen the vehicles based on network performance indicators to obtain candidate cooperative vehicles;

[0093] Step 3.2.1: Determine the event type;

[0094] Specifically, the maximum and minimum values ​​of the three network performance indicators, network delay D, packet loss rate L, and occupied network throughput T, are determined respectively, and the number of divided numerical intervals is set. Based on the maximum and minimum values ​​of the three network performance indicators and the number of divided numerical intervals, the three network performance indicators are divided into a number of numerical intervals, each numerical interval representing an event type, such as low network delay, medium network delay, high network delay, low packet loss rate, medium packet loss rate, high packet loss rate, and low network throughput;

[0095] The boundaries of each numerical interval are:

[0096] edges=linspace(min,max,bins+1)

[0097] Among them, edges is the boundary of the numerical interval, linspace is the average calculation instruction in MATLAB, min is the minimum value of a certain network performance indicator, max is the maximum value of a certain network performance indicator, and bins is the number of set numerical intervals;

[0098] Step 3.2.2: Define that when the data value of a certain network performance indicator of a vehicle is within the numerical range of a certain event type, it means that an event of the event type has occurred on the vehicle;

[0099] Step 3.2.3: Based on the data values ​​of the three network performance indicators of each vehicle obtained within a period of time t, calculate the number of events of each event type that occurred for the vehicle;

[0100]

[0101] in, is the data value of the network performance indicator, Counts is the number of network performance indicators that fall within a certain numerical interval, that is, the number of events of a certain event type; histcounts is a function in MATLAB used to calculate the number of times a data value falls within a specified interval;

[0102] Step 3.2.4: Calculate the probability distribution of each event type for each vehicle;

[0103] prob=counts / sum(counts)

[0104] Among them, prob is the probability distribution of a certain event type, and sum is the summation function;

[0105] Step 3.2.5: Based on the probability distribution of each event type, determine the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle within a period of time t;

[0106] The process is expressed as:

[0107]

[0108] Among them, indies is the event type corresponding to the data value of a certain network performance indicator, and discretize returns the boundary of the data interval to the data value of the network performance indicator, thereby determining the event type to which the data value of the network performance indicator belongs;

[0109] prob*=prob(indies)

[0110] Where prob* is the probability of occurrence of an event type corresponding to the data value of a certain network performance indicator. That is, the probability distribution of the event type (prob) is used to find the probability of occurrence of an event type corresponding to the data value of the network performance indicator using the event type (indies) corresponding to the data value of the network performance indicator.

[0111] P = reshape(prob*, n, 1)

[0112] Where P is a one-dimensional vector, in which each element is the probability of an event corresponding to a data value of a network performance indicator. Reshape means that the probability of an event corresponding to the data value of each network performance indicator of a vehicle is arranged in order to ensure that the data value of the network performance indicator, the corresponding event type, and the probability of the corresponding event are one-to-one corresponding.

[0113] Step 3.2.6: Calculate the information content I(l) of each event based on the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle;

[0114] The information content of each event is calculated using the information theory formula. Information content represents the amount of information provided by an event. Generally, information content is correlated with the probability of an event occurring; events with lower probability of occurrence provide greater information.

[0115] The information quantity formula is:

[0116] I(l)=-log2(P)

[0117] Where l is the event number;

[0118] Step 3.2.7: Based on the information volume of all events for each vehicle, calculate the information entropy of network delay D, packet loss rate L, and occupied network throughput T respectively, and sum them to obtain the total information entropy of the vehicle;

[0119] Information entropy represents the uncertainty or randomness of the entire system or data set. Usually, information entropy is the weighted average of the information content of all events.

[0120] The information entropy formula is expressed as:

[0121]

[0122] Among them, H(n) is the information entropy of a certain network performance indicator; P(l) is the probability of occurrence of a certain event corresponding to the network performance indicator;

[0123] Step 3.2.8: Calculate the average network delay D, the average packet loss rate L, and the average occupied network throughput T for each vehicle;

[0124] ave(D)=mean(D)

[0125] ave(L)=mean(L)

[0126] ave(T)=mean(T)

[0127] Where ave(D) is the average value of the vehicle network delay, ave(L) is the average value of the vehicle packet loss rate, ave(T) is the average value of the vehicle network throughput, and mean is a function in MATLAB used to calculate the average value.

[0128] Step 3.2.9: Based on real-world research and analysis, a set of thresholds is set. The total information entropy, average network delay D, average packet loss rate L, and average occupied network throughput T of each vehicle are compared with the set thresholds to select candidate cooperative vehicles. The thresholds include the total information entropy threshold, network delay threshold, packet loss rate threshold, and occupied network throughput threshold.

[0129] The comparison and screening process is specifically as follows: because the threshold is the dividing point of whether the vehicle network is stable or not and the quality of the network is good or bad, determined through actual conditions and research analysis, the basis for judging whether the vehicle's network condition meets the standard is whether the vehicle's total information entropy, the average value of the network delay D, the average value of the packet loss rate L, and the average value of the occupied network throughput T are lower than the corresponding threshold. If they are lower than the threshold, the vehicle is qualified, the vehicle number is recorded, and it is used as a candidate cooperative vehicle for subsequent reverse auctions. If it is higher than the threshold, it is determined that the vehicle network condition is unstable or the network quality is poor.

[0130] A MATLAB simulation experiment was conducted to assess vehicle network status and select candidate cooperative vehicles. The goal was to verify that using vehicle information entropy can identify vehicles with unstable network conditions, a key area where traditional selection methods struggle. This simulation selected 10 vehicles, numbered 1 to 10, and measured their network latency, packet loss rate, and network throughput over 100 minutes, collecting data every minute. The information entropy, average network latency, average packet loss rate, and average network throughput were calculated for each vehicle.

[0131] ① In this simulation, the information entropy threshold is set to 1.2, the average network delay threshold is set to 72, the average packet loss rate threshold is set to 0.1, and the average occupied network throughput threshold is set to 80. The information entropy and network performance indicators are standardized. If the data is greater than 1, it exceeds the threshold, which is used to determine whether they are qualified.

[0132] ② Use the standardized data to calculate the average and standard deviation of the information entropy and each network performance index before and after screening.

[0133] ③Through Figure 2 The box plot shown observes the distribution before and after screening.

[0134] Table 1 Mean and standard deviation of information entropy and performance indicators before and after screening by this scheme

[0135] Information entropy Network latency Packet loss rate Occupied network throughput Average value before screening: 0.9484 0.7174 0.7238 0.8586 Standard deviation before screening: 0.0500 0.1835 0.0740 0.0783 Filtered average: 0.9337 0.6446 0.6967 0.8276 Standard deviation after screening: 0.0444 0.1536 0.0546 0.0412

[0136] Table 1 shows that after filtering using this solution, the average values ​​of vehicle network latency, packet loss rate, and occupied network throughput are lower, indicating a better network status. The standard deviation is also reduced, indicating better network stability. The change in information entropy before and after filtering also illustrates this result.

[0137] Depend on Figure 2 (a) Boxplot before screening and Figure 2 (b) The box plot after screening shows that after screening by this scheme, the boxes of each network indicator are smaller, which indicates that the data has become more concentrated; there are fewer outliers outside the range of 1.5 times the IQR, and the data quality is higher.

[0138] Depend on Figure 3 As can be seen, the average values ​​of vehicle 6's network delay, packet loss rate, and occupied network throughput are normal, but the information entropy of these indicators exceeds the threshold. This indicates that the average value and standard deviation are significantly affected by interference data. Using information entropy combined with a linear regression model can reduce interference.

[0139] Step 4: The requesting vehicle (vr) sends a download request to the roadside unit (RSU). The RSU broadcasts the download request. After receiving the download request, the candidate partner vehicle generates a quote Pi, where i is the candidate partner vehicle number. The download request includes the download data type and budget.

[0140] Step 5: Encrypt the quote generated by the candidate cooperative vehicle, generate a blind signature and zero-knowledge proof, and submit it to the roadside unit (RSU);

[0141] Step 5.1: The candidate cooperative vehicle Vi generates a private key private_key and a public key public_key;

[0142] The private key is kept confidential and is used to sign bids or other data generated by the vehicle. The public key is used to verify the vehicle's signature and can be securely submitted to a trusted third party, such as an auction system administrator or other bidders. They will use the vehicle's public key to verify the signature submitted by the vehicle to ensure that the bid is indeed from the vehicle and has not been tampered with.

[0143] Step 5.2: Blind the quote Pi according to the public key public_key to obtain the blinded information;

[0144] The blinded information Mi is:

[0145] Mi=Pi*(bi**public_key)

[0146] Among them, bi is the blinding factor;

[0147] Step 5.3: Sign the blinded information using the private key private_key to generate a blind signature σi;

[0148] σi=Blindsign(private_key,Mi)

[0149] Among them, Blindsign is the blind signature algorithm in the cryptography database in Python;

[0150] Step 5.4: The candidate cooperative vehicle Vi generates a zero-knowledge proof;

[0151] Specifically, vehicle Vi selects a random value r, calculates the challenge value ci and the response value s, and then uses the zero-knowledge proof protocol to generate a zero-knowledge proof Ti = (ci, s);

[0152] The challenge value ci is:

[0153] ci=hash(statement+r)

[0154] Here, hash refers to a hash function, and statement refers to a public statement or statement to be proved.

[0155] The response value s is:

[0156] s=(r+ci*secret)%prime

[0157] Among them, secret refers to the information to be hidden and encrypted, that is, the bid of the cooperating vehicle in the reverse auction, and prime is a prime number;

[0158] Step 5.5: The candidate cooperative vehicle Vi submits the blind signature and zero-knowledge proof to the roadside unit (RSU);

[0159] Step 6: The roadside unit (RSU) unblinds the blind signature and verifies the validity of the unblinded signature and zero-knowledge proof to ensure the authenticity and security of the bid. Vehicles that fail the verification are no longer considered as candidate cooperative vehicles; the unblinded signature includes the bid.

[0160] Step 6.1: The roadside unit (RSU) unblinds the blind signature of the candidate cooperative vehicle to obtain the unblinded signature;

[0161] The signature after unblinding is:

[0162]

[0163] Among them, Signature is the signature after unblinding;

[0164] Step 6.2: Verify the unblinded signature of the candidate vehicle. If the verification is successful, proceed to step 6.3. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, the vehicle is no longer a candidate vehicle.

[0165] In this embodiment, the first step is to obtain the unblinded signature. The signature is initially encrypted, and after unblinding, the plaintext signature value is obtained, which contains a clear data quote. Next, a predetermined hash algorithm (such as SHA-256) is used to recalculate the hash value. This hash value is usually based on the public statement and the prover's public key (public_key), and is calculated using the formula `hash(statement+public_key)`, where the statement is the public information participating in the proof, and the public key is the public key of the cooperating vehicle. Finally, the calculated hash value is compared with the unblinded signature. If the unblinded signature is equal to the hash value, that is, `DecryptedSignature=hash(statement+public_key)`, then the verification is successful, indicating that the signature is valid and the prover's public key is consistent with the statement; if they are inconsistent, the verification fails, which means that the signature is invalid or has been tampered with. If there is reasonable suspicion that the verification failure is due to a technical problem, the vehicle can be re-submitted with a signature for secondary verification. If the verification fails multiple times, the vehicle is marked as a malicious node and disqualified from participating.

[0166] Step 6.3: Verify the zero-knowledge proof of the selected cooperative vehicle. If the verification is successful, proceed to step 7. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, it will no longer be considered as a candidate cooperative vehicle.

[0167] Set a challenge value C', then take a prime number from the challenge value C' to get C'_prime:

[0168] C'_prime=hash(statement+(public_key**ci))

[0169] Re-verify the challenge value ci == C'_prime. Are they consistent? As above, if they are, the zero-knowledge proof succeeds; if not, verification fails. In some cases, verification failure may be caused by a transmission error or technical malfunction. You can request the prover to submit a new proof and perform a new verification. If verification fails multiple times, the prover will be marked as a potential malicious actor or untrusted node, and will be completely excluded from future transactions.

[0170] In this implementation, assume that the auction data bid of a cooperative vehicle is 10,000. Without the data bid encryption process proposed in this solution, the leaked information is the real information 10,000, which poses a risk to data security. However, after the data bid encryption process combining blind signatures and zero-knowledge proofs in this solution, even if the data is leaked, only a string of encrypted information data that is difficult to crack is obtained. The data output result is as follows:

[0171] The output of the “original code” leak:

[0172] #Price=10000;

[0173] The output of the code "Combining blind signature and zero-knowledge proof" is:

[0174] Blind signature:b'T\x17\xddN\xb6\xdc\xd05\xc9dVfW\x10\xfd2q\xa6O\x80\xa0\xcai\x80q<W3s\x9c^\x0b?~\xf8Df\x81(\xb0\x06\x90\xc6\x90*\xdd\xf0\x9ab\xfa!\x1e\x87 / \x97 / \xca\x91\x1d\xc8\x95\xf5{\xcf\xc9\x10\x01\xd5gk\xdfRS_!ex\xa4\x06\x0c\xc3O\x88\xbc\xc4\xea\xc9\xe8}V{\xcaF\xc7TV\xfc\x0b\xd5\xcd\xcdq?\xdb|:\x85\x03\xc7\xf8\xe1\xe6f\x1e\x87\xb9L~\x01B\xc7\xed\x00q\xb5*\xec\x89\xac\x04\x01@r\xc3\xd2\xe0\x1bE9ce\xda\xb2\x89\x9c\xed\xe1\xc12Y\xa5|\xa0\xb5\xf3\xfc\\\x9cP#W\xd7d}\x8b\t#\xf4B\x18\x88\x9d\x84\xb3=\x8bWfn\xe3r\xc6\x03\xeb\xda\x0b\xe6\xb8Z\xb5\xeb\xa0-\xf1@?\x12L\xf4\x8e\x00\x94\x04A\x1b\x95\x90g.(\x9d\x91i\xae\x18DICp\xa3G\x9c\xce\xe6\x034\x14\xd0\x1d\x03c\xdcM+\xb8\x91%\x81\xbc\x8bC\xc7\xc7<\xae\x8a~\xfe\xf2q\x9aL\xfbHt\x01A7\xb9\xf9\xd1W\xe8\x97\x05k\xf2n\x8f\xe7\xf3\xb8\xb9\xbc\x91\xd93}\x9aJZt\x9d\xebW\xce+\x1bv\x9c\xc5\x18\xa5\tM\\<\xd50\xb8\x00j\xc14,r\xbc\xcc\x8b\x10\x98\xac,\xb4\xd3\x06\x19\xc5\xd13B\x85\x1f\xa9\x9cPKG\x9b\xb7\x86\x81\x9d\xbc\xc6\x94{\xf5\xee\x1fA\x00\x89\x

[0175] Zero-knowledgeproof:Zero-knowledgeproof

[0176] Blind signature verification successful.

[0177] Zero-knowledgeproof verification successful.

[0178] The above simulation, run in Python on PyCharm, demonstrates how a vehicle uses a combination of blind signatures and zero-knowledge proofs to encrypt and hide the bids of partner vehicles and subsequently verify them. The output shows the encrypted content of the blind signature and the corresponding zero-knowledge proof, indicating that both the blind signature and the zero-knowledge proof were successful. These results demonstrate the feasibility of this mechanism and offer a new approach to encrypting and verifying bids from partner vehicles in VANETs.

[0179] Combining blind signature technology and zero-knowledge proof technology can not only provide encryption functions for vehicle quotation information and data pricing, but also give full play to the unique properties and characteristics of each, with advanced privacy protection, anonymity, verifiability, and the characteristics of increasing the participation of cooperative vehicles, achieving the effect of "1+1>2".

[0180] 1. Advanced Privacy: Blind signature technology protects the privacy of the signing vehicle, preventing it from learning the message content during the signing process. Combined with zero-knowledge proof technology, this further ensures that even after receiving the signature, the verification authority cannot obtain the signing vehicle's identity or the message content, providing an even higher level of privacy protection.

[0181] 2. Anonymity: The combination of blind signature technology and zero-knowledge proof technology can achieve anonymous signature, that is, the signing vehicle can sign the vehicle quotation and data pricing without revealing its identity, and prove the validity of the signature to the verification agency.

[0182] 3. Verifiability: Combining blind signatures with zero-knowledge proof technology ensures the validity of signatures. This means that verification agencies can verify the validity of signatures without knowing the identity of the signing vehicle or the bid price. This prevents malicious vehicles from colluding with auctioneers to engage in shady operations and unfair competition, greatly enhancing the credibility and security of the signatures.

[0183] 4. Increase the participation of cooperating vehicles: Since vehicle quotes and data pricing are encrypted and privacy-protected, vehicles’ trust in the system increases, and they may be more willing to participate in the auction process, thereby increasing the number of cooperating vehicles and significantly improving the efficiency and performance of the system.

[0184] Step 7: Set the reserve price, soft ceiling price and ceiling price for the reverse auction;

[0185] Step 7.1: The roadside unit (RSU) uses a linear regression model to estimate the reserve price of the reverse auction;

[0186] Step 7.1.1: Collect N historical transaction data of the requested vehicle It includes the number of transactions q, transaction time ti, transaction conditions c, transaction floor price p and cost cost within a period of time;

[0187] Step 7.1.2: Collect historical transaction data of the requested vehicle Calculate the information entropy of transaction quantity, transaction time and transaction conditions respectively;

[0188] Similar to the above method, determine the event type based on the vehicle's historical transaction data Find the probability distribution of event types corresponding to transaction quantity, transaction time, and transaction conditions, calculate the probability of each corresponding event, and then obtain the amount of information. Based on the amount of information, calculate the information entropy of transaction quantity, transaction time, and transaction conditions respectively;

[0189] Step 7.1.3: Based on the application research, determine the weights of the information entropy of transaction quantity, information entropy of transaction time, and information entropy of transaction conditions, and calculate the total information entropy;

[0190] entropy=w q *entropy_q+w ti *entropy_ti+w c *entropy_c

[0191] Among them, w q ,w ti ,w c are the weights of the information entropy of transaction quantity, the information entropy of transaction time, and the information entropy of transaction conditions, respectively. entropy_q is the information entropy of transaction quantity, entropy_ti is the information entropy of transaction time, and entropy_c is the information entropy of transaction conditions.

[0192] Step 7.1.4: Normalize the transaction quantity, transaction time, and transaction conditions in the historical transaction data to obtain the standardized transaction quantity, transaction time, and transaction conditions;

[0193] Step 7.1.5: Build a linear regression model;

[0194] Y=β0+β1X1+β2X2+……+β N X N +ε

[0195] Where X is the independent variable, β is the linear regression model parameter, i.e. the regression coefficient, ε is the random error term, and Y is the predicted value of the reserve price;

[0196] Step 7.1.6: Construct a training sample set Each sample in the training sample set includes the standardized transaction quantity, the standardized transaction time, the standardized transaction conditions, the total information entropy, the cost cost and the transaction floor price p;

[0197] Step 7.1.7: Fit the constructed linear regression model using the training sample set to obtain a fitted linear regression model;

[0198] Specifically: Use a linear regression model to fit the standardized transaction quantity, transaction time, transaction conditions, total information entropy, cost and transaction floor price p, and determine the optimal parameter estimate by minimizing the residual sum of squares

[0199] The process is expressed as:

[0200]

[0201] Among them, LinearRegression.fit is the training of the linear regression model, and p[:N] is the observed value of the reserve price;

[0202]

[0203] Among them, p i To minimize the predicted value, To minimize the observed value;

[0204] Step 7.1.8: Use the fitted linear regression model to estimate the reserve price and obtain the reserve price for the reverse auction;

[0205]

[0206] Among them, price.Ir is the predicted reserve price, and LinearRegression.predict is the reserve price predicted by the trained linear regression model;

[0207] This example simulates 200 historical transaction records for different vehicles. The number of transactions, transaction time, transaction conditions (here set as the transaction range), and the cost of a successful transaction are calculated. The actual reserve price at the time of the auction is determined through manual analysis and market research to determine the accuracy of the estimated reserve price.

[0208] The original solution directly used the information entropy model to estimate the reserve price. This solution uses a linear regression model combined with the information entropy model to estimate the reserve price.

[0209] The ridge regression model expression is as follows:

[0210] in, is the parameter estimate of the ridge regression model, Y is the dependent variable, X is the independent variable, β is the model parameter, and α is the regularization parameter, which is used to control the influence of the regularization term.

[0211] To prevent the linear regression model from overfitting, the experiment used a ridge regression model for further comparison. If the reserve price predicted by the linear regression model differs significantly from the reserve price estimated by the ridge regression model, this indicates that the linear regression model may be overfitting. If the curves after their simulations are not much different, this may indicate that the linear regression model is able to capture the main trends and relationships in the data when fitting the training data, without overfitting noise or outliers.

[0212] As the amount of training data gradually increases, the performance indicators of the three models in estimating the reserve price are recorded, including the mean square error (MSE), mean square absolute error (MAE), and coefficient of determination (R 2 ) situation.

[0213] Mean Squared Error (MSE): measures the difference between the predicted value and the actual value by squared error and is more sensitive to large errors.

[0214] Mean Absolute Error (MAE): measures the difference between the predicted value and the actual value by the absolute error, and weights all errors equally.

[0215] Depend on Figure 4 , Figure 5 From the two pictures we can see:

[0216] ①The MSE and MAE values ​​of this scheme are both smaller than those of the original scheme, which shows that the combination of linear regression model and information entropy model can better predict the reserve price.

[0217] ② This scheme uses a linear regression model combined with an information entropy model. When the training data is small, the MSE and MAE are smaller, and they increase with the increase of training data. The ridge regression model combined with the information entropy model has larger MSE and MAE, and they decrease with the increase of training data.

[0218] This is because the linear regression model is overfitting, but it stabilizes as the training data increases. (This simulation is limited by the training data set and fails to capture the characteristics of the variables and predicted values. If it were real data, the MSE and MAE of the linear regression model would begin to decline after the training set increases to a certain value.) The ridge regression model combined with the information entropy model has a greater impact on the predicted value due to the influence of the regularization term, but this impact decreases as the number of training sets increases. Finally, after the number of training sets exceeds 150, the MSE and MAE of the linear regression model and the ridge regression model converge, indicating that the linear regression model is not overfitting and can more accurately predict the reserve price.

[0219] Coefficient of determination (R 2 ): Measures how well the model explains the data. 2 The closer it is to 1, the stronger the explanatory power of the model. It is suitable for evaluating the overall fitting results of the model. Figure 6 It can be seen that:

[0220] ③ When there are enough training sets (greater than 150), the determination coefficient of this scheme is close to 0.8, which is higher than the original scheme, indicating that the linear regression model combined with the information entropy model can better explain the original data and has a better overall fitting effect.

[0221] Step 7.2: Set a soft ceiling price and a ceiling price based on the reverse auction reserve price;

[0222] Two-level maximum price strategy: the soft maximum price is considered to be a reasonable price ceiling, and the maximum price is considered to be the final price limit that relaxes the price ceiling to take into account the urgent needs of the requester and does not affect the balance of market transaction prices.

[0223] A price ceiling is a maximum price limit set during a transaction, above which transactions are prohibited. It serves to prevent excessively high vehicle transaction prices, protect the interests of requesting vehicles, control market fluctuations, standardize transaction processes, and ensure the rational use of resources.

[0224] The soft ceiling price is set above the reserve price, leaving a margin for the trading vehicle and taking into account the expected value of the requesting vehicle in this reverse auction. If the requesting vehicle urgently needs assistance from another vehicle, but the trading vehicle's bid is above the soft ceiling price, the requesting vehicle may be willing to accept a price above the soft ceiling. Therefore, a two-tiered ceiling price strategy is adopted.

[0225] During the implementation of the two-level maximum price strategy, in order to reduce the risk of price limit setting, macro-monitoring of market research was introduced. Based on market analysis, two margins were artificially set: the minimum margin and the maximum margin, which were used to set the two price limits respectively.

[0226] Since the soft maximum price is a guide price, it should be higher than the reserve price in the reverse auction. min Based on the acceptance of the vehicle's expected margin α and the macro-monitoring of market research - the minimum margin b min The combined impact of

[0227] price soft_max =price min *(α*w1+b min *w2)

[0228] Among them, price soft_max is the set soft maximum price, w1 and w2 are the expected margin α and minimum margin b of the requested vehicle respectively min The weight value of

[0229] The maximum price is needed to limit the transaction price to prevent price gouging. It should be above the reserve price of the reverse auction. min Based on the macro monitoring of market research - maximum margin b max and the impact of the supply-demand relationship μ;

[0230] price limit_max =price min *(b max *w2+μ*w3)

[0231] Among them, price limit_max is the maximum price set, b max is the maximum margin, μ is the supply and demand relationship in the market, and w3 is the weight value of the supply and demand relationship in the market;

[0232] Case 1: When the number of vehicles participating in the bidding exceeds the standard number of vehicles: μ is 1; the standard number of vehicles refers to the number of vehicles participating in the bidding, and the reverse auction market supply and demand relationship reaches normal;

[0233] Case 2: When the number of vehicles participating in the bidding is less than the minimum number of participating vehicles: μ is The maximum value of μ max ; The μ max It is the maximum value set for the market supply and demand relationship. The minimum number of participating vehicles is the number of standard vehicles divided by μ max to determine;

[0234] Case 3. When the number of vehicles participating in the bidding is between the minimum number of participating vehicles and the standard number of vehicles: μ is calculated based on To determine, the value range of μ is [1, μ max ];

[0235] In this embodiment, the predicted reserve price is:

[0236] Predictedprice for the new transaction with entropy assumption:34.25447852798585

[0237] If there are 10 vehicles participating in the bidding, assume that the expected margin of the requested vehicle is 0.6. After market research, the lowest margin is 0.4 and the highest margin is 1.0. The weight values ​​are w1=0.4, w2=0.6, and w3=0.4.

[0238] Set the soft ceiling price and the ceiling price to:

[0239] max_price_values:[50.69662822]

[0240] max_limit_price_values:[75.35985276]

[0241] Vehicle quotation as Figure 7 shown. Figure 7 In the chart, the horizontal lines from bottom to top represent the set floor price, soft ceiling price, and ceiling price.

[0242] Step 8: The roadside unit (RSU) screens candidate vehicles based on the reverse auction's reserve price, soft ceiling price, and ceiling price. It then uses a combination of a perfect reverse auction (PRA) and a randomized reverse auction (RRA) to screen candidate cooperative vehicles again and obtain the winning vehicle set.

[0243] Step 8.1: Determine whether any candidate partner vehicle's bid is between the reserve price and the soft maximum price. If so, select one or more vehicles with the lowest bid as the winner and proceed to Step 8.3. Otherwise, proceed to Step 8.2.

[0244] Step 8.2: When the bids of the candidate vehicles are above the soft ceiling price, the requesting vehicle will receive a prompt asking whether to raise the price cap to continue the transaction. If it chooses not to, the transaction ends. Otherwise, the process continues to determine whether the bids of the candidate vehicles are below the ceiling price. If so, the vehicle or vehicles with the lowest bids are selected as the winners and step 8.3 is executed. If all bids exceed the ceiling price, the transaction ends.

[0245] The purpose of establishing two maximum price levels is to adjust to the needs of requesting vehicles. We prefer that vehicle transactions be completed within the first price limit to maintain price stability and control market volatility. Therefore, even if a transaction is successfully completed within the second price limit but outside the first price limit, certain penalties will be imposed on both parties (for example, the requesting vehicle's reputation will be reduced; if the winning vehicle participates in auctions outside the first price limit too many times within a certain period of time, its reward and penalty factors will be affected, reducing its remuneration).

[0246] Step 8.3: Use the perfect reverse auction (PRA) mechanism to select the winning vehicle;

[0247] Step 8.4: Use the random group reverse auction (RRA) mechanism to further screen the winning vehicles and obtain the final winning vehicle set;

[0248] Step 8.4.1: The RSU randomly divides the winners into two groups (e.g., group T and group W), with each group allocated half of the RSU's budget;

[0249] Step 8.4.2: Apply the perfect reverse auction (PRA) mechanism to each group, but introduce a random factor in the auction process. After the auction, determine the winning vehicle of each group;

[0250] The random factors include randomly adjusting the bid range and randomly selecting some vehicles to participate, so as to increase the flexibility and fairness of the auction;

[0251] Step 8.4.3: Combine the two sets of winning vehicles from the auction to form the final set of winning vehicles;

[0252] Taking into account the influence of random factors, this set may include some or all of the preliminary winning vehicles, and may also include new vehicles added due to random factors.

[0253] Step 9: The roadside unit (RSU) sends a notification to the vehicles in the winning set. After the winning vehicle confirms its intention to cooperate, it prepares to start data downloading.

[0254] Step 10: The winning vehicle downloads the data from the roadside unit (RSU) and passes the data packet to the requesting vehicle (VR) via V2V communication.

Claims

1. A VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction, characterized in that: The steps include: Step 1: Deploy roadside units along the vehicle route and deploy onboard units on each vehicle; Step 2: The roadside unit registers and authenticates the vehicle, recording the reputation value of the verified vehicle. Then, the onboard unit samples the three network performance indicators of the verified vehicle n times over a period of time and reports them to the roadside unit, i.e., obtaining n data values ​​for each network performance indicator. Step 3: Preliminary screening of vehicles based on their reputation and network performance indicators to obtain candidate cooperative vehicles; Step 4: The requesting vehicle sends a download request to the roadside unit. The roadside unit broadcasts the download request. After receiving the download request, the candidate partner vehicle generates a bid. Step 5: Encrypt the quote generated by the candidate cooperative vehicle, generate a blind signature and zero-knowledge proof, and submit it to the roadside unit; Step 6: The roadside unit unblinds the blind signature and verifies the validity of the unblinded signature and zero-knowledge proof. Vehicles that fail the verification are no longer considered as candidate cooperative vehicles; the unblinded signature includes the bid. Step 7: Set the reserve price, soft ceiling price and ceiling price for the reverse auction; Step 8: The roadside unit screens candidate vehicles based on the reserve price, soft ceiling price, and ceiling price of the reverse auction. It then uses a combination of a perfect reverse auction and a random group reverse auction to screen candidate cooperative vehicles again and obtain the winning vehicle set. Step 9: The roadside unit sends a notification to the vehicles in the winning set. After the winning vehicle confirms its intention to cooperate, it prepares to start data downloading. Step 10: The winning vehicle downloads the data from the roadside unit and passes the data packet to the requesting vehicle via V2V communication.

2. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 1 is characterized in that: The three network performance indicators mentioned in step 2 include network delay D, packet loss rate L and occupied network throughput T.

3. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 2 is characterized in that: Step 3 specifically includes: Step 3.1: The roadside unit uses a reputation-based incentive mechanism to evaluate the reputation of vehicles and excludes vehicles with reputation values ​​below the threshold; Step 3.2: Further screen the vehicles based on the network performance indicators to obtain candidate cooperative vehicles.

4. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 3 is characterized in that: Step 3.2 specifically includes: Step 3.2.1: Determine the event type; Specifically, the maximum and minimum values ​​of the three network performance indicators, namely, network delay D, packet loss rate L, and occupied network throughput T, are determined respectively, and the number of divided numerical intervals is set. Based on the maximum and minimum values ​​of the three network performance indicators and the number of divided numerical intervals, the three network performance indicators are divided into a number of numerical intervals, each numerical interval being an event type; Step 3.2.2: Define that when the data value of a certain network performance indicator of a vehicle is within the numerical range of a certain event type, it means that an event of the event type has occurred on the vehicle; Step 3.2.3: Based on the data values ​​of the three network performance indicators of each vehicle obtained within a period of time t, calculate the number of events of each event type that occurred for the vehicle; Step 3.2.4: Calculate the probability distribution of each event type for each vehicle; Step 3.2.5: Based on the probability distribution of each event type, determine the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle within a period of time t; Step 3.2.6: Calculate the information content I(l) of each event based on the probability of occurrence of the event corresponding to the data value of the network performance indicator of each vehicle; Step 3.2.7: Based on the information volume of all events for each vehicle, calculate the information entropy of network delay D, packet loss rate L, and occupied network throughput T respectively, and sum them to obtain the total information entropy of the vehicle; Step 3.2.8: Calculate the average network delay D, the average packet loss rate L, and the average occupied network throughput T for each vehicle; Step 3.2.9: Set a set of thresholds and compare each vehicle's total information entropy, average network delay D, average packet loss rate L, and average occupied network throughput T with the set thresholds to select candidate cooperative vehicles; the thresholds include a total information entropy threshold, a network delay threshold, a packet loss rate threshold, and an occupied network throughput threshold; The comparative screening process is specifically as follows: determine whether the vehicle's total information entropy, average network delay D, average packet loss rate L, and average occupied network throughput T are lower than the corresponding thresholds. If so, the vehicle is qualified and the vehicle number is recorded as a candidate cooperative vehicle. If so, the vehicle is unqualified.

5. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 1 is characterized in that: Step 5 specifically includes: Step 5.1: The candidate cooperation vehicle Vi generates a private key private_key and a public key public_key, where i is the number of the candidate cooperation vehicle; Step 5.2: Blind the quote Pi according to the public key public_key to obtain the blinded information; Step 5.3: Sign the blinded information using the private key private_key to generate a blind signature σi; Step 5.4: The candidate cooperative vehicle Vi generates a zero-knowledge proof; Specifically, vehicle Vi selects a random value r, calculates the challenge value ci and the response value s, and then uses the zero-knowledge proof protocol to generate a zero-knowledge proof Ti = (ci, s); Step 5.5: The candidate cooperative vehicle Vi submits the blind signature and zero-knowledge proof to the roadside unit.

6. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 1 is characterized in that: Step 6 specifically includes: Step 6.1: The roadside unit unblinds the blind signature of the candidate cooperative vehicle to obtain the unblinded signature; Step 6.2: Verify the unblinded signature of the candidate vehicle. If the verification is successful, proceed to step 6.

3. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, the vehicle is no longer a candidate vehicle. Step 6.3: Verify the zero-knowledge proof of the selected cooperative vehicle. If the verification is successful, proceed to step 7. Otherwise, re-obtain the blind signature for verification. If the verification fails within the set number of times, it will no longer be considered as a candidate cooperative vehicle.

7. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 1 is characterized in that: Step 7 specifically includes: Step 7.1: The roadside unit uses the linear regression model to estimate the reserve price of the reverse auction; Step 7.2: Set a soft ceiling price and a ceiling price based on the reverse auction reserve price; The soft maximum price is: price soft_max =price min *(α*w1+b min *w2) Among them, price soft_max Is the soft maximum price set, price min is the reserve price of the reverse auction, α is the expected margin of the requested vehicle, b min is the minimum margin, w1 and w2 are the expected margin α and the minimum margin b of the requesting vehicle respectively min The weight value of The stated maximum price: price limit_max =price min *(b max *w2+μ*w3) Among them, price limit_max is the maximum price set, b max is the maximum margin, μ is the supply and demand relationship in the market, and w3 is the weight value of the supply and demand relationship in the market.

8. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 7 is characterized in that: Step 7.1 specifically includes: Step 7.1.1: Collect N historical transaction data of the requested vehicle It includes the number of transactions q, transaction time ti, transaction conditions c, transaction floor price p and cost cost within a period of time; Step 7.1.2: Collect historical transaction data of the requested vehicle Calculate the information entropy of transaction quantity, transaction time and transaction conditions respectively; Step 7.1.3: Determine the weights of the entropy of transaction quantity, the entropy of transaction time, and the entropy of transaction conditions, and calculate the total entropy. entropy=w q *entropy_q+w ti *entropy_ti+w c *entropy_c Among them, w q ,w ti ,w c are the weights of the information entropy of transaction quantity, the information entropy of transaction time, and the information entropy of transaction conditions, respectively. entropy_q is the information entropy of transaction quantity, entropy_ti is the information entropy of transaction time, and entropy_c is the information entropy of transaction conditions. Step 7.1.4: Normalize the transaction quantity, transaction time, and transaction conditions in the historical transaction data to obtain the standardized transaction quantity, transaction time, and transaction conditions; Step 7.1.5: Build a linear regression model; Y=β0+β1X1+β2X2+……+β N X N +e Where X is the independent variable, β is the linear regression model parameter, i.e. the regression coefficient, ε is the random error term, and Y is the predicted value of the reserve price; Step 7.1.6: Construct a training sample set Each sample in the training sample set includes the standardized transaction quantity, the standardized transaction time, the standardized transaction conditions, the total information entropy, the cost cost and the transaction floor price p; Step 7.1.7: Fit the constructed linear regression model using the training sample set to obtain a fitted linear regression model; Specifically: Use a linear regression model to fit the standardized transaction quantity, transaction time, transaction conditions, total information entropy, cost, and transaction floor price p, and determine the estimated values ​​of the optimal linear regression model parameters by minimizing the residual sum of squares; Step 7.1.8: Use the fitted linear regression model to estimate the reserve price and obtain the reserve price for the reverse auction.

9. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 1 is characterized in that: Step 8 specifically includes: Step 8.1: Determine whether any candidate partner vehicle's bid is between the reserve price and the soft maximum price. If so, select one or more vehicles with the lowest bid as the winner and proceed to Step 8.

3. Otherwise, proceed to Step 8.

2. Step 8.2: When the bids of the candidate vehicles are above the soft ceiling price, the requesting vehicle will receive a prompt asking whether to raise the price cap to continue the transaction. If it chooses not to, the transaction ends. Otherwise, the process continues to determine whether the bids of the candidate vehicles are below the ceiling price. If so, the vehicle or vehicles with the lowest bids are selected as the winners and step 8.3 is executed. If all bids exceed the ceiling price, the transaction ends. Step 8.3: Use the perfect reverse auction mechanism to select the winning vehicles; Step 8.4: Use the random group reverse auction mechanism to further screen the winning vehicles and obtain the final set of winning vehicles.

10. The VANET cooperative download method based on multi-dimensional screening and encrypted reverse auction according to claim 9, characterized in that: Step 8.4 specifically includes: Step 8.4.1: The roadside unit randomly divides the winners into two groups, with each group allocated half of the roadside unit's budget; Step 8.4.2: Apply the perfect reverse auction mechanism to each group, introduce a random factor into the auction process, and determine the winning vehicle of each group after the auction ends; Step 8.4.3: Combine the two sets of auction winning vehicles to form the final set of winning vehicles.

Images