Method for evaluating ip geolocation database based on urban latency characteristics

Abstract

The application provides an IP geolocation database evaluation method based on city delay characteristics, which evaluates IP geolocation databases from the perspective of cities by constructing a high-quality fusion reference database. First, a minimum network segment matching mechanism is proposed to fuse network segments of multiple IP geolocation databases, and candidate cities of the fused network segments are screened; then, through self-defined city boundary delay range BDRC and city delay characteristics CDC and other judgment indexes, positioning judgment is performed on the candidate cities of the fused network segments to obtain a fusion reference database; finally, based on the fusion reference database and two reliability indexes, the IP geolocation database is evaluated in multiple levels and in the full range.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of IP geolocation database evaluation, in particular to an IP geolocation database evaluation method based on city time delay characteristics. BACKGROUND

[0002] IP geolocation service is a basic network service of the Internet, which has important theoretical significance and application value in many fields such as personalized advertising, cloud service supervision, screening of sensitive network entities, Internet illegal behavior evidence collection, deployment strategy of network infrastructure, and fault node discovery. At present, most Internet users obtain the positioning information of IP address by querying the existing IP positioning database, but how to choose a suitable positioning database from many IP geolocation databases has become a difficult problem. Most of the current IP positioning databases do not disclose the construction method of the database, and many databases also have problems such as small number of covered cities, serious lack of city-level / county-level positioning information, and low positioning accuracy. The current IP geolocation database evaluation method is either limited by the data size of the verification set, the evaluation range is not comprehensive, and the IP geolocation database cannot be comprehensively evaluated, or the evaluation dimension is small, and the reliability of the evaluation result is seriously dependent on other IP geolocation databases, and the credibility is low. SUMMARY

[0003] The present application proposes an IP geolocation database evaluation method based on city time delay characteristics to solve the technical problems of the current IP geolocation database evaluation method, such as non-comprehensive evaluation range, less evaluation dimension, and low credibility of evaluation result. The method uses the characteristics that IP addresses have different time delay distributions in different cities, proposes Boundary time Delay Range of Cities (BDRC) and City time Delay Characteristic (CDC), and then based on the minimum network segment matching mechanism, the network segments of multiple IP geolocation databases are fused, and the candidate cities of the fused network segment are obtained. Based on the city time delay characteristics, the fused network segment is positioned and judged, and a fused reference database with high credibility and full coverage of cities is constructed. Finally, the reliability of commercial IP geolocation database is evaluated based on the fused reference database.

[0004] Therefore, the technical scheme of the present application is an IP geolocation database evaluation method based on city time delay characteristics, and the specific steps are as follows:

[0005] Step 1, constructing city IPsame address set, selecting IP addresses with the same positioning from multiple well-known commercial IP geolocation databases as IPsame data set, and then dividing by city to construct city IPsame data set;

[0006] Step 2, select the appropriate detection point for the target city, select multiple detection points in the global, calculate the distance between the target city and all detection nodes based on latitude and longitude, select the nearest detection point as the time delay detection point of the target city;

[0007] Step 3, obtain the city delay characteristics of the target city, based on the divided detection points, measure the IPsame data set of the target city actively, obtain the original delay data of the city, and then use the method of truncation and low frequency elimination to clean the original delay data, extract the city boundary delay range BDRC and city delay characteristics CDC;

[0008] Step 4, network segment fusion and candidate city screening, rearrange all network segments in the IP geographic positioning library in the form of first IP-end IP, then arrange in ascending order, based on the minimum network segment matching mechanism, perform network segment fusion, and then store in the Fusion_before database, at the same time, record all positioning cities of the fusion network segment in different IP positioning library as candidate cities;

[0009] Step 5, positioning city judgment of fusion network segment, for routable network segment, based on city boundary delay range BDRC and city delay characteristics CDC, positioning judgment is performed; for non-routable network segment, positioning judgment is performed based on city delay characteristic value, and a fusion reference database is constructed;

[0010] Step 6, overall evaluation and city-level evaluation of IP geographic positioning library, by comparing with the fusion reference database, the positioning information of the IP positioning library is evaluated in multiple aspects;

[0011] In step 3, the reserved delay T' = {t1, t2, …, t m} after data cleaning is taken as the city boundary delay range BDRC, and the occurrence probability p of each delay in the city boundary delay range BDRC is taken together with the corresponding delay to construct the city delay characteristics, which is expressed as:

[0012] CDC = [{t1, p1}, {t2, p2}, …, {t m : p m}]

[0013] Where, t n represents different delays appearing in the reserved delay T', and p n represents the frequency of the delay;

[0014] In step 4, one network segment is taken out from each IP geolocation database respectively, and network segment fusion based on minimum network segment matching mechanism is performed, and the network segment fusion is performed in a multi-round iteration comparison manner to select the minimum network segment, and the minimum network segment is the fusion network segment;

[0015] In the first round of comparison, the initial first IP is the same, the network segment Fk-Ek with the minimum end IP is stored in the Fusion_before database from all the network segments compared in the first round, and the positioning city of the network segment in all IP geolocation databases is recorded as the candidate positioning city;

[0016] The minimum network segment Fk-Ek in the first round of comparison is removed, and the next network segment is read from the IP geolocation database providing the minimum network segment Fk-Ek to participate in the next round of comparison, and the first IP of the remaining network segment is replaced by Ek+1 to form a new network segment to participate in the next round of comparison, and the process is repeated until the minimum network segment matching and candidate city screening of all network segments are completed;

[0017] In step 5, the candidate cities in the fusion network segment are positioned and determined based on the city boundary delay range BDRC and the city delay characteristic CDC, and the network segment delay used for positioning judgment of each candidate city in the fusion network segment is obtained by the detection point corresponding to the candidate city, and the network segment delay can only be used for positioning judgment of the candidate city in the network segment,

[0018] For a routable network segment, the network segment delay is the mode delay of all routable IPs; for a non-routable network segment, the network segment delay is the average delay of the last visible IP address of the traceroute path corresponding to all IPs in the network segment;

[0019] For a routable network segment, if the network segment delay is in the BDRC of the corresponding candidate city, the candidate city is considered as the actual positioning city; if there are multiple candidate cities satisfying the BDRC condition, the candidate city with the maximum probability p in the CDC is selected; if the network segment delay does not satisfy the BDRC requirement of any candidate city, further positioning is performed, and the fusion network segment accurately determined is stored in the Fusion database;

[0020] For a non-routable network segment and the network segment not satisfying the BDRC condition, the network segment most adjacent to the network segment is found from the Fusion database, and the positioning city of the most adjacent network segment is selected as a candidate city and supplemented into the candidate city list of the network segment;

[0021] For a non-routable network segment and the network segment not satisfying the BDRC condition, the city delay characteristic value of the candidate city of the network segment is calculated City delay characteristic value is referred to as: the weighted average of the city delay value t and the occurrence probability p in the city delay characteristic CDC = [{t1, p1}, {t2, p2},..., {tn, pn}], which is expressed as:

[0022]

[0023] wherein, represents the city delay characteristic value of the candidate city, representing the expected delay from a specific probe point to a specific city;

[0024] For the non-routable network segment and the network segment that does not satisfy the BDRC condition, the delay difference between the network segment delay t and the city delay characteristic value of the corresponding candidate city The candidate city with the smallest delay difference is selected as the positioning city of the network segment.

[0025] Preferably, in step 1, the city IPsame data set is obtained in units of cities, and the IP geolocation library containing the target city is compared to select the IP addresses that are all located in the same city as the city IPsame data, which is expressed as:

[0026] IPsame←Screen(DB n ,IP,City)(1)

[0027] wherein DB n represents n IP location libraries containing the target city, City represents the target city, and IP represents the IP address located in the target city in the n IP location libraries.

[0028] Preferably, in step 2, a plurality of probe points are selected globally, and the longitude and latitude (Lo, La) of the city where the probe point is located are recorded. For any target city i, the longitude and latitude are (Lo i ,La i ), based on the longitude and latitude, the Euclidean distance to all probe points is calculated, and the probe point with the shortest distance is selected as the delay probe point of the city, which is expressed as:

[0029]

[0030] wherein P represents the probe point, and [1, k] represents the first probe point to the kth probe point.

[0031] Preferably, in step 3, based on the partitioned probe points, the IPsame data set of the target city is actively measured to obtain the original delay data T = {t1, t2,..., t nThen, the original delay data is cleaned by the head and tail cutting method and the low frequency elimination method to obtain reliable delay data for constructing the city boundary delay range (BDRC) and the city delay characteristic (CDC) of the target city.

[0032] Preferably, in step 3, the head and tail cutting method refers to taking the mode t n in the city original delay T = {t1, t2, …, t w as the center, eliminating the delay less than a × t and the delay greater than b × t w , and only keeping the delay information in the middle range.

[0033] The low frequency elimination method refers to counting the frequency f of all delays in the remaining delay range after the head and tail cutting operation, and then eliminating the delay less than a × f.

[0034] Wherein, a and b are both positive real parameters.

[0035] Preferably, the parameters a and b are determined through multiple experiments, and the delay distribution range less than 100 ms and the original data retention ratio of 90% are taken as the measurement standards. Finally, the parameters a = 20 and b = 1000 are determined.

[0036] Preferably, in step 6, the IP location library is compared with the fusion IP geographic location library (Fusion) to count the number of IP IP consistent that are consistent in positioning, and calculate the overall reliability of the IP location library. The overall reliability refers to the ratio of the number of IP IP consistent that are consistent in positioning between the IP location library and the Fusion database to the total number of IP IP F_all of the fusion IP location library, which is expressed as:

[0037]

[0038] Wherein, Reliability represents the overall reliability of the IP geographic location library at the city level positioning granularity, which is used to evaluate the overall IP geographic location library.

[0039] The city reliability of the IP geographic location library in the city is calculated, which refers to the ratio of the number of IP IP consistent that are consistent in positioning between the IP location library and the Fusion database in the city to the total number of IP IP S_all of the IP location library in the city, which is expressed as:

[0040]

[0041] ​City_Reliability = (IPsame / IPtotal) * 100

[0042] The beneficial effect of the present application is that a fusion reference database with a large number of network segments, full city coverage, high city-level coverage of network segments, and reliable city-level positioning results is constructed based on the minimum network segment matching mechanism and city delay characteristics. This is a method that can simultaneously locate and judge routable network segments and non-routable network segments. Compared with the IP geolocation library, the proposed overall reliability and city reliability indicators provide a multi-level and highly reliable evaluation of the IP geolocation library. In addition, the fusion method of the IP geolocation library proposed in the present application can effectively increase the number of cities, the number of network segments, the city-level coverage of IP addresses, and the city-level positioning accuracy of the IP geolocation library, and has a significant contribution to improving the city-level geographic positioning of IP addresses. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is the overall flowchart of the IP geolocation library evaluation method based on city delay characteristics of the present application;

[0044] Figure 2 is the calculation flowchart of city delay characteristics of the present application;

[0045] Figure 3 is the construction flowchart of the fusion reference database;

[0046] Figure 4 is the overall reliability evaluation diagram of the four experimental IP geolocation libraries;

[0047] Figure 5 is the city reliability evaluation diagram of the four experimental IP geolocation libraries. DETAILED DESCRIPTION

[0048] The present application will be further described below in conjunction with examples.

[0049] Figures 1-5 is an embodiment of the IP geolocation library evaluation method based on city delay characteristics of the present application, and the specific steps are as follows:

[0050] Step 1, constructing the city IPsame dataset. Select multiple well-known commercial IP geolocation libraries, compare the IP addresses located in the destination city, and select the same IP addresses as the city IPsame dataset. If a certain IP geolocation library does not contain the city, the database is excluded and then compared.

[0051] In step 1, the city IPsame dataset is obtained in units of cities, and the IP geolocation library containing the target city is compared to select IP addresses that are all located in the city as the city IPsame data, denoted as:

[0052] IPsame←Screen(DB n ,IP,City)(1)

[0053] wherein DB n contains n IP geolocation libraries of the target city, City represents the target city, and IP represents IP addresses located in the target city in the n IP geolocation libraries. Since different IP geolocation libraries cover different numbers of cities, the source IP geolocation library for constructing the IPsame dataset of different cities is different. The construction process of the city IPsame dataset is a dynamic process, and IP geolocation libraries that do not contain the city do not participate in the construction of the city IPsame dataset.

[0054] Step 2, selecting appropriate probe points for the target city. Select multiple probe points globally, calculate the distance between the target city and all probe nodes based on latitude and longitude, and select the closest probe point as the time delay probe point of the target city.

[0055] In step 2, multiple probe points are selected globally, and the latitude and longitude (Lo, La) of the city where the probe point is located are recorded. For any target city i, its latitude and longitude are (Lo i ,La i ), the Euclidean distance to all probe points is calculated based on latitude and longitude, and the closest probe point is selected as the time delay probe point of the city. Denoted as:

[0056]

[0057] wherein P represents the probe point, and [1, k] represents the first probe point to the kth probe point.

[0058] Step 3, obtaining the city time delay characteristics of the target city. From a specific probe point, the IPsame dataset of the target city is actively measured to obtain the original time delay data of the city. Then, the original time delay data is cleaned by using the method of truncation and tail elimination and low frequency elimination, and the city boundary delay range BDRC and city delay characteristics CDC are extracted.

[0059] In step 3, based on the partitioned probe points, the IPsame dataset of the target city is actively measured to obtain the original time delay data T = {t1, t2, …, t nThen, the data cleaning of the original delay data is performed by the head and tail cutting method and the low frequency rejection method to obtain reliable delay data for constructing the boundary delay range of the target city (BDRC) and the city delay characteristics (CDC).

[0060] The head and tail cutting method is to take the mode t n in the city original delay T = {t1, t2, …, t w as the center, remove the delay less than a × t w , and only keep the delay information in the middle range. It is expressed as:

[0061]

[0062] wherein Cut_HT represents the delay range retained after the head and tail cutting method. The purpose of this step is to narrow the delay range of the city, extract the typical delay characteristics of the city, and improve the calculation efficiency.

[0063] The low frequency rejection method is to count the frequency f of all delays in the remaining delay range after the head and tail cutting operation, and then remove the delay with f , which is expressed as:

[0064]

[0065] wherein f tn represents the frequency of the delay tn, f all represents the total number of all delays, and b represents the threshold of the low frequency rejection method.

[0066] In step 3, the parameters a and b are determined through multiple experiments, taking the delay distribution range less than 100 ms and the retention ratio of the original data of 90% as the measurement standard. Finally, the parameters a = 20 and b = 1000 are determined.

[0067] In step 3, in order to avoid the distortion of the city delay characteristics caused by too few reliable delays of the city, the cities with less than 10 reliable delays are re-screened by IPsame. The screening method is to remove the IP positioning library that provides the least number of IPs in the city, and then reconstruct the city IPsame dataset.

[0068] In step 3, the retained delay T′ = {t1, t2, …, t m after data cleaning is taken as the boundary delay range BDRC of the city. The occurrence probability p of each delay in the city boundary delay range BDRC is constructed together with the corresponding delay to construct the city delay characteristics, which is expressed as:

[0069] ​​CDC = [{t1, p1}, {t2, p2},..., {t m m}]

[0070] where t n represents different time delays appearing in the reserved time delay T', and p n represents the frequency of the time delay.

[0071] Step 4, network segment fusion and candidate city screening. All network segments in the IP positioning library are rearranged in the form of first IP-end IP, and then sorted in ascending order. Network segment fusion is performed based on the minimum network segment matching mechanism, and the fused network segment and all positioning cities in different IP positioning libraries are stored in the Fusion_before database. The fused network segment can have multiple candidate cities.

[0072] In step 4, a network segment is taken out from all IP geographic positioning libraries respectively, and network segment fusion based on the minimum network segment matching mechanism is performed. Network segment fusion uses multiple rounds of iteration comparison to select the minimum network segment, which is the fused network segment.

[0073] In the first round of comparison, the initial first IP is the same. From all network segments compared in the first round, the network segment Fk-Ek with the minimum end IP is stored in the Fusion_before database, and the positioning cities of the network segment in all IP geographic positioning libraries are recorded as candidate positioning cities.

[0074] The minimum network segment Fk-Ek in the first round of comparison is removed, and the next network segment from the IP geographic positioning library providing the minimum network segment Fk-Ek is read to participate in the next round of comparison. The first IP of the remaining network segments is replaced by Ek+1 to form a new network segment to participate in the next round of comparison. This process is repeated until the minimum network segment matching and candidate city screening of all network segments are completed.

[0075] Step 5, positioning city judgment of the fused network segment. For routable network segments, positioning determination is made based on the city boundary delay range BDRC and the city delay characteristic CDC; for non-routable network segments, positioning determination is made based on the city delay characteristic value.

[0076] Based on BDRC and CDC, the candidate cities in the fused network segment are positioned. The network segment delay used for positioning judgment of each candidate city in the fused network segment is obtained by the probe point corresponding to the candidate city, and the network segment delay can only be used for positioning judgment of the candidate city in the network segment.

[0077] ​For routable network segment, the segment latency is the mode latency of all routable IPs; for non-routable network segment, the segment latency is the average latency of the last visible IP address of the traceroute path of all IP pairs in the network segment.

[0078] For routable network segment, if the segment latency is in the BDRC of the corresponding candidate city, the candidate city is considered as the actual positioning city; if there are multiple candidate cities satisfying the BDRC condition, the candidate city with the maximum probability p in the CDC is selected; if the segment latency does not satisfy the BDRC requirement of any candidate city, further positioning is performed. The fusion network segment that is accurately determined is stored in the Fusion database.

[0079] For non-routable network segment and other network segments that do not satisfy the BDRC condition, the most adjacent network segment to the network segment is found from the Fusion database, and the positioning city of the most adjacent network segment is selected as a candidate city to supplement the candidate city list of the network segment.

[0080] For non-routable network segment and other network segments, the city latency characteristic value of the candidate city of the network segment is calculated City latency characteristic value is referred to as the weighted average of the city latency value t and the occurrence probability p in the city latency characteristic CDC = [{t1, p1}, {t2: p2},..., {tn: pn}], and is represented as:

[0081]

[0082] wherein, represents the city latency characteristic value of the candidate city, representing the expected latency of a specific probe point to a specific city.

[0083] For non-routable network segment and other network segments, the latency difference between the segment latency t and the city latency characteristic value of the corresponding candidate city is calculated, and the candidate city with the smallest latency difference is selected as the positioning city of the network segment, represented as:

[0084]

[0085] wherein, t k represents the segment latency of the kth candidate city, represents the city latency characteristic value of the kth candidate city.

[0086] Step 6, reliability evaluation of IP geolocation library. Through comparison with the fusion reference database, the positioning information of the IP positioning library is evaluated in multiple aspects for reliability.

[0087] In step 6, the IP location database is compared with the Fusion IP geolocation database, and the number of IP addresses IP consistent that are consistent in positioning is counted to calculate the overall reliability of the IP location database. The overall reliability is the ratio of the number of IP addresses IP consistent that are consistent in positioning between the IP location database and the Fusion database to the total number of IP addresses IP F_all of the IP location database, and is expressed as:

[0088]

[0089] where Reliability represents the overall reliability of the IP geolocation database at the city level, and is used to evaluate the overall performance of the IP geolocation database.

[0090] The city reliability of the IP geolocation database in a city is calculated. The city reliability is the ratio of the number of IP addresses IP consistent that are consistent in positioning between the IP location database and the Fusion database to the total number of IP addresses IP S_all of the IP location database in the city, and is expressed as:

[0091]

[0092] where City_Reliability represents the city reliability of the IP geolocation database in the city, and is used to evaluate the reliability of the positioning information provided by the IP geolocation database in the city, providing a reference for users using the IP geolocation database in the city, and completing the evaluation of the IP geolocation database.

[0093] The city reliability City_Reliability can further evaluate the performance of the IP geolocation database in city clusters, provinces, and states, and the overall reliability Reliability can evaluate the reliability of the entire database in the country, continent, and entire database.

[0094] An IP geolocation database evaluation method based on city latency characteristics is provided, taking the IP UU, IP2Location Lite, GeoLite2, and IP2Region four known IP geolocation databases as examples for further illustration.

[0095] Step 1, construct a city IP same address set.

[0096] First, as Figure 2As shown, IP addresses locating the target city are found from four databases, and the IP addresses that are consistent across all databases are selected as the IPsame address set for the target city. If a certain IP geolocation database does not contain the target city, it is not included in the construction of the IPsame address set for that city. This is represented as: IPsame ← Screen(DB) n IP, City)

[0097] Step 2: Select a suitable detection point for the target city.

[0098] Multiple detection points are selected globally. For any city, the distance between it and all detection nodes is calculated based on latitude and longitude. The detection point with the closest distance is selected as the time delay detection point for that target city.

[0099] Step 3: Obtain the urban latency characteristics of the target city.

[0100] like Figure 2 As shown, IPsame data for each city is actively measured from specific probe points. Cities with fewer than 10 reliable latency values ​​are filtered out, and the IPsame dataset is retrained. The training method involves removing the IP geolocation database with the fewest IPs provided to that city and then obtaining a new IPsame dataset.

[0101] Take a = 20, and with the mode of delay W as the center, remove delays less than 0.05*W and delays greater than 20*W, and only retain the delay information in the middle range.

[0102] Take b = 1000, calculate the frequency of all delays within the reasonable delay range, and then remove delays with a frequency lower than 0.001.

[0103] The cleaned latency dataset is used as the city's reliable latency dataset, and the latency range of this dataset is the city's latency range (BDRC). The frequency of occurrence of reliable latency is calculated, which constitutes the city's latency feature (CDC).

[0104] Step 4: Network segment fusion and candidate city selection.

[0105] like Figure 3 As shown, the network segments in the four selected experimental databases IPUU, IP2LocationLite, GeoLite2, and IP2Region are sorted in ascending order. Then, the smallest network segment is searched for in each database. The network segment fusion adopts a multi-round iterative comparison method to filter the smallest network segment. The smallest network segment is divided by the smallest network segment matching mechanism and then stored in the 'Fusion_before' database.

[0106] At the same time, record all the positioning cities of the fusion network segment in the four given IP positioning databases as candidate cities.

[0107] Step 5, positioning city judgment of the fusion network segment.

[0108] For the network segment with routable IP, take the mode delay of the network segment as the network segment delay. If the network segment delay is within the city delay range BDRC of the candidate city, it is considered that the candidate city is the correctly positioned city of the network segment; if there are multiple candidate cities that meet the BDRC condition at the same time, select the city with the maximum probability p in the city delay characteristic CDC.

[0109] If the network segment delay is not within the city delay range BDRC of all candidate cities, it is considered that the network segment is incorrectly positioned in the four databases and needs to be repositioned.

[0110] For the non-routable network segment and other network segments that do not meet the BDRC condition, find the most adjacent network segment to the network segment from the ‘Fusion’ database, and select the positioning city of the most adjacent network segment as a candidate city to supplement into the candidate city list of the network segment.

[0111] For the non-routable network segment and other network segments, calculate the city delay characteristic value of the candidate city of the network segment and the corresponding network segment delay. The network segment delay of the non-routable network segment is composed of the average value of the last IP in all IP paths in the network segment.

[0112] For the non-routable network segment and other network segments, calculate the delay difference of the network segment delay t to the city delay characteristic value of the corresponding candidate city, and select the candidate city with the smallest delay difference as the positioning city of the network segment, represented as:

[0113] wherein t k represents the network segment delay of the kth candidate city, represents the city delay characteristic value of the kth candidate city.

[0114] Step 6, overall evaluation and city-level evaluation of the IP geographic positioning database.

[0115] At the overall level, for the four given example positioning databases, compare each IP positioning database with the fusion IP geographic positioning database ‘Fusion’, count the number of consistent IP positions IP consistent , and calculate the overall reliability Reliability of the IP positioning database.

[0116] At the city level, the city reliability City_Reliability of each city is calculated for four different IP geolocation databases in the city.

[0117] As shown in Figures 4-5 By comparing the overall reliability of IPUU, IP2LocationLite, GeoLite2 and IP2Region calculated by analysis and the city reliability in some cities, the IP geolocation database can be evaluated and analyzed.

[0118] The above method can comprehensively integrate the advantages of each IP positioning database in the number of network segments, the number of covered cities, the city coverage rate of network segments and the positioning accuracy, construct a high-reliability fusion IP positioning database covering the whole IPv4 space, 71915 city areas and 99.99% of the city coverage rate of network segments, and effectively improve the accuracy and coverage of the IP positioning database at the city level positioning granularity, which plays a significant role in improving the accuracy of IP positioning.

[0119] The application realizes the positioning and judgment functions of routable and non-routable network segments at the same time, which is a problem that most evaluation methods cannot handle, and the evaluation method can effectively evaluate the reliability of each IP geolocation database in specific cities, city groups, countries, continents and the whole database, and realize the functions of evaluating IP full coverage, multi-level evaluation range and high-reliability evaluation results.

[0120] The above is only a specific embodiment of the application, and cannot limit the scope of the application, so the replacement of equivalent components or equivalent changes and modifications made within the scope of the patent protection of the application should still fall within the scope of the claims of the application.

Claims

1. A method for evaluating IP geolocation database based on urban latency features, characterized in that, The specific steps are as follows: Step 1, constructing a city IP same address set, screening IP addresses with the same location from multiple well-known commercial IP geographic positioning libraries as the IP same data set, then dividing by city to construct a city IP same data set; Step 2, selecting a suitable detection point for the target city, selecting multiple detection points globally, calculating the distance between the target city and all detection nodes based on latitude and longitude, and selecting the closest detection point as the time delay detection point of the target city; Step 3, obtaining the city time delay characteristics of the target city, based on the divided detection points, actively measuring the IP same data set of the target city, obtaining the original time delay data of the city, and then using the methods of truncation and low frequency elimination to clean the original time delay data, extracting the city boundary delay range BDRC and the city delay characteristic CDC; Step 4, network segment fusion and candidate city screening, rearranging all network segments in the IP geographic positioning library in the form of first IP-end IP, then ascending arrangement, network segment fusion based on the minimum network segment matching mechanism, then storing in the Fusion_before database, and recording all positioning cities of the fusion network segment in different IP positioning libraries as candidate cities; Step 5, positioning city judgment of the fusion network segment, for routable network segments, positioning judgment based on city boundary delay range BDRC and city delay characteristic CDC; for non-routable network segments, positioning judgment based on city delay characteristic value, constructing a fusion reference database; Step 6, overall evaluation and city-level evaluation of IP geographic positioning library, comparing with the fusion reference database to evaluate the positioning information of the IP positioning library from multiple aspects; In step 3, the remaining time delay T' = {t1, t2, …, t m} after data cleaning is taken as the boundary delay range BDRC of the city. The occurrence probability p of each time delay in the boundary delay range BDRC is constructed together with the corresponding time delay to form the city delay feature, denoted as: CDC = [{t1,p1},{t2:p2},...,{t m :p m}] where t n denotes the different delays occurring in the reserved delay T', p n denotes the frequency of occurrence of the delay; In step 4, a network segment is taken from all IP geographic positioning libraries, and network segment fusion based on the minimum network segment matching mechanism is performed, and the minimum network segment is selected by multiple rounds of iteration comparison, and the minimum network segment is the fusion network segment; In the first round of comparison, the initial first IP is the same, from all network segments compared in the first round, the network segment Fk-Ek with the minimum end IP is stored in the Fusion_before database, and the positioning cities of the network segment in all IP geographic positioning libraries are recorded as candidate positioning cities; The minimum network segment Fk-Ek in the first round of comparison is removed, and the next network segment is read from the IP geographic positioning library providing the minimum network segment Fk-Ek to participate in the next round of comparison, and the first IP of the remaining network segments is replaced by Ek+1 to form a new network segment to participate in the next round of comparison, and so on, until the minimum network segment matching and candidate city screening of all network segments are completed; In step 5, the candidate cities in the fusion network segment are positioned based on the city boundary delay range BDRC and the city delay characteristic CDC, and the network segment delay of each candidate city in the fusion network segment is obtained by the corresponding probe point, which can only be used for positioning of the candidate city in the network segment, For a routable network segment, the network segment delay is the mode delay of all routable IPs; for a non-routable network segment, the network segment delay is the average delay of the last visible IP address of the traceroute path corresponding to all IPs in the network segment; For a routable network segment, if the network segment delay is within the BDRC of the corresponding candidate city, the candidate city is considered to be the actual positioning city; if there are multiple candidate cities satisfying the BDRC condition, the candidate city with the maximum probability p in the CDC is selected; if the network segment delay does not satisfy the BDRC requirement of any candidate city, further positioning is performed, and the fusion network segment is stored in the Fusion database; For a non-routable network segment and the network segment not satisfying the BDRC condition, the most adjacent network segment to the network segment is found from the Fusion database, and the positioning city of the most adjacent network segment is selected as a candidate city and supplemented into the candidate city list of the network segment; For the non-routable network segment and the network segment not satisfying the BDRC condition, calculate the city delay characteristic value of the candidate city of the network segment city delay characteristic value The city delay characteristic CDC is referred to as the weighted average value of the city delay value t and the occurrence probability p in the city delay characteristic CDC = [{t1, p1}, {t2, p2},..., {tn, pn}], and is expressed as: wherein, Ci represents the city latency eigenvalue of the candidate city, representing the expected latency from a specific probe point to a specific city. For the non-routable network segment and the network segment not satisfying the BDRC condition, the network segment delay t is calculated to the eigenvalue of the city delay of the corresponding candidate city The candidate city with the minimum delay difference is selected as the positioning city of the network segment.

2. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 1, characterized in that, In step 1, the city IPsame data set is obtained in units of cities, and the IP geolocation library containing the target city is compared to select the IP addresses with the same positioning in the city as the city IPsame data, represented as: IPsame <- Screen(DB n , IP, City) (1) Wherein, DB n represents n IP positioning libraries containing the target city, City represents the target city, and IP represents IP addresses positioned in the target city in the n IP positioning libraries.

3. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 1, characterized in that, In step 2, a plurality of probe points are selected globally, and the longitude and latitude (Lo, La) of the city where each probe point is located is recorded. For any target city i, the longitude and latitude thereof is (Lo i ,La i ). The Euclidean distance from the target city to all probe points is calculated based on the longitude and latitude, and the probe point closest to the target city is selected as the time delay probe point of the target city, denoted as: Where P represents a probe point, and [1, k] represents the first probe point to the kth probe point.

4. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 1, characterized in that, In step 3, based on the partition-based probe points, the IPsame dataset of the target city is actively measured to obtain the original delay data T = {t1, t2, …, t n} of the city, and then the data cleaning of the tail-cutting method and the low-frequency elimination method is performed on the original delay data to obtain reliable delay data used to construct the city boundary delay range BDRC and the city delay characteristic CDC of the target city.

5. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 4, characterized in that, In step 3, the truncated tail method refers to taking the mode t n in the original urban delay T = {t1, t2, …, t w as the center, eliminating the delay less than a and the delay greater than a × t w , and only keeping the delay information in the middle range. The low frequency elimination method is to count the frequency of all the time delays in the remaining time delay range after the "truncation" operation, and then eliminate the time delay with the frequency f. ​ Where a and b represent a positive real number parameter.

6. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 5, characterized in that, After multiple rounds of experiments, the delay distribution range less than 100 ms and the original data retention ratio of 90% are used as the measurement standards, and finally the parameters a = 20 and b = 1000 are determined.

7. The IP geolocation database evaluation method based on urban time delay characteristics according to claim 1, characterized in that, In step 6, the IP positioning database is compared with the Fusion IP geolocation database, and the number of IP addresses that are consistent in positioning is counted IP consistent The overall reliability of the IP positioning database is calculated, which is the ratio of the number of IP addresses that are consistent in positioning between the IP positioning database and the Fusion database IP consistent to the total number of IP addresses in the Fusion IP positioning database IP F_all , and is expressed as: Where Reliability represents the overall reliability of the IP geolocation library at the city level positioning granularity, which is used to evaluate the overall reliability of the IP geolocation library. In the city unit, the city reliability of the IP geolocation database in the city is calculated, which is the ratio of the number of IP addresses that the IP geolocation database is consistent with the Fusion database positioning in the city IP consistent to the total number of IP addresses of the IP geolocation database IP S_all , which is expressed as: Where City_Reliability represents the city reliability of the IP geolocation library in the city.

Images