CN108874941B - Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping - Google Patents
Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping Download PDFInfo
- Publication number
- CN108874941B CN108874941B CN201810562678.7A CN201810562678A CN108874941B CN 108874941 B CN108874941 B CN 108874941B CN 201810562678 A CN201810562678 A CN 201810562678A CN 108874941 B CN108874941 B CN 108874941B
- Authority
- CN
- China
- Prior art keywords
- convolution
- url
- hash
- mapping
- big data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Information Transfer Between Computers (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a big data URL duplication removing method based on convolution characteristics and multiple hash mapping, wherein a BitSet is established by using a fast characteristic mapping algorithm based on the convolution characteristics and the multiple hash function mapping, each URL is mapped to a plurality of bits through one convolution operation and a plurality of hash functions, the collision probability is greatly reduced, and the URL in an independent web log is identified. Compared with the traditional duplicate removal calculation method, the method of the invention saves more resources, greatly reduces the probability of hash collision and has very high recognition speed.
Description
Technical Field
The invention relates to the technical field of URL (uniform resource locator) deduplication, in particular to a big data URL deduplication method based on convolution features and multiple hash mapping.
Background
One of the existing big data URL processing technologies is to store the accessed URL by HashSet, and whether a URL is accessed can be checked only by approaching the cost of O (1). The method has the following defects: the memory is consumed too much, the occupied memory is more and more along with the increase of the URLs, even if only one hundred million URLs exist, each URL is calculated into 50 characters, 5GB memory is needed, and the number of actual large data processing is much larger.
The second of the existing big data URL processing technology is that the URL is subjected to one-way hash such as MD5 or SHA-1 and then stored in a HashSet or a database, and because the length of an information summary of a character string after being processed by MD5 is only 128 bits, and the length of an information summary of the character string after being processed by SHA-1 is only 160 bits, the memory is saved by several times. This method has the following disadvantages: the hash mapping is used, so that the memory is saved by several times, but a plurality of hash table conflicts still exist in the processing process of the data volume of billions, the number of false reports reaches a non-negligible number, and the duplicate removal effect is poor.
Terminology of the related art
Convolution: in functional analysis, Convolution (Convolution) is a mathematical algorithm that generates a third function by two functions f and g, characterizing the mapping of the overlapping parts of the function f and g, which are flipped and translated.
URL: a uniform resource locator is a compact representation of the location and access method of a resource available from the internet, and is the address of a standard resource on the internet.
A hash function: the keywords in the data elements are mapped to the hash table through a certain functional relationship.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a big data URL duplicate removal method based on convolution characteristics and multiple hash mapping, which is used for solving the problems of low URL duplicate removal processing speed and poor effect of a big data WEB access log, can quickly screen out high-value independent URLs from a huge amount of data and is convenient for post processing.
In order to solve the technical problems, the invention adopts the technical scheme that:
a big data URL duplication removing method based on convolution characteristics and multiple hash mapping comprises the following steps:
step 1: extracting a WEB access log from a WEB server or traditional WAF equipment, and then screening out required domain names HOST and URLs;
step 2: performing sequence convolution and multi-hash function mapping on the URL field;
defining a convolution kernel containing a number string by self, wherein the number sequence of the convolution kernel is determined to be any one, each character string needing to be de-duplicated is determined to be a number sequence according to a mapping table, and the result of one convolution operation is the multiplication of corresponding numbers; defining a step size, and sequentially convolving the convolution kernel on the digital sequence to generate a plurality of convolution values;
setting the number k of hash functions, the size m of bit groups and the number n of added character strings, wherein the possibility of false alarm of a hash table under the condition is as follows:
setting the value of k to be k-ln 2 m/n, the lowest false alarm rate is reached:
determining the number of hash functions to be used according to the expected false alarm rate according to the formula;
and step 3: the convolution function is f, and k hash functions are used for mapping, the hash functions are h1, h2, h3, h4... hk, and the binary bit assignment principle of the convolution characteristic value and the hash function mapping value in the BitSet is as follows: initializing a bit array of m bits (each bit is initially set to be 0), wherein the convolution output and the output of each hash function are numbers (corresponding to the index of the bit array) between (0, m-1);
input x, for each hash function, compute j ═ hi (x), set m _ bit [ j ] to 1; similarly, for each convolution operation of the convolution function, c ═ f (x) is calculated, and m _ bit [ c ] is set to 1;
and 4, step 4: and according to the assigned BitSet, the assigned BitSet is used as a keyword or a label of the URL, and then the independent URL is identified.
Further, the step 1 specifically comprises:
1) filtering redundant fields in the WAF log;
2) performing label matching by using the script, and filtering the static file and the URL which does not meet the requirement of the status code;
3) and splicing the domain name HOST and the URL character string and exporting.
Further, in step 1, "filter out unnecessary fields in the WAF logs" performs filtering using the filter method in the scala language, that is, each log is first partitioned by a space character using the split method in scala, and then a filter method is customized, wherein the filter method includes the rule to be filtered.
Further, when the domain name HOST and the URL character string are spliced, a character string combination method "+" built in scala is used for splicing.
Compared with the prior art, the invention has the beneficial effects that: the method utilizes the convolution characteristic value representing the independent characteristic of the URL and the mapping of the multi-Hash function, and the generated BitSet represents the multi-character URL, so that resources are saved more than that of a traditional duplicate removal calculation method, the probability of Hash collision is greatly reduced, and the identification speed is very high.
Drawings
FIG. 1 is a flowchart illustrating a big data URL deduplication method based on convolution features and multiple hash mappings according to the present invention.
Fig. 2 is a diagram of the convolution value and hash function value mapping principle.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
According to the invention, a BitSet is established by using a fast feature mapping algorithm based on convolution features and multiple hash function mapping, and each URL is mapped to a plurality of bits through one convolution operation and a plurality of hash functions, so that the collision probability is greatly reduced, and the URLs in independent web logs are identified. The details are as follows:
first, extract the WEB access log from the WEB server or the traditional WAF device, and then screen out the required HOST and URL (uniform resource locator). The specific screening method comprises the following steps:
1. filtering redundant fields in a WAF log
The filtering can be done using the filter method in scala language. Firstly, each log is divided by a space character by using a split method in scala:
val fields=line.split("")
then, a filter method is customized, wherein the rule which needs to be filtered is contained. The rule setting method can be performed by using a scala regular matching mode:
def myfilter()
val ho ═ gov.cn ". r// case: filtering only government web sites
ho.findFirstMatchIn(host)!=None
2. Using scripts for tag matching, filtering static files and URLs that do not meet required state codes
The static file refers to a static page or file accessed by a URL, such as html, xml, js, css, and the like, and these URLs are not needed in the actual deduplication application, and are also performed by using a self-defined filter function in scala:
3. combining HOST and URL, deriving
The filtered host and the filtered URL are both in the form of character strings, and then the duplicate removal operation is performed, all URLs under a domain name need to be removed, so that the host and the URL character strings are spliced by using a character string combination method "+" built in scala:
val fields=line.split("")
val host=fields(8)
val url=fields(9)
val uRL=host+url
secondly, performing sequence convolution and multi-hash function mapping on URL fields
Description of convolution operation: a convolution kernel is customized that includes a string of numbers, wherein the length of the convolution kernel is preferably no less than 6 numbers. The number sequence of convolution kernels is arbitrary, e.g., ("453752"), but once determined cannot be modified, all subsequent convolution operations use this convolution kernel. And determining each character string needing to be deduplicated as a digit sequence according to a mapping table (for example, "abcde" is mapped to "12345"), wherein the result of one convolution operation is the multiplication of corresponding digits, for example, 123 convolution 234, and the result is 1 × 2+ 3 × 4 ═ 20. Defining a step size, the sequential convolution of the convolution kernel over the digital sequence can produce many convolution values, such as convolution kernel "123" convolved over "23456", 123 convolved with 234 to produce a value, and 123 convolved with 345 to produce a value (step size 1).
The hash mapping algorithm may use the general MD5 algorithm. Setting the number k of hash functions, the size m of bit groups and the number n of added character strings, wherein the possibility of false alarm of a hash table under the condition is as follows:
setting the value of k to be k-ln 2 m/n, the lowest false alarm rate is reached:
and determining the number of hash functions to be used according to the expected false alarm rate according to the formula.
And thirdly, the convolution function is f, k hash functions are used for mapping, the hash functions are h1, h2, h3 and h4... hk, and the principle of binary bit assignment of the convolution characteristic value and the hash function mapping value in the BitSet is shown in FIG. 2. Initializing a bit array with m bits, wherein the convolution output and the output of each hash function are numbers between (0, m-1); input x, for each hash function, compute j ═ hi (x), set m _ bit [ j ] to 1; similarly, for each convolution operation of the convolution function, c ═ f (x) is calculated, and m _ bit [ c ] is set to 1.
Note: many values are generated in the convolution process (a large step length generates few values, a small step length generates many values), the step length is defined according to the needs of the user, and each convolution value is mapped into the BitSet (only one convolution value generated by convolution is shown in an image in fig. 2).
The multiple hash function mapping process part Java code is as follows:
and fourthly, according to the assigned BitSet, the assigned BitSet can be easily used as a keyword or a label of the URL, and then the independent URL is identified. For example, if a BitSet to which a URL maps already exists, the URL is considered to be duplicated.
And fifthly, the identified partial results are as follows:
www.xxxx.com/piwik.phpaction_name=www.wdzj.com%2F%E7%A4%BC%E5%BE%B7%E8%B4%A2%E5%AF%8C%E7%BD%91%E8%B4%B7%E6%A1%A3%E6%A1%88_%E7%A4%BC%E5%BE%B7%E8%B4%A2%E5%AF%8C%E5%AE%98%E7%BD%91%E8%B5%84%E6%96%99_p2p%E5%B9%B3%E5%8F%B0%E6%A1%A3%E6%A1%88_%E7%BD%91%E8%B4%B7%E4%B9%8B%E5%AE%B6&idsite=1&rec=1&r=931653&h=23&m=31&s=47&url=https%3A%2F%2Fwww.wdzj.com%2Fdangan%2Fldcf1%2F&urlref=https%3A%2F%2Fwww.wdzj.com%2Fdangan%2Fsearch%3Ffilter%3De1-b41-n44%26show%3D1&_id=747107e1f17b5566&_idts=1521648124&_idvc=3&_idn=0&_refts=1521732597&_viewts=1521732597&_ref=https%3A%2F%2Fwww.google.com%2F&send_image=0&pdf=1&qt=0&realp=0&wma=0&dir=0&fla=0&java=0&gears=0&ag=0&cookie=1&res=1440x900&cvar=%7B%223%22%3A%5B%22www%22%2C%22%22%5D%2C%225%22%3A%5B%22uid%22%2C%220%22%5D%7D>_ms=888>-1
www.xxxx.com/m/c.ashxs=35&u=100000&c=4&P=170663&Fl=https%3A//www.google.com.hk/>3
www.xxxx.com/user/actionevent_type=load&curt_id=7f8745b8-2d39-11e8-897e-00163e131d5b&prev_id=&event_info=%7B%22ad_uuid%22%3A%22add_trwgzad1tkva%22%7D&event=ad_exposure&target=http%3A%2F%2Fwww.shixiseng.com%2Ftc%2Frpo&uuid=9f2cd019-7402-8948-9c0b-501353d6a9e5&url=https%253A%2F%2Fwww.shixiseng.com%2F&referrer=https%3A%2F%2Fwww.google.com%2F&uri=%2F&source=pc>---1
www.xxxx.com/user/actionevent_type=load&curt_id=54c06304-2d8a-11e8-97ea-00163e131d5b&prev_id=42b3df23-dfea-4651-a86f-88ba92a4e42d&event_info=%7B%22ad_uuid%22%3A%22add_77mcl4cyo2uu%22%7D&event=ad_exposure&target=%2Fcom%2Fcom_qrf1ioxwhvxk&uuid=6e65a594-9034-97f3-ad00-b1e7a46d39ca&url=https%253A%2F%2Fwww.shixiseng.com%2F&referrer=https%3A%2F%2Fwww.google.com%2F&uri=%2F&source=pc>---1
www.xxxx.com/user/actionevent_type=load&curt_id=d72ad59e-2dcc-11e8-97ea-00163e131d5b&prev_id=&event_info=%7B%22ad_uuid%22%3A%22add_5zj7701ibn7t%22%7D&event=ad_exposure&target=http%3A%2F%2Fcampus.51job.com%2Funiqlo%2F&uuid=f6878545-7dff-4c90-9ec4-d3f0b0be2cb7&url=https%253A%2F%2Fwww.shixiseng.com%2F&referrer=https%3A%2F%2Fwww.google.com%2F&uri=%2F&source=pc>---1
www.xxxx.com/user/actionevent_type=load&curt_id=dc320a40-2dd5-11e8-86b1-00163e0e0af8&prev_id=&event_info=%7B%22ad_uuid%22%3A%22add_q5h0sozpfgsg%22%7D&event=ad_exposure&target=%2Fcom%2Fcom_ohgsahcs55rv&uuid=4d6eab5f-7e0a-da6c-a70a-75cacb5b8e2f&url=https%253A%2F%2Fwww.shixiseng.com%2F&referrer=https%3A%2F%2Fwww.google.com.hk%2F&uri=%2F&source=pc>1
www.xxxx.com/user/actionevent_type=load&curt_id=0ae48218-2db9-11e8-99e6-00163e040372&prev_id=&event_info=%7B%22ad_uuid%22%3A%22add_trwgzad1tkva%22%7D&event=ad_exposure&target=http%3A%2F%2Fwww.shixiseng.com%2Ftc%2Frpo&uuid=21bf3ae9-d652-1e0a-867a-1f9c29660cd5&url=https%253A%2F%2Fwww.shixiseng.com%2F&referrer=https%3A%2F%2Fwww.google.com.hk%2F&uri=%2F&source=pc>1
Claims (4)
1. a big data URL duplication removing method based on convolution characteristics and multiple hash mapping is characterized by comprising the following steps:
step 1: extracting a WEB access log from a WEB server or traditional WAF equipment, and then screening out required domain names HOST and URLs; splicing the domain name HOST and the URL character string;
step 2: performing sequence convolution and multi-hash function mapping on the URL field;
defining a convolution kernel containing a number string by self, wherein the number sequence of the convolution kernel is determined to be any one, each character string needing to be de-duplicated is determined to be a number sequence according to a mapping table, and the result of one convolution operation is the multiplication of corresponding numbers; defining a step size, and sequentially convolving the convolution kernel on the digital sequence to generate a plurality of convolution values;
setting the number k of hash functions, the size m of bit groups and the number n of added character strings, wherein the possibility of false alarm of a hash table under the condition is as follows:
setting the value of k to be k-ln 2 m/n, the lowest false alarm rate is reached:
determining the number of hash functions to be used according to the expected false alarm rate according to the formula;
and step 3: the convolution function is f, and k hash functions are used for mapping, the hash functions are h1, h2, h3, h4... hk, and the binary bit assignment principle of the convolution characteristic value and the hash function mapping value in the BitSet is as follows: initializing a bit array with m bits, wherein the convolution output and the output of each hash function are numbers between (0, m-1); input x, for each hash function, compute j ═ hi (x), set m _ bit [ j ] to 1; similarly, for each convolution operation of the convolution function, c ═ f (x) is calculated, and m _ bit [ c ] is set to 1;
and 4, step 4: and according to the assigned BitSet, the assigned BitSet is used as a keyword or a label of the URL, and then the independent URL is identified.
2. The big data URL deduplication method based on convolution features and multiple hash mapping as claimed in claim 1, wherein the step 1 specifically is:
1) filtering redundant fields in the WAF log;
2) performing label matching by using the script, and filtering the static file and the URL which does not meet the requirement of the status code;
3) and splicing the domain name HOST and the URL character string and exporting.
3. The big data URL deduplication method based on convolution features and multiple hashes mapping as claimed in claim 2, wherein "filtering out unnecessary fields in WAF logs" in step 1 is performed by filtering in scala language using filter method, i.e. each log is first partitioned by space character using split method in scala, and then a filter method is customized, wherein the rule that the filtering is wanted is contained.
4. The big data URL deduplication method based on convolution features and multiple hashmaps as claimed in claim 3, wherein when domain name HOST and URL string are spliced, the concatenation is performed using a built-in string combination method "+" of scala.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810562678.7A CN108874941B (en) | 2018-06-04 | 2018-06-04 | Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810562678.7A CN108874941B (en) | 2018-06-04 | 2018-06-04 | Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108874941A CN108874941A (en) | 2018-11-23 |
CN108874941B true CN108874941B (en) | 2021-09-21 |
Family
ID=64335996
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810562678.7A Active CN108874941B (en) | 2018-06-04 | 2018-06-04 | Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108874941B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110334251B (en) * | 2019-07-01 | 2021-10-19 | 南京邮电大学 | Element sequence generation method for effectively solving rehash conflict |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101958883A (en) * | 2010-03-26 | 2011-01-26 | 湘潭大学 | Bloom Filter and open-source kernel-based method for defensing SYN Flood attack |
CN103106219A (en) * | 2011-11-15 | 2013-05-15 | 盛乐信息技术(上海)有限公司 | Method and system of web page grabbing |
CN104657350A (en) * | 2015-03-04 | 2015-05-27 | 中国科学院自动化研究所 | Hash learning method for short text integrated with implicit semantic features |
CN104809182A (en) * | 2015-04-17 | 2015-07-29 | 东南大学 | Method for web crawler URL (uniform resource locator) deduplicating based on DSBF (dynamic splitting Bloom Filter) |
CN105956068A (en) * | 2016-04-27 | 2016-09-21 | 湖南蚁坊软件有限公司 | Webpage URL repetition elimination method based on distributed database |
CN106295629A (en) * | 2016-07-15 | 2017-01-04 | 北京市商汤科技开发有限公司 | Structured text detection method and system |
CN106599022A (en) * | 2016-11-01 | 2017-04-26 | 中山大学 | User portrait forming method based on user access data |
CN107832476A (en) * | 2017-12-01 | 2018-03-23 | 北京百度网讯科技有限公司 | A kind of understanding method of search sequence, device, equipment and storage medium |
CN107871014A (en) * | 2017-11-23 | 2018-04-03 | 清华大学 | A kind of big data cross-module state search method and system based on depth integration Hash |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8626474B2 (en) * | 2010-04-19 | 2014-01-07 | Altera Corporation | Simulation tool for high-speed communications links |
-
2018
- 2018-06-04 CN CN201810562678.7A patent/CN108874941B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101958883A (en) * | 2010-03-26 | 2011-01-26 | 湘潭大学 | Bloom Filter and open-source kernel-based method for defensing SYN Flood attack |
CN103106219A (en) * | 2011-11-15 | 2013-05-15 | 盛乐信息技术(上海)有限公司 | Method and system of web page grabbing |
CN104657350A (en) * | 2015-03-04 | 2015-05-27 | 中国科学院自动化研究所 | Hash learning method for short text integrated with implicit semantic features |
CN104809182A (en) * | 2015-04-17 | 2015-07-29 | 东南大学 | Method for web crawler URL (uniform resource locator) deduplicating based on DSBF (dynamic splitting Bloom Filter) |
CN105956068A (en) * | 2016-04-27 | 2016-09-21 | 湖南蚁坊软件有限公司 | Webpage URL repetition elimination method based on distributed database |
CN106295629A (en) * | 2016-07-15 | 2017-01-04 | 北京市商汤科技开发有限公司 | Structured text detection method and system |
CN106599022A (en) * | 2016-11-01 | 2017-04-26 | 中山大学 | User portrait forming method based on user access data |
CN107871014A (en) * | 2017-11-23 | 2018-04-03 | 清华大学 | A kind of big data cross-module state search method and system based on depth integration Hash |
CN107832476A (en) * | 2017-12-01 | 2018-03-23 | 北京百度网讯科技有限公司 | A kind of understanding method of search sequence, device, equipment and storage medium |
Non-Patent Citations (3)
Title |
---|
BloomFilter 确定合适的 k(hash函数个数) 值;F_URY;《https://blog.csdn.net/u012400327/article/details/62222922》;20170315;第1页 * |
Character-level Convolutional Networks for Text Classification;Xiang Zhang 等;《arXiv:1509.01626v3》;20160404;第2-4页第2部分 * |
网络爬虫技术在云平台上的研究与实现;刘小云;《中国优秀硕士学位论文全文数据库 信息科技辑》;20170215(第02期);第18-21页第3.3.2节 * |
Also Published As
Publication number | Publication date |
---|---|
CN108874941A (en) | 2018-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10579831B2 (en) | Verification of data set components using digitally signed probabilistic data structures | |
US8380680B2 (en) | Piecemeal list prefetch | |
CN110162750B (en) | Text similarity detection method, electronic device and computer readable storage medium | |
US10216848B2 (en) | Method and system for recommending cloud websites based on terminal access statistics | |
US10078687B2 (en) | Deletion of elements from a probabilistic data structure | |
CN105975459B (en) | A kind of the weight mask method and device of lexical item | |
WO2020228182A1 (en) | Big data-based data deduplication method and apparatus, device, and storage medium | |
US20190179933A1 (en) | Generating sub-indexes from an index to compress the index | |
US10263784B2 (en) | Signature verification for data set components using probabilistic data structures | |
WO2017036348A1 (en) | Method and device for compressing and decompressing extensible markup language document | |
CN110750731B (en) | Method and system for removing duplicate of news public opinion | |
JP2017532655A (en) | Compress cascading style sheet files | |
CN104065736A (en) | URL redirection method, device, and system | |
US20140325327A1 (en) | Replacing problem web links using context information | |
CN111563199A (en) | Data processing method and device | |
US20210109945A1 (en) | Self-orchestrated system for extraction, analysis, and presentation of entity data | |
US20190354455A1 (en) | Skew detector for data storage system | |
CN108874941B (en) | Big data URL duplication removing method based on convolution characteristics and multiple Hash mapping | |
CN112765966B (en) | Method and device for removing duplicate of associated word, computer readable storage medium and electronic equipment | |
CN103618742A (en) | Method and system for acquiring sub domain names and webmaster permission verification method | |
US20140129543A1 (en) | Search service including indexing text containing numbers in part using one or more number index structures | |
CN104361094A (en) | Storage method and device for file in search result, and browser client | |
CN115080684B (en) | Network disk document indexing method and device, network disk and storage medium | |
US11593824B2 (en) | System and method for domain name valuation | |
CA2997922C (en) | Deletion of elements from a bloom filter |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CP02 | Change in the address of a patent holder | ||
CP02 | Change in the address of a patent holder |
Address after: 9/F, Block C, No. 28 Tianfu Avenue North Section, Chengdu High tech Zone, China (Sichuan) Pilot Free Trade Zone, Chengdu City, Sichuan Province, 610000 Patentee after: CHENGDU KNOWNSEC INFORMATION TECHNOLOGY Co.,Ltd. Address before: 610000, 11th floor, building 2, No. 219, Tianfu Third Street, hi tech Zone, Chengdu, Sichuan Province Patentee before: CHENGDU KNOWNSEC INFORMATION TECHNOLOGY Co.,Ltd. |