JP2009104400A - Email filtering device, method for filtering email, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically filter out spam when email are received, and reduce incorrect decision. <P>SOLUTION: Email received by a user are sorted into normal mail and spam by predetermined filter rules, and features of header information of the sorted email are collated and extracted to generate feature vectors. The feature vectors of the normal mail and spam are analyzed to generate initial filter rules, and received email are separated into normal mail and spam by the generated filter rules. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a filtering device at the time of receiving an e-mail, and more particularly to an e-mail filtering device that automatically filters spam mail when an e-mail is received, an e-mail filtering method, and a program.

  In recent years, with the development of networks, anyone can easily and easily send and receive e-mail (hereinafter simply referred to as “mail” as necessary), so-called spam mail (spam mail) The number is also increasing. Here, “spam mail” means mail that is sent indiscriminately and in large quantities without ignoring the recipient's intention and without prior request or consent. Synonyms for this spam mail include “junk mail”, “junk mail”, “UCE (Unsolicited Commercial Email)”, “UBE (Unsolicited Bulk Email)”, and the like.

  Such spam e-mails include virus infections due to attachments, etc., decrease in recipient business productivity and efficiency due to an increase in unnecessary e-mails, increased load on servers and networks due to increased traffic, and fraudulent sites It can be a threat regardless of individuals or organizations in terms of privacy due to guidance or leakage of confidential information.

  The problems caused by spam emails as described above have already reached the level of social problems. In view of the fact that email addresses are available at low cost and that high-speed communications with a flat rate are provided at a low price, spam emails are expected to increase without decreasing in the future. Effective measures against email are urgently required.

  In such a situation, regarding the filtering method of the email stored in the mail server, the terminal having the filtering condition related to the email transmission address and the reply address is used to store the header information of the email stored in the mail server. A technique for requesting and determining header information of an email received based on a filtering condition, sending a filtering instruction based on the determination result to the mail server, and executing a filtering process based on the filtering instruction by the mail server is known. (For example, refer to Patent Document 1).

  In addition, when it is determined whether or not the mail is a junk mail based on the junk mail frequency information of the identified relay device on the mail distribution route, and the e-mail delivered through the distribution route is a junk mail In addition, a technique for reliably detecting this is also known (for example, see Patent Document 2).

In addition, a blacklist-based filter that determines spam mail by referring to the spam mail sender IP address in the database, a keyword-based filter that determines spam mail by matching keywords in the mail, and Bayes' theorem are applied. The spam mail is judged based on the statistical processing. It learns whether a specific word, word group, byte stream, etc. are included, sorts spam mail from regular mail, Bayesian filter, spam mail based on rules created manually using mail containing header A technique for distributing mails with high accuracy by combining a plurality of filter techniques such as a filter based on a heuristic rule to be determined is also known (see Non-Patent Document 1).
JP 2004-171169 A JP 2006-260515 A Spam Assassin, http: // spamassin. apache. org /

  However, in the technique described in Patent Document 1, since the filtering condition of the system is that the e-mail transmission address and the reply address do not match, even if it is a normal mail distributed on the mailing list The above-mentioned conditions may be satisfied, and there is a problem that erroneous detection may occur.

  In the technique described in Patent Document 2, since only the address of the relay device is used, it functions when the transmission source is constant like a spammer, but an unspecified address such as a botnet. When using, there is a problem that it may not function.

  Furthermore, in the technology described in Patent Document 1, the black list is not effective due to the transmission of spam mail from the botnet, or the keyword database filter is passed by the transition of the keyword. When embedded in an image, normal mail is determined to be spam mail or spam mail is determined to be normal mail for reasons such as passing through a filter based on keyword databases, Bayesian, and heuristic rules. There's a problem.

  Therefore, the present invention has been made in view of the above circumstances, and automatically filters spam mail when receiving an email, and reduces an erroneous determination of spam mail, and email filtering. An object is to provide a method and a program.

  The present invention proposes the following items in order to solve the above-described problems.

(1) The present invention classifies electronic mail received by a user into normal mail and spam mail according to a predetermined filter rule (for example, equivalent to the classification unit 12 in FIG. 1), and the classified Feature vector generation means (for example, equivalent to the feature vector generation unit 13 in FIG. 1) for generating a feature vector by collating and extracting features of header information of each electronic mail, the normal mail and spam mail Filter rule generation means (for example, equivalent to the filter rule generation unit 14 in FIG. 1) that analyzes the feature vector and generates an initial filter rule, and receives received e-mail according to the generated filter rule as normal mail and spam E-mail filtering characterized by comprising separation means (e.g., equivalent to the separation unit 15 in FIG. 1) for separation into mail It has proposed a location.

  According to this invention, the classification means classifies the email received by the user into normal mail and spam mail according to a predetermined filter rule, and the feature vector generation means includes the header information of each classified email. The feature vector is generated by collating and extracting the features. The filter rule generation unit analyzes the feature vectors of normal mail and spam mail and generates an initial filter rule. Then, the separating means separates the received electronic mail into normal mail and spam mail according to the generated filter rule. Therefore, it is possible to automatically filter spam mails when receiving e-mails and reduce erroneous determination of spam mails.

  (2) The present invention relates to the e-mail filtering device of (1), storing means for storing the separated normal mail and spam mail (for example, in the normal mail storage unit 16 and the spam mail storage unit 17 of FIG. And the normal mail stored in the storage means and the spam mail, the characteristics of the header information are collated and extracted, the characteristic vectors of the normal mail and the spam mail are analyzed, and the generated initial The present invention proposes an e-mail filtering device characterized by comprising filter rule updating means (for example, corresponding to the filter rule updating unit 19 in FIG. 1) for updating the filter rule.

  According to this invention, the normal mail and spam mail separated by the storage means are stored, and the header information features are collated and extracted from the normal mail and spam mail stored by the feature vector regenerating means. Then, the filter rule update means analyzes the feature vectors of normal mail and spam mail and updates the generated initial filter rule. Therefore, since the updated filter rule is always used, the erroneous determination of spam mail can be further reduced.

  (3) According to the present invention, in the electronic mail filtering device of (1) or (2), the feature vector generation means (for example, corresponding to the feature vector generation unit 13 in FIG. 1) The present invention proposes an electronic mail filtering device characterized by generating a feature vector by collating and extracting at least a feature of domain information of a sender.

  According to this invention, the feature vector generation means generates a feature vector by collating and extracting at least the feature of the sender's domain information from the header information of the electronic mail. Therefore, since feature vectors are generated based on the presence or absence of fields, locations that do not match or match in the email, or the domain itself, feature vectors can be generated with simple processing, and spam can be received when an email is received. It can automatically filter emails and reduce spam email misjudgments.

  (4) According to the present invention, in the electronic mail filtering device of (1) or (2), the feature vector generation means (for example, corresponding to the feature vector generation unit 23 in FIG. 5) The present invention proposes an e-mail filtering device characterized by generating a feature vector by collating and extracting features of server information through which at least the e-mail passes.

  According to this invention, the feature vector generation means generates a feature vector by collating and extracting at least the feature of the server information through which the electronic mail has passed through the header information of the electronic mail. Therefore, since the feature vector is generated based on the context of the MTA, the degree of coincidence between the self-named host information and the actual host information, the IP address, etc., the feature vector can be generated by a simple process, and when receiving an email It can automatically filter spam emails and reduce spam email misjudgments.

  (5) According to the present invention, in the electronic mail filtering device of (1) or (2), the feature vector generation means (for example, corresponding to the feature vector generation unit 33 in FIG. 8) The electronic mail filtering apparatus is characterized by generating a feature vector by collating and extracting at least a feature of the sender's mail environment information.

  According to this invention, the feature vector generation means generates a feature vector by collating and extracting at least the feature of the sender's mail environment information from the header information of the electronic mail. Therefore, since feature vectors are generated based on date, character settings, etc., feature vectors can be generated with simple processing, spam emails are automatically filtered when emails are received, and spam emails are misjudged Can be reduced.

  (6) The present invention includes a first step (for example, corresponding to step S101 in FIG. 3) for classifying the mail into a normal mail and a spam mail according to the default settings of the known spam mail filter, A second step (for example, corresponding to step S102 in FIG. 3) for collating and extracting features of the header information of the email to generate a feature vector, and analyzing the feature vectors of the normal mail and spam mail A third step of generating an initial filter rule (for example, corresponding to step S103 of FIG. 3), and a fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule Step (equivalent to steps S104 and S105 in FIG. 4), and e-mail filtering It has proposed a method.

  According to the present invention, according to the default settings of the known spam mail filter, it is classified into normal mail and spam mail, and the feature information of the header information of each classified email is collated and extracted, and the feature vector is obtained. Generate. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, regular mail and spam mail are classified according to the default settings of the known spam mail filter, so it is possible to classify regular mail and spam mail in a simple way, and when receiving email It can automatically filter spam emails and reduce spam email misjudgments.

  (7) According to the present invention, an e-mail filtered under the condition that the determination criterion of the known spam mail filter is set higher than the default value is set as spam mail, and the determination criterion of the known spam mail filter is set lower than the default value. A first step (for example, corresponding to step S201 in FIG. 6) of classifying an email filtered under the above conditions as a normal email, and collating and extracting features of header information of each classified email And a second step of generating feature vectors (for example, corresponding to step S202 of FIG. 6), and a third step of analyzing the feature vectors of the normal mail and spam mail and generating an initial filter rule (For example, corresponding to step S203 in FIG. 6) and receiving according to the generated filter rule A fourth step of separating child mail into normal mail and spam mail (for example, corresponding to steps S204 to S207 in FIG. 7) is proposed. .

  According to the present invention, an email filtered under the condition that the determination criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the determination criterion of the known spam mail filter is set lower than the default value. The e-mail filtered by the condition is classified as a normal mail, and the feature information is generated by collating and extracting the features of the header information of each classified e-mail. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, e-mail filtered under the condition that the criterion of the known spam mail filter is set higher than the default value is regarded as spam mail, and is filtered under the condition that the criterion of the known spam mail filter is set lower than the default value. E-mails are classified as regular mails, so regular mails and spam mails can be reliably distinguished, and spam mails are automatically filtered when receiving e-mails, reducing false judgments of spam mails. it can.

  (8) The present invention collects spam mails published on the web, and collects e-mails in the user's mailbox as normal mails (for example, in step S301 in FIG. 9). A second step (for example, corresponding to step S302 in FIG. 9) for generating a feature vector by collating and extracting features of header information of each collected e-mail, Analyzing the feature vectors of mail and spam mail and generating an initial filter rule (for example, equivalent to step S303 in FIG. 9), and in accordance with the generated filter rule, receive e-mail as a normal mail And a fourth step (for example, equivalent to steps S304 and S305 in FIG. 10) for separating the mail into spam mails. It has proposed a method of filtering e-mail which is characterized.

  According to the present invention, the spam mails published on the web are collected, the electronic mails in the user's mailbox are collected as normal mails, and the header information of each collected electronic mail is characterized. A feature vector is generated by performing collation and extraction. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, it collects spam emails published on the web and collects emails in the user's mailbox as regular emails to classify regular emails and spam emails, ensuring that Mail and spam mail can be discriminated, and spam mail is automatically filtered when an e-mail is received, and erroneous determination of spam mail can be reduced.

  (9) The present invention is a program for causing a computer to execute an e-mail filtering method, and a first step (for example, classifying into normal e-mail and spam e-mail according to a default setting of a known spam e-mail filter) , Corresponding to step S101 in FIG. 3) and a second step (for example, in step S102 in FIG. 3) to generate a feature vector by collating and extracting features of header information of each classified e-mail. The third step (equivalent to step S103 in FIG. 3) for analyzing the feature vectors of the normal mail and spam mail and generating an initial filter rule, and receiving according to the generated filter rule The fourth step (for example, the step of FIG. 4) for separating the e-mail to be processed into normal mail and spam mail. We propose a program for executing the corresponding) to flop S104, S105, to the computer.

  According to the present invention, according to the default settings of the known spam mail filter, it is classified into normal mail and spam mail, and the feature information of the header information of each classified email is collated and extracted, and the feature vector is obtained. Generate. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, regular mail and spam mail are classified according to the default settings of the known spam mail filter, so it is possible to classify regular mail and spam mail in a simple way, and when receiving email It can automatically filter spam emails and reduce spam email misjudgments.

  (10) The present invention is a program for causing a computer to execute an e-mail filtering method. An e-mail filtered under a condition in which a determination criterion of a known spam mail filter is set higher than a default value is spam mail. A first step (for example, corresponding to step S201 in FIG. 6) of classifying an email filtered as a normal email under a condition in which a known spam mail filter criterion is set lower than the default value; A second step (for example, equivalent to step S202 in FIG. 6) of collating and extracting features of header information of each classified email to generate a feature vector, and the normal email and spam email A third step of analyzing the feature vector and generating an initial filter rule ( For example, it corresponds to step S203 in FIG. 6 and a fourth step (for example, corresponding to steps S204 to S207 in FIG. 7) for separating received e-mail into normal mail and spam mail according to the generated filter rule. ) And a program for causing a computer to execute.

  According to the present invention, an email filtered under the condition that the determination criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the determination criterion of the known spam mail filter is set lower than the default value. The e-mail filtered by the condition is classified as a normal mail, and the feature information is generated by collating and extracting the features of the header information of each classified e-mail. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, e-mail filtered under the condition that the criterion of the known spam mail filter is set higher than the default value is regarded as spam mail, and is filtered under the condition that the criterion of the known spam mail filter is set lower than the default value. E-mails are classified as regular mails, so regular mails and spam mails can be reliably distinguished, and spam mails are automatically filtered when receiving e-mails, reducing false judgments of spam mails. it can.

  (11) The present invention is a program for causing a computer to execute an e-mail filtering method, collecting spam e-mails published on the web, and e-mails in a user's mailbox. A first step (for example, corresponding to step S301 in FIG. 9) that is collected as mail, and a feature vector is generated by collating and extracting features of header information of each collected email. Step (for example, equivalent to step S302 in FIG. 9), and a third step for analyzing the normal mail and spam mail feature vectors and generating an initial filter rule (for example, equivalent to step S303 in FIG. 9) And separating received e-mail into normal mail and spam mail according to the generated filter rule. The fourth step (e.g., corresponding to step S304, S305 of FIG. 10) proposes a program for executing a, to the computer.

  According to the present invention, the spam mails published on the web are collected, the electronic mails in the user's mailbox are collected as normal mails, and the header information of each collected electronic mail is characterized. A feature vector is generated by performing collation and extraction. Then, the characteristic vectors of normal mail and spam mail are analyzed, an initial filter rule is generated, and the received electronic mail is separated into normal mail and spam mail according to the generated filter rule. Therefore, it collects spam emails published on the web and collects emails in the user's mailbox as regular emails to classify regular emails and spam emails, ensuring that Mail and spam mail can be discriminated, and spam mail is automatically filtered when an e-mail is received, and erroneous determination of spam mail can be reduced.

  According to the present invention, since it is possible to refer to the characteristics of past spam mails, it is possible to reliably block electronic mails from senders that have been determined to be spam mails or emails having information related to known spam mails. There is an effect that can be.

  According to the present invention, since it is possible to refer to past spam mails and known spam mails, it is possible to effectively deal with spam mail senders whose methods change over time, such as changing images and keywords. There is an effect that can be done.

  According to the present invention, since only the characteristics of the header part in the e-mail are referred to, it is possible to effectively deal with spam mail senders that change the technique over time, such as changing images and keywords. There is an effect that can be done.

  According to the present invention, since the characteristics of the header part in the electronic mail are referred to, there is an effect that filtering can be performed accurately even for a botnet or the like having similar characteristics to the spam mail and the transmission method.

  There is an effect that a unique filter corresponding to a small network such as an in-house system can be generated. In general, higher accuracy can be achieved by cooperating with a used filter.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
Note that the constituent elements in the present embodiment can be appropriately replaced with existing constituent elements, and various variations including combinations with other existing constituent elements are possible. Therefore, the description of the present embodiment does not limit the contents of the invention described in the claims.

<First Embodiment>
The electronic mail filtering apparatus according to this embodiment will be described with reference to FIGS. 1 to 4 and FIG.

<Configuration of e-mail filtering device>
As shown in FIG. 1, the electronic mail filtering device according to the present embodiment includes a mail receiving unit 11, a classification unit 12, a feature vector generation unit 13, a filter rule generation unit 14, a separation unit 15, and a normal mail. The storage unit 16 includes a spam mail storage unit 17 and a filter rule update unit 19.

  The classification unit 12 classifies the electronic mail received by the user at the mail reception unit 11 into a normal mail and a spam mail according to a predetermined filter rule, and creates a data set of the spam mail and the normal mail. In the present embodiment, an example in which the default setting of a known spam mail filter is used as the predetermined filter rule will be described. This makes it possible to create a normal mail and a known spam mail data set by a simple method.

  The predetermined filter rule is not limited to this, and an e-mail filtered under a condition in which a determination criterion of a known spam mail filter is set higher than a default value is regarded as spam mail, and a determination criterion of a known spam mail filter E-mail filtered under the condition that is set to a value lower than the default value may be used as normal mail, spam e-mail published on the web is collected, and e-mail in the user's mailbox is collected. A method of classification by collecting as normal mail may be used.

  The feature vector generation unit 13 collates and extracts features of the header information of each electronic mail classified by the classification unit 12 to generate a feature vector. In the present embodiment, an example will be described in which feature vectors are generated by collating and extracting features of sender's domain information from the header information of an email. In the example of FIG. 11, “Return-Path: <test@example.com>” or “from example2.com”, Message-Id: <200709121116.18CBGDr9020678@example. The feature vector is generated by collating and extracting features such as com>. Thereby, since the feature vector is generated based on the presence / absence of the field, the location that does not match the location in the mail, or the domain itself, the feature vector can be generated by a simple process.

  The header information is not limited to this, but may be server information through which an e-mail passes, mail environment information of a sender, or the like.

  The filter rule generation unit 14 analyzes feature vectors of known spam mail and normal mail, and generates an initial filter rule. Generate initial filter rules. The separating unit 15 separates the electronic mail received by the mail receiving unit 11 into normal mail and spam mail according to the generated filter rule. The normal mail storage unit 16 stores the normal mail separated by the separation unit 15. The spam mail storage unit 17 stores the spam mail separated by the separation unit 15.

  Here, as shown in FIG. 2, for example, there are the following methods for creating a database of feature vectors of normal mail and spam mail in the normal mail storage unit 16 and the spam mail storage unit 17, as shown in FIG. The mail itself may be stored in a database.

  As a method for creating a database, as shown in FIG. 2A, each feature vector of each e-mail is converted into a database as it is. Here, mail1, mail2, mail3, and mail4 are mails received by the mail receiving unit 11, w is a Return-Path domain (for example, XXX.com), and x is a From domain (for example, XXX). .Com), y indicates a From domain (for example, ΔΔΔ.com), and z indicates a Message-ID domain (for example, OO.co.jp).

  FIG. 2B is a method of creating a database by matrixing the feature vectors and the numbers in the feature vectors. Here, as the feature vector, for example, XXX. com, □□□. The domain of the com and From fields matches the domain of the Return-Path field? , W is the number of Return-Path domains, x is the number of From domains, and y is the number of yes.

  FIG. 2C shows a method for creating a database of all the mail headers. Here, w is a Return-Path domain, x is a From domain, y is a Message-ID domain, and z is an IP address. In this example, the Return-Path domain of mail1 is “◯”. XX.com ”, the From domain is“ XXX.co.jp ”, the Message-ID domain is“ − ”, the IP address is“ 100.100.XX ”, and the return-Path domain of mail2 is “XXX”, the From domain is “ΔΔΔ.com”, the Message-ID domain is “ΔΔΔ.com”, and the IP address is “100.100.XY”. -Path domain is “xxx.com”, From domain is “□□□ .com” The Message-ID domain is “-”, the IP address is “100.100.X.Z”, the Return-Path domain of mail4 is “ΔΔΔ.com”, the From domain is “-”, and the Message-ID domain. Is “-” and the IP address is “100.X.X.X”.

  The filter rule update unit 19 analyzes the feature vectors of normal mail and spam mail and updates the filter rule. This can further reduce spam mail misjudgment. When storing the mail itself in the normal mail storage unit 16 and the spam mail storage unit 17, the feature vector is regenerated based on the header information of these emails, and the filter rule is updated. Also good.

<Processing of e-mail filtering device>
Processing of the e-mail filtering apparatus according to the present embodiment will be described with reference to FIGS. 3 and 4. The processing of the e-mail filtering apparatus according to the present embodiment will be described separately in the procedure for creating the initial filter rule and the procedure for classifying spam mail and normal mail according to the generated filter rule.

<Initial filter rule creation procedure>
First, the classification unit 12 classifies electronic mails received by the mail reception unit 11 into normal mails and spam mails based on existing filter rules (step S101). The feature vector generation unit 13 collates the feature of the sender's domain information among the header information of each email classified by the classification unit 12 to generate a feature vector (step S102).

  The filter rule generation unit 14 analyzes the feature vectors of the known spam mail and normal mail generated by the feature vector generation unit 13 and generates an initial filter rule (step S103).

<Classification of spam mail and normal mail>
It is verified whether or not a certain e-mail is determined as a spam mail by the initial filter rule generated by the filter rule generation unit 14 (step S104).

  At this time, if it is determined that the email is a spam mail (“Yes” in step S104), the user confirms whether or not the electronic mail is determined as a normal mail (step S105). If it is determined that the electronic mail is not a normal mail (“No” in step S105), the electronic mail is stored in the spam mail storage unit 17 (step S106).

  On the other hand, when it is determined in step S104 that it is not spam mail (“No” in step S104) and in step S104, it is determined that it is spam mail (“Yes” in step S104), but in step S105. When the user determines that the e-mail is a normal mail (“Yes” in step S105), the e-mail is stored in the normal mail storage unit 16. Note that after filtering processing, data is added to the normal mail storage unit 16 and the spam mail storage unit 17, respectively, and the feature vector regeneration unit is based on the latest normal mail storage unit 16 and the database in the spam mail storage unit 17. 18 and the filter rule update unit 19 sequentially update the filter rules.

  Therefore, according to the present embodiment, in order to classify normal mail and spam mail according to the default settings of the known spam mail filter, a data set of known spam mail and normal mail is created by a simple method, An initial filter rule can be created, and spam mail is automatically filtered when an e-mail is received, and erroneous determination of spam mail can be reduced.

<Second Embodiment>
The e-mail filtering device according to the present embodiment will be described with reference to FIGS. 5 to 7 and FIG.

<Configuration of e-mail filtering device>
As shown in FIG. 5, the e-mail filtering apparatus according to the present embodiment includes a mail receiving unit 11, a classification unit 22, a feature vector generation unit 23, a filter rule generation unit 14, a separation unit 15, and a normal mail. The storage unit 16 includes a spam mail storage unit 17 and a filter rule update unit 19. In addition, about the component which attaches | subjects the same code | symbol as 1st Embodiment, since it has the same function, the detailed description is abbreviate | omitted.

  The classification unit 22 classifies the electronic mail received by the user at the mail reception unit 11 into normal mail and spam mail according to a predetermined filter rule. In the present embodiment, as the predetermined filter rule, an email filtered under a condition in which the determination criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the determination criterion of the known spam mail filter is the default. An example of using a filter rule in which an e-mail filtered under a condition set lower than the value is a normal mail will be described. Thereby, it is possible to reliably discriminate between normal mail and spam mail.

  Note that the predetermined filter rule is not limited to this, and may be one using a default setting of a known spam mail filter, collecting spam mail published on the web, and storing it in the user's mailbox. A method of classifying certain electronic mails by collecting them as normal mails may be used.

  The feature vector generation unit 13 collates the features of the header information of each electronic mail classified by the classification unit 12 to generate a feature vector. In the present embodiment, an example will be described in which the feature vector is generated by collating the features of the server information through which the e-mail passes out of the header information of the e-mail. In the example of FIG. 11, “Received: from example2.com by example2.com (8.14.1 / 8.12.11) with ESMTP id 18CBDGdj207111 for <test2@example3.com>; Wed, 12 Sep 2007: 16:40 +0900 "or" Received: by example4.com (Postfix) id 51A59EC8EB; Wed, 12 Sep 2007 11:27:56 +0900 (JST) "," Received: from localhost. ]) By example4.com (Postfix) with ESMTP id 35D6CEC8D9 for <t st2@example3.com>; Wed, 12 Sep 2007 11:27:56 +0900 (JST) "," Received: from example.com (example111.com [XXX.XXX.XXX.XXX.XXX]) by example8. 14.1 / 8 / 8.11) with SMTP id 18CBGDr9020678 for <test2@example3.com>; Wed, 12 Sep 2007 20:16:13 +0900 ”,“ Received: from example5.comb. 14.1 / 8.11.11) with SMTP id 18CBGDr9020678for <test2@example3.com>; Wed, 12 Sep 2 07 20:16:13 +0900 ", and the like by performing a collation of the features of, and generates a feature vector. Thereby, since the feature vector is generated based on the context of the MTA, the degree of matching between the self-named host information and the actual host information, the IP address, etc., the feature vector can be generated by simple processing.

  The header information is not limited to this, but may be sender domain information, sender mail environment information, or the like.

  In addition, when storing the mail itself in the normal mail storage unit 16 and the spam mail storage unit 17, the feature vector is regenerated based on the header information of these emails, and the filter rule is updated. Also good.

<Processing of e-mail filtering device>
The processing of the e-mail filtering device according to this embodiment will be described with reference to FIGS. The processing of the e-mail filtering apparatus according to the present embodiment will be described separately in the procedure for creating the initial filter rule and the procedure for classifying spam mail and normal mail according to the generated filter rule.

<Initial filter rule creation procedure>
First, the e-mail received by the classifying unit 12 at the mail receiving unit 11 under the condition that the determination criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the known spam mail filter The e-mail filtered under the condition that the determination criterion is set lower than the default value is classified as a normal mail (step S201). The feature vector generation unit 13 collates the feature of the sender's domain information among the header information of each electronic mail classified by the classification unit 12, and generates a feature vector (step S202).

  The filter rule generation unit 14 analyzes the normal mail and spam mail feature vectors generated by the feature vector generation unit 13 and generates an initial filter rule (step S203).

<Classification of spam mail and normal mail>
First, it is determined whether the value of a certain e-mail is an e-mail having a higher value or a lower value than the default value of the existing filter (step S204). If the value of this e-mail is higher or lower than the default value of the existing filter (“Yes” in step S204), then whether or not this e-mail value is higher than the default value. Is determined (step S205).

At this time, if the value of this e-mail is higher than the default value (“Yes” in step S205), the e-mail is determined to be spam mail and data is stored in the spam mail database (step S208).

On the other hand, if the value of this e-mail is lower than the default value (“No” in step S205), the e-mail is determined to be a normal mail, and data is stored in a normal mail database (step S209). ).

Furthermore, in step S204, when the value of the e-mail is not higher or lower than the default value of the existing filter (“No” in step S204), the initial filter rule generated by the filter rule generation unit 14 Whether or not it is determined as spam mail is verified (step S204).

  At this time, when it is determined that the email is spam mail (“Yes” in step S206), the user confirms whether or not the electronic mail is determined to be a normal mail (step S207). If it is determined that this e-mail is not a normal mail (“No” in step S207), the e-mail is stored in the spam mail storage unit 17 (step S208).

  On the other hand, when it is determined in step S206 that it is not a spam mail (“No” in step S206) and in step S206, it is determined that it is a spam mail (“Yes” in step S206), but in step S207 When the user determines that the e-mail is a normal mail (“Yes” in step S207), the e-mail is stored in the normal mail storage unit 16. After the filtering process, data is added to the normal mail storage unit 16 and the spam mail storage unit 17, respectively, and the filter rule update unit is based on the latest normal mail storage unit 16 and the database in the spam mail storage unit 17. 19, the filter rules are sequentially updated.

  Therefore, according to the present embodiment, an email filtered under the condition that the determination criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the determination criterion of the known spam mail filter is set to be higher than the default value. E-mails filtered with low criteria are classified as regular mails, so you can create a dataset of known spam mails and regular mails in a simple way, and automatically receive spam mails when you receive them Filtering as well as reducing spam mail misjudgments.

<Third Embodiment>
The e-mail filtering device according to this embodiment will be described with reference to FIGS.

<Configuration of e-mail filtering device>
As shown in FIG. 6, the electronic mail filtering device according to the present embodiment includes a mail receiving unit 11, a classification unit 32, a feature vector generation unit 33, a filter rule generation unit 14, a separation unit 15, and a normal mail. The storage unit 16 includes a spam mail storage unit 17 and a filter rule update unit 19. In addition, about the component which attaches | subjects the same code | symbol as 1st Embodiment, since it has the same function, the detailed description is abbreviate | omitted.

  The classification unit 32 classifies the electronic mail received by the user at the mail reception unit 11 into normal mail and spam mail according to a predetermined filter rule. In the present embodiment, as the predetermined filter rule, spam mail published on the web is collected, and normal mail and spam mail are collected by collecting e-mail in the user's mailbox as normal mail. An example using a filter rule for classifying Thereby, it is possible to reliably discriminate between normal mail and spam mail.

  The predetermined filter rule is not limited to this, and may be one using a default setting of a known spam mail filter, or filtered under a condition in which the judgment criterion of the known spam mail filter is set higher than the default value. A method may be used in which the received e-mail is classified as spam mail, and the e-mail filtered under the condition that the determination criterion of the known spam mail filter is set lower than the default value is classified as a normal mail.

  The feature vector generation unit 13 collates the features of the header information of each electronic mail classified by the classification unit 12 to generate a feature vector. In the present embodiment, an example will be described in which a feature vector is generated by comparing features of a sender's mail environment in header information of an email. In the example of FIG. 11, feature vectors such as “Date: Wed, 12 Sep 2007 11:27:22 +0900” and “Content-Type: text / plain; charset = US-ASCII” are collated to generate a feature vector. To do. Thereby, since the feature vector is generated based on the date, Content-Type, or the like, the feature vector can be generated by a simple process.

  The header information is not limited to this, and may be the sender's domain information, server information via e-mail, or the like.

  Further, when the mail itself is stored in the normal mail storage unit 16 and the spam mail storage unit 17, the feature vector may be regenerated based on the header information of these electronic mails.

<Processing of e-mail filtering device>
The processing of the e-mail filtering apparatus according to this embodiment will be described with reference to FIGS. 9 and 10. The processing of the e-mail filtering apparatus according to the present embodiment will be described separately in the procedure for creating the initial filter rule and the procedure for classifying spam mail and normal mail according to the generated filter rule.

<Initial filter rule creation procedure>
First, the classification unit 32 collects spam mails published on the web, and classifies normal mails and spam mails by collecting electronic mails in the user's mailbox as normal mails (steps). S301). The feature vector generation unit 33 compares the features of the sender's mail environment among the header information of each electronic mail classified by the classification unit 32, and generates a feature vector (step S302).

  The filter rule generation unit 14 analyzes the feature vectors of normal mail and spam mail generated by the feature vector generation unit 33, and generates an initial filter rule (step S303).

<Classification of spam mail and normal mail>
First, it is verified whether or not an e-mail is determined to be spam mail by the initial filter rule generated by the filter rule generation unit 14 (step S304).

  At this time, if it is determined that the email is spam mail (“Yes” in step S304), the user confirms whether or not the electronic mail is determined to be a normal mail (step S305). If it is determined that this electronic mail is not a normal mail (“No” in step S305), the electronic mail is stored in the spam mail storage unit 17 (step S306).

  On the other hand, if it is determined in step S304 that it is not a spam mail (“No” in step S304) and it is determined in step S304 that it is a spam mail (“Yes” in step S304), but in step S305 When the user determines that the e-mail is a normal mail (“Yes” in step S305), the e-mail is stored in the normal mail storage unit 16. After the filtering process, data is added to the normal mail storage unit 16 and the spam mail storage unit 17, respectively. Based on the latest normal mail storage unit 16 and the database in the spam mail storage unit 17, the filter rule update unit 19 is added. Thus, the filter rules are sequentially updated.

  Therefore, according to the present embodiment, the spam mails published on the web are collected, the emails in the user's mailbox are collected as normal mails, and the normal mails and the spam mails are classified. Therefore, it is possible to reliably discriminate between normal mail and spam mail, and automatically filter spam mail at the time of receiving an electronic mail, and reduce erroneous determination of spam mail.

  The processing of the electronic mail filtering apparatus is recorded on a computer-readable recording medium, and the electronic mail filtering apparatus of the present invention is realized by causing the electronic mail filtering apparatus to read and execute the program recorded on the recording medium. be able to. The computer system here includes an OS and hardware such as peripheral devices.

  Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW (World Wide Web) system is used. The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.

  The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

  The embodiments of the present invention have been described in detail with reference to the drawings. However, the specific configuration is not limited to the embodiments, and includes designs and the like that do not depart from the gist of the present invention. For example, the filter rule creation method is not particularly mentioned, but a classification model may be created by data mining by learning k-NN, decision tree, neural network, etc., or a rule base such as class association. A classification model using data mining may be created.

1 is a configuration diagram of an e-mail filtering device according to a first embodiment. FIG. It is the figure which illustrated the structure of the database in a normal mail preservation | save part and a spam mail preservation | save part. It is the figure which showed the creation procedure of the initial stage filter rule in the email filtering apparatus which concerns on 1st Embodiment. It is the figure which showed the classification | category procedure of the spam mail and normal mail in the electronic mail filtering apparatus which concerns on 1st Embodiment. It is a block diagram of the electronic mail filtering apparatus which concerns on 2nd Embodiment. It is the figure which showed the preparation procedure of the initial stage filter rule in the email filtering apparatus which concerns on 2nd Embodiment. It is the figure which showed the classification | category procedure of the spam mail and normal mail in the electronic mail filtering apparatus which concerns on 2nd Embodiment. It is a block diagram of the electronic mail filtering apparatus which concerns on 3rd Embodiment. It is the figure which showed the preparation procedure of the initial stage filter rule in the email filtering apparatus which concerns on 3rd Embodiment. It is the figure which showed the classification | category procedure of the spam mail and normal mail in the electronic mail filtering apparatus which concerns on 3rd Embodiment. It is a figure which shows the specific example of header information.

Explanation of symbols

  11 ... mail receiving unit, 12, 22, 32 ... classifying unit, 13, 23, 33 ... feature vector generating unit, 14 ... filter rule generating unit, 15 ... separating unit, 16. ..Normal mail storage unit, 17 ... Spam mail storage unit, 19 ... Filter rule update unit

Claims (11)

A classification means for classifying email received by the user into regular mail and spam mail according to a predetermined filter rule;
A feature vector generating means for generating a feature vector by collating and extracting features of header information of each classified e-mail;
A filter rule generating means for analyzing the feature vectors of the normal mail and the spam mail and generating an initial filter rule;
Separating means for separating received e-mail into normal mail and spam mail according to the generated filter rule;
An e-mail filtering device comprising: Storage means for storing the separated normal mail and spam mail;
From the normal mail and spam mail stored in the storage means, the header information features are collated and extracted, the normal mail and spam mail feature vectors are analyzed, and the generated initial filter rule is A filter rule update means for updating;
The electronic mail filtering apparatus according to claim 1, further comprising:   3. The feature vector generation unit generates a feature vector by collating and extracting at least a feature of domain information of a sender from header information of the e-mail. Email filtering device.   3. The feature vector generation means generates a feature vector by collating and extracting at least features of server information through which the e-mail has passed from header information of the e-mail. The e-mail filtering device described in 1.   3. The feature vector generation unit generates a feature vector by collating and extracting at least the feature of the sender's mail environment information from the header information of the electronic mail. E-mail filtering device. A first step of classifying normal mail and spam mail according to the default settings of known spam mail filters;
A second step of collating and extracting features of header information of each classified e-mail to generate a feature vector;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
An e-mail filtering method comprising: E-mail filtered with the condition that the criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the e-mail that is filtered with the condition of the criterion of the known spam mail filter set lower than the default value A first step of classifying mail as normal mail;
A second step of collating and extracting features of header information of each classified e-mail to generate a feature vector;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
An e-mail filtering method comprising: A first step of collecting spam mail published on the web and collecting e-mail in the user's mailbox as normal mail;
A second step of collating and extracting features of header information of each collected email to generate a feature vector;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
An e-mail filtering method comprising: A program for causing a computer to execute an e-mail filtering method,
A first step of classifying normal mail and spam mail according to the default settings of known spam mail filters;
A second step of collating and extracting features of header information of each classified e-mail to generate a feature vector;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
A program that causes a computer to execute. A program for causing a computer to execute an e-mail filtering method,
E-mail filtered with the condition that the criterion of the known spam mail filter is set higher than the default value is defined as spam mail, and the e-mail that is filtered with the condition of the criterion of the known spam mail filter set lower than the default value A first step of classifying mail as normal mail;
A second step of collating and extracting features of header information of each classified e-mail to generate a feature vector;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
A program that causes a computer to execute. A program for causing a computer to execute an e-mail filtering method,
A first step of collecting spam mail published on the web and collecting e-mail in the user's mailbox as normal mail;
A second step of generating feature vectors by matching features of the header information of each collected email;
A third step of analyzing feature vectors of the normal mail and spam mail and generating an initial filter rule;
A fourth step of separating received e-mail into normal mail and spam mail according to the generated filter rule;
A program that causes a computer to execute.

Images