JP2009037346A - Unwanted e-mail exclusion system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that since only a mail manager manages a database in which a keyword as a criterion of the unwanted e-mail is stored in the conventional unwanted e-mail exclusion system, it is necessary for each receiver to request a mail manager to register information in the database each time receiving the unwanted e-mail. <P>SOLUTION: A data repeater system 3 for enabling an enterprise share information of electronic mail to be relayed between an Internet server 1 and terminal equipment 2 inside the enterprise is installed. The plurality of words/phrases of the reception mail are compared with SPAM probability calculated from the degree of appearance of each word/phrase of a SPAM information management database 4 so that the SPAM/non-SPAM of the reception mail can be determined by a reception mail decision means 32, and the reception mail is classified into each log. When the reception mail is not correctly classified due to erroneous determination, each receiver is made to correctly classify and learn the reception mail by using a spam registration means 34a and a spam registration releasing means 34b. Thus, it is possible to improve the spam decision precision by the user inside the enterprise. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention is to install a data relay device between an Internet server and a company internal terminal to share a log of relayed e-mail, and receive the spam probability of the received mail from the lexical extracted from the e-mail shared with the received mail The present invention relates to a junk e-mail elimination system that determines the degree of spam / non-spam by calculating the degree of error, and in the case of erroneous determination for each recipient, corrects the classification by each recipient and learns it.

  Conventional spam e-mail elimination methods include a method in which only a mail administrator manages a dictionary in which keywords and addresses for determining junk e-mail are registered, and a method for eliminating spam e-mails using words contained in e-mails There is.

  In Patent Document 1, only a mail administrator has the authority to register, refer to, and delete a keyword that is a determination criterion for mail that rejects incoming calls in a database, and includes a keyword registered in a received mail. In this case, a technique is disclosed in which the received mail is discarded and not delivered. The method for eliminating junk e-mail disclosed in Patent Document 1 includes a database that stores the address of a subscriber of a junk e-mail incoming rejection contract and a keyword that is a criterion for determining junk e-mail, and a contract that stores the address of a junk mail incoming rejection contract. When the person information registration function means and the transmission destination address of the received mail from the outside are registered in the database, it is determined whether or not the keyword stored in the database is included. Cancels the received mail, and if it is not included, the mail incoming rejection / permission function means for storing in the system device or the mail spooler, the keyword reference function means for referring to the keyword registered in the database, Keyword registration function means for registering a keyword in a database, and registration of the database Those having a keyword delete function means that you want to delete the keyword.

Further, Patent Document 2 detects word tendency information when there is a word in a database in which the spam / non-spam tendency of the word is registered among the words included in the e-mail to be discriminated. Discloses a technique for determining whether or not the e-mail is a spam mail by adding word tendency information as a word having a non-spam tendency when the e-mail is not registered. The method of eliminating junk e-mails in Patent Literature 2 includes a word tendency information holding unit that holds whether a word has a tendency of spam / non-spam, and the magnitude of this tendency as word tendency information for each word; An email acquisition unit that acquires email, a word extraction unit that extracts words included in the email, and a word tendency information for words registered in the word tendency information holding unit, while detecting unregistered words Has a word tendency information acquisition unit that gives word tendency information as a word having a non-spam tendency, and a determination unit that determines whether the word is spam mail from the word tendency information of a word extracted from an e-mail Is.
JP 2004-133508 A JP 2007-122148 A

  However, as described above, in the conventional spam mail rejection system, the authority to register or delete the keyword that is the criterion for judging spam mail is limited to only the mail administrator. Each time a message is received, it is necessary to convey the information to the mail manager and register it in the database. For this reason, if the registration / deletion process by the email administrator is omitted, or if the contact of each recipient to the email administrator is leaked, the same spam mail is not excluded and continues to be delivered to each recipient. In addition, when a delay occurs in the work by the mail manager, there is a problem that it takes time until each receiver receives a request and reflection of spam mail is reflected. Since the received mail determined to be junk mail is discarded without being stored in the mail spooler, each recipient can know the existence of the received mail and confirm the contents of the received mail later. There were no problems.

  In addition, if the word contained in the received mail is not registered in the database in which the word spam / non-spam tendency is registered, whether the word is spam mail by adding word tendency information of non-spam tendency to the word. Therefore, there is a problem that detection errors increase while detection errors are reduced.

  The present invention has been made in view of the above-mentioned problems, and a data relay device is installed between an Internet server and a terminal in a company, and information on all electronic mail relayed by the data relay device. In the junk mail rejection system that eliminates the reception of junk e-mail among the relayed e-mails, the e-mail or the data relay device already shared Corresponding to the log identification information by the lexical extraction means for extracting each lexical word appearing from the header and body of the relayed received mail, the log identification information indicating the log of the shared electronic mail, and the lexical extraction means Each lexical word extracted from the e-mail, and each extracted lexical word contains a spam log for storing junk mail and a reception log for storing non-spam mail. A spam information management database for managing the appearance degree appearing in a plurality of words, a plurality of words extracted by the word extraction means from the received mail addressed to each terminal in the company, and the appearance of the spam information management database The received mail determining means for determining spam / non-spam of the received mail by calculating the spam probability of the received mail by comparing the spam probability of each lexical word calculated from the degree, and based on the determination result The received mail classification means for classifying the received mail into the spam log or the received log, and the received mail classified into the spam log or the received log according to the determination result against each recipient's will, Spam / non-spam instructions are directly given by the recipient's judgment from the terminal inside the company, It solves by being constituted by a learning means for increasing the spam and non-spam determination accuracy of Shin-email determining unit.

  In the present invention, the learning means is classified into the reception log based on the determination result of the received mail determination means as non-spam mail, but the received mail is determined to be spam mail by each recipient. When a spam registration button provided on the list of received logs or the body text screen of the received mail is clicked once by an instruction of each recipient, the log identification information and each extracted A spam registration unit is provided for registering a lexical phrase and the appearance degree and moving the received mail from the reception log to the spam log.

  Furthermore, in the present invention, the learning means is classified into the spam log according to the determination result of the received mail determining means as spam mail, but the received mail is determined as non-spam mail by each recipient. When the spam exclusion button provided on the spam log list or the received mail text screen is clicked once by the instruction of each recipient, the log identification information registered in the spam information management database, the extraction And a spam deregistration means for moving the received mail from the spam log to the reception log.

  As described above, in the spam mail rejection system of the present invention, a data relay device that shares an electronic mail log inside the company is installed between the Internet server and the terminal in the company, and appears from the relayed received mail. The spam probability calculated from the appearance degree of each lexical in the spam information management database that extracts each lexical phrase and manages the spam probability information of each e-mail shared with a plurality of lexical words of the received received mail. In comparison, spam / non-spam is determined from spam probabilities of a plurality of lexical characters of received mail and classified into spam logs or received logs. With respect to the received mail classified against each recipient's will, spam / non-spam instructions are directly given by each recipient, and the accuracy of determination of spam / non-spam is improved. As a result, there is no need to set up a special email administrator who has the authority to manage the database that is used to determine spam, and each user within the company performs spam registration and release of spam. Since it is possible to manage a database that is a criterion for e-mail, the contents of each user's processing are reflected in all users in the company. And since each user can register and cancel spam directly, the time until it is reflected in the determination of spam / non-spam can be greatly shortened, and the degree of appearance of each lexical in the database, which is the judgment criterion for spam, is also the user. Everyone can learn.

  Furthermore, since all the emails relayed by the data relay device are shared by a plurality of users inside the company, even users other than each recipient can read the emails of the recipients in the reception log. When a certain spam mail or non-spam mail is found in the spam log, the spam registration and spam cancellation buttons are pressed by each recipient or a user other than the recipient, and the spam registration means and spam registration cancellation means Processing is started. As a result, all users in the company can manage spam and non-spam, so that a spam mail eliminating system that does not require a mail manager can be realized. Note that “sharing” here is not limited to a state in which all emails relayed by all users are shared, but also includes a state in which a sharing range or a monitoring range is set for each user or each department. In the latter case, other users' e-mails can be registered as spam or released as long as each user can browse.

  Also, according to the present invention, each recipient is a junk mail, but the received mail is determined to be a non-spam mail by the received mail judging means and classified into the reception log. By clicking a spam registration button provided on the list or the received mail text screen, each information is registered in the spam information management database and the received mail is moved to the spam log. This corrects erroneous discrimination due to omission of detection by the received mail determination means. It is also possible to select a plurality of spam mails from the received log list and register spam at a time.

  Furthermore, according to the present invention, each recipient is a non-spam mail, but the received mail is judged to be a spam mail by the received mail judging means and classified into the spam log. By clicking one button on the spam exclusion button provided on the list or the received mail text screen, each information in the spam information management database is excluded and the received mail is moved to the reception log. Thereby, the erroneous determination due to the erroneous detection of the received mail determination means can be corrected. You can also select multiple non-spam emails from the spam log list to exclude spam at once.

  Furthermore, by learning the accuracy of spam / non-spam determination for all users of the information of all emails relayed through the data relay device, the number of emails that are the basis for the occurrence of each word in the spam information management database increases. . This improves the accuracy of the spam probability calculated from the appearance level of each lexical word, reduces the number of unregistered words in the spam information management database, and causes errors due to both false detection and missed detection when determining spam / non-spam. Judgment can be greatly reduced.

  Furthermore, by storing the received mail determined as spam by the received mail determining means and the received mail registered as spam by the spam registering means in the spam log, it is possible to leave unsolicited mail without discarding it from the data relay device. In addition, each user can check the received mail stored in the spam log due to the erroneous determination later.

  Furthermore, by recording the information of all electronic mail relayed to the data relay apparatus as a log, even the electronic mail deleted from the recipient terminal can be referred to by the data relay apparatus. It is possible to process spam registration and spam registration for mail.

  Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. 1 to 4.

  First, FIG. 1A shows a configuration diagram of the spam mail rejection system of the present invention, and FIG. 1B shows a diagram for explaining a spam information management database of the present invention.

  As shown in FIG. 1A, a spam mail rejection system according to the present invention includes an Internet server 1 installed inside a provider outside the company or inside the company, and a terminal inside the company connected to the Internet server 1 via the Internet. 2 and a data relay device 3 installed between the Internet server 1 and the terminal 2.

  The Internet server 1 is a computer that is installed inside a provider outside the company or inside the company, and provides the client computer with its own functions and data. Specifically, it is a mail server, and the mail server 1 is a computer that is always connected to the Internet and transmits and receives a user's e-mail in its own network.

  The terminal 2 is a computer installed in a company that is connected to the Internet and can send and receive electronic mail. Terminals 2 in the company are connected by a LAN, and data is relayed to / from an external network via a router (not shown).

  The data relay device 3 is installed between a router that relays internal data to the mail server 1 of a provider outside the enterprise and the terminal 2, or between the mail server 1 and the terminal 2 inside the enterprise, A device that relays electronic mail transmitted and received between the server 1 and the terminal 2 and records the relayed electronic mail information in a log as history information. The mail server (Internet server) 1 outside the company or inside the company The data relay device 3 can be input alone with almost no influence on the setting. The data relay device 3 includes a spam information management database 4, a lexical extraction means 31, a received mail determination means 32, a received mail classification means 33, and a learning means 34.

  The spam information management database 4 is a database in which log information of electronic mail relayed through the data relay device 3 and spam information of each electronic mail are recorded. Specifically, the log information is log identification information indicating mail data of each electronic mail (not shown), and spam information is extracted from the electronic mail corresponding to the log identification information by the lexical extraction means 31 described later. Each lexical word and the lexical word from which the spam probability is calculated are information on the degree of appearance that appears in the spam log storing spam mail and the reception log storing non-spam mail (see FIG. 1B). These are managed by the spam information management database 4.

  The log identification information is a set of all e-mail information relayed through the data relay device 3, and includes at least transmission / reception date / time, sender name, recipient name, e-mail subject, e-mail text, sender e-mail address, recipient's e-mail address. Information such as an e-mail address and an e-mail data size. In addition, log identification information is not divided into areas for each user in the company and history information is managed, but all history information is managed in a unified manner and shared by all users. Even if the e-mail has been deleted at the terminal 2 of each user, it remains in the spam information management database 4, so that the processing of each means described later can be performed on these e-mails. Note that “sharing” here is not limited to a state in which all emails relayed by all users are shared, but also includes a state in which a sharing range or a monitoring range is set for each user or each department. Even in the latter case, each means described below can be performed on another person's e-mail within a range that can be browsed by each user.

  The appearance degree of the reception log is a value calculated from (a1) “total number of non-spam mails” and (b1) “number of occurrences of the word in all non-spam mails”. This is a value calculated from (a2) “total number of spam mails” and (b2) “number of occurrences of the word in all spam mails”. The word spam probability is calculated from these values (a1), (a2), (b1), and (b2).

Specifically, the values of (a1) and (a2) are calculated by aggregating the number of emails in the received log and spam log from the log identification information, and the values of (b1) and (b2) are [1 The number of appearances is adjusted at the timing of receiving an e-mail and classifying spam, and [2] registering or canceling e-mail as spam. Here, [1] receiving e-mail and classifying spam is extracting lexical characters from the e-mail and counting up the number of lexical occurrences in the spam information management database 4 according to the number of each lexical word. Specifically, when the email is identified as spam, the value of (b2) of each lexical word is counted up. [2] To register or cancel an e-mail as spam, the e-mail is read using log identification information, extracted from the e-mail, and lexically extracted from the e-mail. When registered, the value of (b2) is counted up for each lexical phrase, and (b1) is counted down.

  A log is a record that changes over time so that it can be seen later. Specifically, the log storage destination stores emails with received logs, spam logs, and viruses. Virus logs to be sent, send logs to save outgoing mail, and trash to evacuate each email that has been instructed to be deleted.

  Spam probability refers to the degree of occurrence of whether each received lexical phrase often appears in the received log or the spam log. It is a probability value indicating whether it is non-spam. The lexical spam probability and the e-mail spam probability are calculated in the range of 0 to 1. When the calculation result of the spam probability of an electronic mail is equal to or higher than a threshold value of 0.9, the mail is regarded as spam.

  The word / phrase will be described later.

  Here, the log information and spam information of e-mail are recorded in the spam information management database 4 and managed by one database. However, only the spam information is managed in the spam information management database 4, and the log information is You may make it acquire from the email which is sharing the email corresponding to a determination result by the received mail determination means 32 mentioned later.

  The lexical extraction means 31 is a means for extracting each lexical word that appears from the header and the body of the email that has already been shared or the received mail relayed through the data relay device 3. A lexical phrase is a part extracted from a character string in a certain unit. Specifically, when a lexical word is extracted from the phrase “data relay device.”, “Data”, “relay”, “device”, “.” Are obtained in word units, and “de”, “-” in character units. “Ta”, “Medium”, “Join”, “Appearance”, “Place”, “.”, “Data Relay Device” in the case of one character unit, and “Data Relay” “Relay Device” in the case of two adjacent words In the case of a set of two adjacent characters, “data”, “data”, “medium”, “relay”, “joining”, “device”, “device”.

The received mail judging means 32 is a spam probability calculated from a plurality of lexical words extracted by the lexical word extracting means 31 from the received mail addressed to the terminal 2 inside the company and the appearance degree of each lexical word in the spam information management database 4. This is a means for determining spam / non-spam of received mail by calculating the spam probability of the received mail from the spam probability of the plurality of words of the received mail. Specifically, as shown in FIG. 1 (B), the spam probability of each lexical word from the degree of appearance of each lexical word in the spam management information database 5 is
-"Complete" spam probability = (36/100) / {(36/100) + 2 x (3/150)} = 0.9
・ "Registration" spam probability = (15/100) / {(15/100) + 2 × (20/150)} = 0.36
・ Free spam probability = (32/100) / {(32/100) + 2 × (8/150)} = 0.75
And can be calculated. Therefore, the spam probability of the string "complete registration free"
・ Spam probability of the string “complete registration”
= (0.9 x 0.36 x 0.75) / {(0.9 x 0.36 x 0.75) + (1-0.9) x (1-0.36) x (1-0.75)}
= 0.938
Is calculated. Since the calculation result 0.938 is equal to or greater than the threshold value 0.9, the received mail determination means 32 determines that “complete registration free” is spam.

  The received mail classification means 33 is a means for classifying received mail into a spam log or a received log based on the determination result of the received mail determination means 32. By storing the spam mail determined to be spam in the spam log without deleting it, it can be used to determine the tendency of spam.

  The learning means 34 refers to the received mail classified into the spam log or the received log by the received mail classification means 33 based on the determination result of the received mail determination means 32 against the intention of each recipient from the terminal 2 inside the company. Spam / non-spam instructions are directly given by each recipient's judgment, and this is means for improving the accuracy of determination of spam / non-spam by the received mail judgment means. Specifically, it has spam registration means 34a and spam registration cancellation means 34b.

  The spam registration unit 34a is classified as a reception log based on the determination result of the received mail determination unit 32 as non-spam mail. However, the received mail determined to be spam by each recipient is received according to the instruction of each recipient. When the spam registration button provided in the log list or the received mail text screen is clicked once, the log identification information, each extracted lexical word and the degree of appearance are registered in the spam information management database 4, and the received mail It is a means to move from the reception log to the spam log. On the other hand, the spam registration canceling unit 34b is classified into a spam log according to the determination result of the received mail determining unit 32 as spam mail, but for each received mail determined as non-spam mail by each receiver, When a spam exclusion button provided on the spam log list or the body text screen of the received mail is clicked by one instruction, the log identification information registered in the spam information management database 4, each extracted lexical word, and the degree of appearance are excluded. A means for moving received mail from the spam log to the received log. The details of the spam registration unit 34a and the spam registration cancellation unit 34b will be described with reference to FIGS.

  By combining the functions of the above-described means and resources such as a database into one data relay device, it can be easily installed without changing the current settings before installation. And, by sharing resources such as the database of the data relay device among all users inside the company connected to the data relay device and below, it has knowledge of the special system and has the authority to register / release in the database Even without a mail administrator, corporate users can reflect spam / non-spam information in the database, improve the accuracy of spam probability, and greatly reduce false positives.

  Next, the operation by the spam registration unit 34a of the present invention will be described with reference to FIG. 2A is an example of a received mail list display screen, and FIG. 2B is an example of a screen showing the text of the received mail.

  As shown in FIG. 2 (A), the received mail list display screen 10a includes a log selection area 11, a received mail list area 12, and an operation instruction area 13.

  In the log selection area 11, a list of logs that are recording destinations of e-mails is displayed. Specifically, the above reception log, transmission log, spam log, virus log, and trash can are displayed in this order.

  The received mail list area 12 includes a mail selection box 12a for selecting a desired received mail from the list, a group selection column 12b for selecting a group such as a department within the company, and a received mail shared between the groups. There are a mail information display column 12c for displaying information such as the date and time of receipt, sender, and the like, and a mail search column 12d for searching with words included in the received mail information. Specifically, when displaying a list of received mails shared by the general affairs department, if “general affairs” is selected from the group selection field 12b, only the corresponding received mails are extracted and displayed in the mail information display field 12c. Then, when the title of the mail information display column 12c is clicked (in this case, it is an e-mail sent from Mr. Ta on 2006/8/30 00:03), the received mail shown in FIG. A body screen 10b is displayed. Although not shown in the figure, if “report” is entered in the mail search field 12d and the search button is pressed, the received mail including the “report” word is extracted and received in the received mail list area 12. Is displayed. Furthermore, if detailed conditions are set with a detailed search button, received mails are narrowed down and extracted.

  In the operation instruction area 13, a button for instructing an operation for a received mail whose mail selection box 12a is checked is arranged. Specifically, an output button 13a for downloading a mail source as a file, a button 13b for moving to a trash box, a button 13c for redistributing to a recipient, and a button 13d for registering spam are arranged.

  As shown in the figure, if the recipient (in this case, Mr. Yama xx) reads the received mail addressed to himself / herself displayed in the received mail list and finds junk mail, the mail of the corresponding received mail A check is given to the selection box 12a (in this case, two cases), and the button 13d for spam registration is pressed. The process of the spam registration means 34a is started by the instruction of the recipient, and the two received mails with the check added are moved to the spam log, and the process of generating the spam probability information to be registered in the spam information management database 4 is performed. It is activated. Specifically, each lexical word extracted from the received mail by the lexical extracting means 31, the degree of appearance of each lexical word calculated by the received mail determining means 32, and the log identification information of the movement destination by the spam registration means 34a are spam information. Registered in the management database 4. Note that even users other than mail recipients can similarly register spam on behalf of mail recipients as long as they are shared received emails.

  Further, as described above, instead of selecting a plurality of received mails in the received mail list area 12 and performing a process such as spam registration at a time, as shown in FIG. You may process spam registration etc. one by one. In this case, an operation instruction may be performed using each button arranged in the operation instruction area 13 'at the bottom of the screen. The source display button 13'e is a button for displaying the source of the received mail, and the other buttons are the same as those in the operation instruction area 13 and are omitted.

  Further, the screen layouts shown in FIGS. 2A and 2B are not limited to this form, and the layout may be different as long as each column is provided.

  Next, the operation by the spam registration cancellation means 34b of the present invention will be described with reference to FIG. FIG. 3A is an example of a list display screen of spam mail, and FIG. 3B is an example of a screen showing the text of spam mail.

  As shown in FIG. 3A, the junk mail list display screen 20 a includes a log selection area 21, a junk mail list area 22, and an operation instruction area 23.

  Similarly to the received mail list display screen 10a described above, the log selection area 21 displays the reception log, the transmission log, the spam log, the virus log, and the trash that are the recording destinations of the e-mail.

  The spam mail list area 22 includes a mail selection box 22a for selecting a desired spam mail from the list, a group selection field 22b for selecting a group such as a department within the company, and a spam mail shared between the groups. There is a mail information display field 22c for displaying the information of the mail, and a mail search field 22d for searching with words and phrases included in the spam mail information. Specifically, in order to display a list of all spam mails to be shared, “display all” may be selected from the group selection field 22b. Then, when the title of the mail information display field 22c is clicked (in this case, it is an email sent from Mr. X village XX on 2006/8/29 13:34), the inconvenience shown in FIG. A mail text screen 20b is displayed. It should be noted that, although not shown, if the e-mail search field 22d to "Hello" and press the search button and enter the lexical, appear in the Junk E-mail list area 22 is junk e-mail extraction, including the lexical of "Hello" The Furthermore, if detailed conditions are set with a detailed search button, spam mails are narrowed down and extracted.

  In the operation instruction area 23, a button for instructing an operation for a junk mail with a check in the mail selection box 22a is arranged. Specifically, an output button 23a for downloading a mail source as a file and a button 23f for excluding it from spam are arranged.

  As shown in the figure, if a recipient (in this case, Mr. Yama XX) finds a non-spam mail addressed to himself / herself in the spam mail list, he / she adds a check to the mail selection box 22a for the corresponding non-spam mail. (In this case, one case) and the button 23f to exclude from spam is pressed. The processing of the spam registration cancellation unit 34b is activated in accordance with the instruction of the recipient, and one non-spam mail with a check is moved to the reception log and log identification information of the non-spam mail registered as spam information , Each lexical phrase and the appearance degree of each lexical phrase are excluded from the spam information management database 4. Even users other than mail recipients can remove non-spam mail from spam in the same manner on behalf of the mail recipients if they are shared spam mails.

  Further, as described above, instead of selecting a plurality of received mails in the spam mail list area 22 and performing processing such as canceling spam registration at a time, as shown in FIG. You may perform processing such as spam registration cancellation one by one. In this case, an operation instruction may be performed using each button arranged in the operation instruction area 23 'at the bottom of the screen. Since each button is the same as described above, description thereof is omitted.

  Further, the screen layouts shown in FIGS. 3A and 3B are not limited to this form, and the layout may be different as long as each column is provided.

  As described above with reference to FIGS. 2 and 3, all e-mails relayed by the data relay device can be read on behalf of the recipients even by users other than the recipients. When a non-spam mail is found in an email or spam log, the spam registration means and the spam registration release means are activated by pressing the spam registration and spam release buttons by each recipient or a user other than the recipient. The And since all users inside the company can manage spam / non-spam, it is possible to realize a spam mail elimination system that eliminates the need for a mail manager.

  Finally, FIG. 4 is a flowchart for explaining processing by each means of the spam mail exclusion system of the present invention.

  First, a received mail addressed to a company internal terminal relays the data relay apparatus 3 (step S1). Subsequently, each word / phrase in the received mail is extracted by the word / phrase extracting means 31 (step S2). Spam probability calculated by the received mail determination means 32 from the appearance of a plurality of lexical characters in the received mail extracted by the lexical extraction means 31 and each lexical word of each shared email managed by the spam information management database 4 The degree of spam probability of the received mail is calculated by comparing the degree of the received mail and spam / non-spam of the received mail is determined (step S3).

  Then, the received mail classification means 33 classifies and stores the received mail in the spam log if the determination result is spam, and classifies and stores the received mail in the reception log if the determination result is non-spam (step S4). ).

  Then, each recipient in the company reads the received mail in the reception log and the spam log, and confirms whether or not the mail has been correctly distributed (step S5). If correctly sorted (Yes in step S5), the judgment of each recipient matches the judgment result of the received mail judgment means 32, so the process ends without moving the received mail to the correct log. To do.

  On the other hand, if it is not correctly distributed (No in step S5), the received mail determination means 32 has made a determination result contrary to the intention of each receiver, so the following instructions are given by each receiver. (Step S6).

[When SW = Exclude button from spam is pressed]
Since the received mail that should be determined as non-spam is sorted in the spam log, each recipient presses the button 23f for canceling spam after giving a check to the mail selection box 22a of the junk mail list screen 20a, A button 23'f for directly canceling spam from the spam mail text screen 20b is pressed. As a result, the spam registration cancellation unit 34b is activated, and the log identification information of each received mail registered in the spam information management database 4, each lexical word, and the appearance degree of each lexical word are excluded, and at the same time, the received mail is (Step S7).

[When SW = spam registration button is pressed]
In the case where spam mail to be determined as spam is distributed to the reception log, each recipient presses the button 13d for registering as spam after checking the mail selection box 12a on the received mail list screen 10a, A button 13'd for directly registering as spam is pressed from the body screen 10b of the received mail. As a result, the spam registration unit 34a is activated, and the spam identification management database 4 newly registers the log identification information of each junk mail, each lexical word, and the appearance degree of each lexical word, and at the same time, the junk mail is moved to the spam log. (Step S8).

  And the spam information management database 4 is learned by step S7 or step S8, and all the processes are completed.

It is the (A) block diagram explaining the spam mail exclusion system of this invention, (B) The block diagram of a spam information management database. It is an example of the screen of (A) received mail list display screen and (B) received mail text of the present invention. It is an example of the screen of (A) Junk mail list display screen of this invention, (B) Junk mail text. It is a flowchart which shows the process of the spam mail exclusion system of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Internet server 2 Terminal in company 3 Data relay apparatus 31 Lexical extraction means 32 Received mail determination means 33 Received mail classification means 34 Learning means 34a Spam registration means 34b Spam registration cancellation means 4 Spam information management database

Claims (3)

A data relay device was installed between the Internet server and the company terminal, and all the email information relayed by the data relay device was recorded in a log and shared by the company terminal. In the junk e-mail rejection system that eliminates the reception of junk e-mail among the e-mails,
Lexical extraction means for extracting each lexical word appearing from the header and body of the received mail relayed through the electronic mail or the data relay device already shared;
Log identification information indicating the log of the shared email, each lexical word extracted from the electronic mail corresponding to the log identification information by the lexical extraction means, and each extracted lexical word A spam information management database for managing the spam log to be stored and the appearance degree appearing in the reception log for storing non-spam mail;
Comparing a plurality of lexical words extracted by the lexical extraction means from the received mail addressed to each terminal in the company and a spam probability degree of each lexical word calculated from the appearance degree of the spam information management database , Received mail determination means for determining spam / non-spam of the received mail by calculating a spam probability of the received mail;
Received mail classification means for classifying the received mail into the spam log or the received log based on a determination result;
For the received mail classified into the spam log or the received log according to the determination result against each recipient's will, an instruction of spam / non-spam is given from the terminal in the company according to the judgment of each recipient. An unsolicited mail elimination system, characterized in that the system is composed directly with learning means for improving the accuracy of spam / non-spam judgment of the received mail judgment means.   The learning means is classified into the reception log according to the determination result of the received mail determination means as non-spam mail, but for each received mail determined to be spam mail by each receiver, When a spam registration button provided on the received log list or the received mail text screen is clicked once by an instruction, the log identification information, each extracted word and phrase, and the degree of appearance are displayed in the spam information management database. The spam mail eliminating system according to claim 1, further comprising a spam registration unit that registers and moves the received mail from the reception log to the spam log.   The learning means is classified into the spam log according to the determination result of spam received by the received mail determination means, but for each received mail determined as non-spam mail by each receiver, When a spam exclusion button provided on the spam log list or the body text screen of the received mail is clicked by one instruction, the log identification information registered in the spam information management database, the extracted words and phrases, The spam mail rejection system according to claim 1, further comprising: a spam registration cancellation unit that excludes the appearance degree and moves the received mail from the spam log to the reception log.

Images