JP2022031421A - Information processor, method for processing information, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a simultaneous sending of mails to addresses which are prohibited from being sent simultaneously.

SOLUTION: The information processor of an embodiment includes: first calculation means for calculating a feature amount for each of combinations of a specific number of combinations of mail addresses set for a sending target mail; and transmission means for transmitting a notification which shows that a combination of mail addresses which are possibly prohibited from being sent simultaneously is set to be the address of the sending target mail, by using the largest feature amount or the smallest feature amount of the feature amounts obtained by the first calculation means.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2022,JPO&INPIT

Description

The present invention relates to an information processing apparatus, an information processing method and a program.

As a countermeasure against erroneous sending of e-mail (hereinafter, also referred to simply as "mail"), the destination address set for the e-mail to be sent is an address that has been sent in the past using the e-mail transmission history. A technique for determining the presence or absence has been conventionally known. With such technology, for example, it is determined that sending an email to an address that has been sent in the past is safe, and sending an email to an address that has not been sent in the past is dangerous (risk). ..

Japanese Unexamined Patent Publication No. 2007-87327

However, with the above-mentioned conventional technique, it is not possible to prevent the transmission of an e-mail when an e-mail address for which simultaneous transmission (broadcasting) is prohibited is set.

For example, it is assumed that the e-mail addresses for which broadcasting is prohibited are "e-mail address A" and "e-mail address B", and both the e-mail address A and the e-mail address B are addresses that have been sent in the past. In this case, the e-mail in which both the e-mail address A and the e-mail address B are set as the destination address is determined to be safe, and the e-mail transmission cannot be prevented.

The present invention has been made in view of the above points, and an object of the present invention is to prevent simultaneous transmission of e-mails to destinations for which simultaneous transmission is prohibited.

In order to achieve the above object, the information processing apparatus according to the embodiment calculates the feature amount for each combination of a predetermined number of the plurality of e-mail addresses set as the destinations of the e-mails to be sent. A combination of an e-mail address that may be prohibited from broadcasting using the first calculation means and the maximum or minimum feature amount among the feature amounts calculated by the first calculation means is sent to the destination. It has a transmission means for transmitting a notification indicating that it is set.

It is possible to prevent simultaneous sending of e-mails to destinations for which simultaneous sending is prohibited.

It is a figure for demonstrating an example of the combination of destinations which may be prohibited from simultaneous transmission. It is a figure which shows an example of the whole structure of the mail erroneous transmission prevention system which concerns on this embodiment. It is a figure which shows an example of the learning data. It is a flowchart which shows an example of the learning model creation process which concerns on this embodiment. It is a figure for demonstrating an example of percentile. It is a flowchart which shows an example of the erroneous transmission risk determination / notification processing which concerns on this embodiment. It is a figure (the 1) for demonstrating a specific example of this embodiment. It is a figure (the 2) for demonstrating a specific example of this embodiment.

Hereinafter, embodiments of the present invention (hereinafter, also referred to as “the present embodiments”) will be described. In the present embodiment, an e-mail erroneous transmission prevention system 1 capable of preventing simultaneous e-mail transmission to a destination for which simultaneous transmission is prohibited will be described. In the present embodiment, the "destination" refers to an e-mail address to which an e-mail is sent. Therefore, the destination refers to the e-mail address set in the To field of the e-mail, the e-mail address set in the Cc field, and the e-mail address set in the Bcc field. However, the destination points to at least one of the e-mail address set in the To field of the e-mail, the e-mail address set in the Cc field, and the e-mail address set in the Bcc field. You may.

In the present embodiment, there is a possibility that simultaneous transmission is prohibited by this learning model after creating a learning model using the transmission history of past emails (hereinafter, also referred to as "transmission log"). Determines whether the combination of recipients is set for the email to be sent. This makes it possible to prevent a situation in which an email is sent to a destination that may be prohibited from simultaneous transmission.

<Destinations where simultaneous transmission may be prohibited>
Here, an example in the case where it is determined that the combination of destinations may be prohibited from simultaneous transmission in the present embodiment will be described with reference to FIG. FIG. 1 is a diagram for explaining an example of a combination of destinations where simultaneous transmission may be prohibited. Note that A, B, C and D in FIG. 1 represent e-mail addresses, respectively.

As shown in FIG. 1, for example, in the transmission log, an e-mail with a destination pattern 1 in which an e-mail address A, an e-mail address B, and an e-mail address D are set as a destination, and an e-mail address A, an e-mail address B, and an e-mail as a destination. The mail of the destination pattern 2 in which the address C is set, the mail of the destination pattern 3 in which the mail address A and the mail address D are set as the destination, and the mail of the destination pattern 4 in which the mail address B and the mail address C are set as the destination. It is assumed that there is an email.

At this time, in the present embodiment, for example, a combination of e-mail addresses that does not exist in the transmission log or a combination of e-mail addresses that are transmitted a small number of times is used as a combination of destinations that may be prohibited from simultaneous transmission. In the example shown in FIG. 1, since the combination of the e-mail address C and the e-mail address D does not exist in the transmission log, this combination is set as a combination in which simultaneous transmission may be prohibited.

The e-mail address A, the e-mail address B, the e-mail address C, and the e-mail address D are e-mail addresses that have been sent in the past by themselves. Therefore, in the conventional technique (for example, a technique for determining whether or not the destination address set in the e-mail to be sent is an address that has been sent in the past), the e-mail address C and the e-mail address D are different. E-mail transmission of the set e-mail is determined to be safe e-mail transmission.

On the other hand, in the present embodiment, a combination that does not exist in the past or a combination that has a small number of appearances (or frequency of appearance) is a combination of e-mail addresses that may be prohibited from simultaneous transmission. Therefore, it is possible to determine that the e-mail transmission of the e-mail in which the e-mail address C and the e-mail address D are set is the e-mail transmission having a risk of erroneous transmission, and it is possible to prevent such an e-mail transmission.

<Overall configuration of email mis-sending prevention system 1>
Next, the overall configuration of the mail erroneous transmission prevention system 1 according to the present embodiment will be described with reference to FIG. FIG. 2 is a diagram showing an example of the overall configuration of the mail erroneous transmission prevention system 1 according to the present embodiment.

As shown in FIG. 2, the mail erroneous transmission prevention system 1 according to the present embodiment includes an erroneous transmission prevention server 10 and one or more mail transmission / reception terminals 20. The erroneous transmission prevention server 10 and each mail transmission / reception terminal 20 are connected to each other so as to be communicable via an arbitrary communication network such as an in-house LAN (Local Area Network).

When the erroneous transmission prevention server 10 receives an email from the email transmission / reception terminal 20, the email address that may be prohibited from simultaneous transmission using a learning model created in advance (hereinafter, also referred to as "simultaneous transmission prohibited address"). It is determined whether or not the combination of (represented) is set as the destination of the mail. Then, the erroneous transmission prevention server 10 indicates that there is a risk of erroneous transmission because the destination includes, for example, a combination of e-mail addresses that may be prohibited from simultaneous transmission, depending on the determination result. Notify the terminal 20. The learning model is created by a statistical method using transmission logs, as will be described later. The erroneous transmission prevention server 10 is realized by one or a plurality of computers.

The erroneous transmission prevention server 10 may be a server that functions as a mail server, or may be a server provided separately from the mail server.

The mail transmission / reception terminal 20 is a terminal used by the sender (user) of the mail. The mail transmission / reception terminal 20 creates and sends mail (including forwarding and replying to mail) according to the operation of the user. Further, the mail transmission / reception terminal 20 receives the mail forwarded from the mail server. The mail transmission / reception terminal 20 is realized by, for example, a PC (personal computer) or the like. However, the mail transmission / reception terminal 20 may be realized by various devices or devices such as smartphones and tablet terminals, for example.

In this embodiment, the sender is a member of a company, an organization, etc. (for example, an employee of the company, an employee of the organization, a student of a school, etc.). Hereinafter, the members of these companies and organizations are simply collectively referred to as "employees".

Here, the erroneous transmission prevention server 10 has a transmission / reception unit 101, a determination unit 102, a notification unit 103, and a model creation unit 104 as functional units. Each of these functional units is realized by a process of causing a CPU (Central Processing Unit) or the like to execute one or more programs installed in the erroneous transmission prevention server 10.

Further, the erroneous transmission prevention server 10 has a learning data storage unit 111 and a learning model storage unit 112 as storage units. Each of these storage units can be realized by using, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), or the like. However, at least one of these storage units may be realized by using a storage device or the like connected to the erroneous transmission prevention server 10 via a communication network.

The transmission / reception unit 101 receives an email from the email transmission / reception terminal 20 (hereinafter, also referred to as “sending target email”). Further, the transmission / reception unit 101 transmits (forwards) the transmission target mail to another device (for example, a mail server or the like) according to the determination result by the determination unit 102.

Using the learning model stored in the learning model storage unit 112, the determination unit 102 determines whether or not the destination of the transmission target mail received by the transmission / reception unit 101 includes a combination of simultaneous transmission prohibited addresses. At this time, the determination unit 102 decomposes the e-mail addresses set as the destinations of the e-mails to be sent into combinations of two, and then calculates the score of each combination by the learning model to obtain the simultaneous transmission prohibited addresses. Determine if the combination is set for the destination.

When it is determined that the destination of the e-mail to be sent includes a combination of addresses for which simultaneous transmission is prohibited, the notification unit 103 notifies the e-mail transmission / reception terminal 20 of the sender of the e-mail to be sent that there is a risk of erroneous transmission. Send.

The model creation unit 104 creates a learning model for determining whether or not the combination of e-mail addresses is a combination of simultaneous transmission prohibited addresses by using the learning data stored in the learning data storage unit 111.

The learning data storage unit 111 stores learning data for creating a learning model. As the learning data, the transmission log of the mail transmitted from the mail transmission / reception terminal 20 in the past is used. Here, an example of learning data is shown in FIG. FIG. 3 is a diagram showing an example of learning data.

As shown in FIG. 3, one learning data (that is, one transmission log) includes a "destination address (To)" indicating an email address set in the To column and an email address set in the Cc column. "Destination address (Cc)" indicating, "Destination address (Bcc)" indicating the e-mail address set in the Bcc field, "Subject" indicating the character string set in the subject field, and the text field. Includes "body" indicating the set character string. The e-mail address set in the destination address (To), the destination address (Cc), and the destination address (Bcc) is the "destination".

It is sufficient that one or more e-mail addresses are set in total for the destination address (To), the destination address (Cc), and the destination address (Bcc). Further, each learning data does not necessarily include the subject and the text, for example, only the destination (destination address (To), destination address (Cc) and destination address (Bcc)) from the transmission log. It may be extracted and used as learning data.

The transmission log as learning data may be acquired from, for example, an on-premises mail server or a cloud-type Web mail server. Further, for example, it may be acquired from a device or device that collects transmission logs between the mail transmission / reception terminal 20 and the mail server. However, at this time, for example, the transmission log in which only one e-mail address is set as the destination may not be used as the learning data (in other words, the transmission in which two or more e-mail addresses are set as the destination). Only the log may be used as training data.)

The learning model storage unit 112 stores the learning model created by the model creation unit 104.

The configuration of the mail mistransmission prevention system 1 shown in FIG. 1 is an example, and may be another configuration. For example, even if the erroneous transmission prevention server 10 is composed of a learning device that creates a learning model and a determination device that determines whether or not a combination of simultaneous transmission prohibited addresses is included in the destination using the learning model. good. In this case, the learning device includes a model creation unit 104 and a learning data storage unit 111, and the determination device includes a transmission / reception unit 101, a determination unit 102, a notification unit 103, and a learning model storage unit 112. Further, at this time, at least one of the learning device and the determination device may be realized by a cloud server connected via a communication network such as the Internet.

Alternatively, for example, the erroneous transmission prevention server 10 may be included in the mail transmission / reception terminal 20.

<Learning model creation process>
Next, a process of creating a learning model using the learning data stored in the learning data storage unit 111 will be described with reference to FIG. FIG. 4 is a flowchart showing an example of the learning model creation process according to the present embodiment. It is assumed that a plurality of learning data (that is, a learning data set) are stored in the learning data storage unit 111.

When the learning model is not stored in the learning model storage unit 112, the subsequent learning model creation process needs to be executed before the erroneous transmission risk determination / notification process described later is executed. The learning model creation process is, for example, periodically executed when the learning model is recreated and the learning model stored in the learning model storage unit 112 is updated, or the management of the erroneous transmission prevention server 10 is performed. It may be executed at a timing or the like specified by a person or the like.

First, the model creation unit 104 acquires training data from the training data storage unit 111 (step S101). At this time, the model creation unit 104 may acquire all the learning data stored in the learning data storage unit 111, or acquire only the learning data satisfying a predetermined condition (for example, the transmission log for the last one year). Is acquired as learning data, etc.).

Next, the model creation unit 104 decomposes the e-mail address set as the destination of each learning data into a combination of two (step S102).

For example, "email address A", "email address B", and "email address" as destinations of certain learning data (destination address (To), destination address (Cc), and destination address (Bcc)). When "C" and "e-mail address D" are set, the model creation unit 104 decomposes these e-mail addresses into the following two combinations.

・ {Email address A, email address B}
・ {Email address A, email address C}
・ {Email address A, email address D}
・ {Email address B, email address C}
・ {Email address B, email address D}
・ {Email address C, email address D}
In this way, the model creation unit 104 decomposes the e-mail addresses set as the destinations of each learning data into a combination of two. Generally, when n e-mail addresses are set as the destination of one learning data, these n e-mail addresses are decomposed into a combination of _{two n} Cs. Hereinafter, for the sake of simplicity, for example, {e-mail address A, e-mail address B} is also simply referred to as {A, B}. Similarly, other combinations of e-mail addresses are also expressed as {A, C}, {A, D}, {B, C} and the like.

Next, the model creation unit 104 calculates the number of appearances for each combination in the entire learning data acquired in step S101 above (step S103).

For example, among the learning data acquired in step S101 above, when the learning data in which the combination of {A, B} is included in the destination is N ₁ , the number of occurrences of {A, B} is N ₁ . Similarly, for example, when there are N ₂ learning data in which the combination of {A, C} is included in the destination among the learning data acquired in step S101 above, the number of occurrences of {A, C} is N ₂ . It becomes.

In this way, the model creation unit 104 calculates the number of appearances of the combination in the entire training data for each combination decomposed in step S102.

Next, the model creation unit 104 converts the number of appearances calculated in step S103 into percentiles, and calculates a score for each combination (step S104). Here, the percentile is an arrangement of combinations in order of the number of appearances. As the score, an arbitrary score using the combination after conversion to the percentile can be used, and for example, the order of the number of appearances (that is, the order of the combination) can be used.

Here, an example of the percentile is shown in FIG. The percentiles shown in FIG. 5 are the combinations arranged in the order of the number of appearances (in descending order of the number of appearances). In the percentile shown in FIG. 5, the order of the number of appearances is used as the score, and is expressed by (score, number of appearances). For example, (1,97) is the score (1) and the number of occurrences (97) corresponding to a certain combination having the highest number of appearances. Further, for example, (2,63) is the score (2) and the number of appearances (63) corresponding to a certain combination having the second highest number of appearances.

In this way, the model creation unit 104 calculates the scores corresponding to the combinations by converting the combinations decomposed in the above step S102 into percentiles. The score and the score corresponding to the combination may be referred to as "feature amount", "feature amount of combination" and the like, respectively.

Next, the model creation unit 104 stores in the learning model storage unit 112 as a learning model a set in which each combination decomposed in the above step S102 and each score calculated in the above step S104 are associated with each other. (Step S105).

For example, the combinations decomposed in step S102 above are {A, B}, {A, C}, {A, D}, {B, C}, {B, D}, {C, D}. The score of {A, B} calculated in step S104 above is s ₁ , the score of {A, C} is s ₂ , the score of {A, D} is s ₃ , and the score of {B, C} is s. _4. When the score of {B, D} is s ₅ and the score of {C, D} is s ₆ , the model creation unit 104 may, for example, {{A, B} = s ₁ , {A, C}. = S ₂ , {A, D} = s ₃ , {B, C} = s ₄ , {B, D} = s ₅ , {C, D} = s ₆ } as a learning model in the learning model storage unit 112 save.

As described above, the erroneous transmission prevention server 10 according to the present embodiment uses the past transmission log as learning data and calculates a score from the number of appearances of the set of e-mail addresses set as the destination of each learning data for learning. Create a model. As a result, a learning model is created in which the set of email addresses with a low number of appearances (or frequency of appearance) has a relatively low score, and the set of email addresses with a high number of occurrences (or frequency of appearance) has a relatively high score. Will be done.

In the present embodiment, the score of the combination is calculated for each combination composed of two e-mail addresses, but the score of the combination is calculated for each combination composed of a predetermined number of e-mail addresses. May be good. As a result, for example, in the erroneous transmission risk determination / notification process described later, it becomes possible to determine whether or not a combination composed of a predetermined number of email addresses is included in the transmission target email.

<Risk judgment / notification processing for erroneous transmission>
Next, using the learning model stored in the learning model storage unit 112, it is determined whether or not there is a risk of erroneous transmission in the email to be transmitted, and if it is determined that there is a risk of erroneous transmission, an email to that effect is sent. The process of notifying the transmission / reception terminal 20 will be described with reference to FIG. FIG. 6 is a flowchart showing an example of erroneous transmission risk determination / notification processing according to the present embodiment.

First, the transmission / reception unit 101 receives the transmission target mail transmitted from the mail transmission / reception terminal 20 (step S201).

Next, the determination unit 102 decomposes the e-mail address set as the destination of the e-mail to be sent received in step S201 into a combination of two (step S202). This is the same as step S102 above.

Next, the determination unit 102 calculates the score of each combination decomposed in the above step S202 using the learning model stored in the learning model storage unit 112 (step S203).

For example, the learning model is {{A, B} = s ₁ , {A, C} = s ₂ , {A, D} = s ₃ , {B, C} = s ₄ , {B, D} = s ₅ , {C, D} = s ₆ }, and when the combinations obtained in step S202 above are {A, B}, {B, D} and {C, D}, the determination unit 102 scores. Calculate s ₁ , s ₅ and s ₆ , respectively.

Next, the determination unit 102 determines whether or not the maximum score (maximum score) among the scores calculated in step S203 is equal to or higher than a predetermined threshold value (step S204). That is, the determination unit 102 compares the maximum score with the threshold value and determines whether or not the maximum score is equal to or greater than the threshold value. It should be noted that such a threshold value may be set for each sender of an email, or may be set for each group (for example, an organization). In addition, an arbitrary value may be set as the threshold value. For example, when the value is not explicitly set, the average value of the threshold values set by other users in the group to which the user belongs is set. May be good. Further, for example, when the frequency of sending and receiving e-mails differs depending on the time, the threshold value may be automatically changed depending on the time.

In this embodiment, a learning model is used in which the score increases as the number of appearances (or frequency of appearance) decreases, but depending on the learning model, the score decreases as the number of appearances (or frequency of appearance) decreases (that is,). , The score increases as the number of appearances (or frequency of appearance) increases). In this case, the determination unit 102 may determine in step S204 whether or not the minimum score is equal to or less than a predetermined threshold value, or the reciprocal of the score or the number obtained by multiplying the score by a negative value is predetermined. It may be determined whether or not it is equal to or greater than the threshold value.

If it is not determined in step S204 that the maximum score is equal to or higher than a predetermined threshold value, the transmission / reception unit 101 forwards the transmission target mail to another device (for example, a mail server or the like) (step S205). In this case, it is considered that the destination of the email to be sent does not include the combination of the simultaneous transmission prohibited addresses (it is unlikely that the combination of the simultaneous transmission prohibited addresses is included), and there is no risk of erroneous transmission. (Or the risk of erroneous transmission is low).

On the other hand, when it is determined in step S204 that the maximum score is equal to or higher than a predetermined threshold value, the notification unit 103 notifies the mail transmission / reception terminal 20 of the sender of the transmission target mail that there is a risk of erroneous transmission. Is transmitted (step S206). In this case, it is considered that the destination of the email to be sent contains a combination of simultaneous transmission prohibited addresses (it is highly likely that a combination of simultaneous transmission prohibited addresses is included), and there is a risk of erroneous transmission. (Or the risk of erroneous transmission is high).

As described above, the erroneous transmission prevention server 10 according to the present embodiment uses the learning model created in advance to determine whether or not the destination of the email to be transmitted includes a combination of simultaneous transmission prohibited addresses. According to this determination result, a notification that there is a risk of erroneous transmission is transmitted to the mail transmission / reception terminal 20. This makes it possible to prevent the occurrence of erroneous transmission such as the transmission target email being broadcast to an email address that may be prohibited from simultaneous transmission.

Finally, a specific example of the present embodiment will be described with reference to FIGS. 7 and 8. 7 and 8 are diagrams for explaining a specific example of the present embodiment.

In the specific example shown in FIG. 7, a case where a learning model is created using four learning data is shown. Note that E to H each represent an e-mail address.

At this time, in the example shown in FIG. 7, the number of appearances of the combination of {E, F}, {E, G}, {E, H}, {F, G}, {G, H} in the entire training data is different, respectively. It is assumed that the score is 2,1,2,2,1,0 and the scores are 20, 70, 20, 20, 70, 100, respectively. On the other hand, it is assumed that the destination address of the mail to be sent includes E, G and H. In this case, since the destination address of the email to be sent includes G and H (that is, a combination that does not appear in the learning data), the maximum score of the email to be sent is 100.

Further, the specific example shown in FIG. 8 shows a case where a learning model is created using a large amount of learning data including the above four learning data.

At this time, in the example shown in FIG. 8, the number of appearances of the combination of {E, F}, {E, G}, {E, H}, {F, G}, {G, H} in the entire training data is different, respectively. It is assumed that the numbers are 100, 20, 70, 80, 30, and 1, and the scores are 3,47,7,6,35,85, respectively. On the other hand, as in the above, it is assumed that E, G and H are included in the destination address of the mail to be sent. In this case, since the destination address of the email to be sent includes G and H (that is, the combination having the least number of appearances in the entire learning data), the maximum score of the email to be sent is 85.

The present invention is not limited to the above-described embodiment disclosed specifically, and various modifications and modifications can be made without departing from the scope of claims.

1 E-mail erroneous transmission prevention system 10 E-mail erroneous transmission prevention server 20 E-mail transmission / reception terminal 101 Transmission / reception unit 102 Judgment unit 103 Notification unit 104 Model creation unit 111 Learning data storage unit 112 Learning model storage unit

Claims (5)

A first calculation means for calculating a feature amount for a predetermined number of combinations among a plurality of e-mail addresses set as destinations of e-mails to be sent, and
Using the maximum or minimum feature amount among the feature amounts calculated by the first calculation means, it is shown that a combination of e-mail addresses for which broadcasting may be prohibited is set for the destination. The means of sending notifications and
Information processing device with. Using a plurality of e-mail addresses set as recipients of each e-mail included in the past e-mail transmission history, the number or frequency of appearance of the combination in the e-mail transmission history is calculated for each predetermined number of combinations. 2 calculation means and
A third calculation means for calculating the feature amount for the combination by converting the number of times or frequency calculated by the second calculation means into percentiles.
A model creation means for creating a model in which the combination and the feature amount for the combination are associated with each other.
Have,
The first calculation means is
The information processing apparatus according to claim 1, wherein the feature amount for the combination is calculated by using the model for each combination. The information processing apparatus according to claim 2, wherein the feature amount calculated by the third calculation means is an order when the number of times or the frequency is arranged in a predetermined order. The first calculation procedure for calculating the feature amount for the combination for each combination of a predetermined number of the plurality of e-mail addresses set as the destinations of the e-mails to be sent, and
Using the maximum or minimum feature amount among the feature amounts calculated by the first calculation procedure, it is shown that a combination of e-mail addresses for which broadcasting may be prohibited is set for the destination. Sending procedure to send notification and
Information processing method that the computer executes. A program for causing a computer to function as the information processing device according to any one of claims 1 to 3.

Images