Information processing system that makes autonomous escalation decisions

Abstract

In autonomous execution of processing based on input messages, even when involving external transmission, external API integration, or handling of confidential information, the system should suppress erroneous execution due to unauthorized access or impersonation, and appropriately escalate to a human when necessary. [Solution] The present invention is an information processing system that acquires an input message, identifies candidate processes for execution, decides whether to execute autonomously or escalate based on urgency information, authorization determination results, and reliability information, executes the process when autonomous execution is performed, and generates and sends notification information to a recipient when escalation is performed.

Description

[[Technical Field]]

[0001] The present invention relates to information processing that executes processing in response to an input message, and particularly relates to an information processing system that controls whether to autonomously execute the processing or escalate it to an authorized person based on multi-axis evaluation. [[Background Art]]

[0002] In recent years, information processing technology that executes processing such as acquiring information, transmitting information, or changing the state of an external service based on a request input by a user in natural language has become widespread.

[0003] In this type of technology, while convenience is improved through process automation, risks such as external transmission that deviates from the requester's authority, unintended state changes via application programming interfaces of external services, or inappropriate handling of confidential information may become apparent.

[0004] Furthermore, when an input message includes expressions emphasizing urgency such as "right now" or "urgent", or includes indirect instructions such as "a third party has given permission", the possibility that social engineering induces inappropriate execution increases.

[0005] Furthermore, if authority and reliability are treated excessively conservatively, escalation occurs frequently which may disrupt operations. On the other hand, if excessive autonomous execution is allowed, accidents may occur. Therefore, an appropriate balance is required in practical operation. [[Prior Art Documents]] [[Patent Documents]]

[0006] [[Patent Document 1]] US20250007907A1 Specification [[Patent Document 2]] US11757872B2 Specification [[Patent Document 3]] US20230297962A1 Specification [[Patent Document 4]] US9123027B2 specification [Patent Document 5] International Publication Specification No. WO2017177076A1 [Overview of the project] [Problems that the invention aims to solve]

[0007] In recent years, there has been an increase in cases where users interact with external services and manipulate information using natural language input messages in information processing systems. However, in these types of automated systems, the risk of user misuse, inappropriate sharing and manipulation of confidential information, and erroneous execution due to social engineering, etc., is increasing.

[0008] Conventionally, risk-based multi-factor authentication technologies and access control technologies disclose dynamic requests and permission decisions for authentication means based on context and risk information (e.g., Patent Documents 1 and 2). Furthermore, while technologies related to approval workflow management provide manual approval processes and hierarchical notification control (e.g., Patent Document 3), they do not comprehensively determine whether processing itself can be autonomously executed in response to requests specified in natural language. In addition, technologies related to social engineering attack detection (e.g., Patent Document 4) and access control technologies for heterogeneous environments (e.g., Patent Document 5) provide detection and scoring for specific threats and environments, respectively, but do not comprehensively handle request meaning understanding, external API integration, permission determination, confidential information verification, and escalation decisions.

[0009] However, these prior technologies are biased towards individual authentication, access control, approval workflows, and attack detection, and do not provide a comprehensive control framework that includes control over processing execution based on natural language input, determination of autonomous execution and manual escalation based on multiple evaluation axes of urgency, authority, and reliability, and confidential information verification flows.

[0010] Therefore, even when processing is performed autonomously based on input messages, if the processing involves API integration with external services or handling of confidential information, there is a need for technology that comprehensively evaluates the scope of the requester's authority, reliability, and urgency, and controls whether or not to execute the processing safely and efficiently by combining confirmation and escalation as needed. [Means for solving the problem]

[0011] An information processing system according to one aspect of the present invention comprises: an acquisition unit that acquires an input message; a processing identification unit that identifies a candidate process to be executed based on the input message; an urgency generation unit that generates urgency information based on the input message and the process; an authority determination unit that determines whether the process is within a predetermined authority scope based on authority information; a reliability generation unit that generates reliability information indicating the reliability of the requester and the instructions contained in the input message; a decision unit that determines whether to autonomously execute the process or escalate it based on the urgency information, the authority determination result and the reliability information; an execution unit that executes the process when autonomous execution is determined; and a notification unit that generates notification information and sends it to a notification destination when escalation is determined. An information processing system is obtained. [Effects of the Invention]

[0012] According to the present invention, since autonomous execution or escalation can be determined by evaluating urgency, authority determination results, and reliability on multiple axes, it is possible to achieve operation that suppresses excessive escalation while suppressing erroneous execution caused by authority abuse or impersonation.

[0013] Furthermore, since the ability to control whether or not to execute processes involving external transmission or information acquisition, transmission, or state changes via application programming interfaces of external services is available, taking into account factors such as whether prior permission has been granted, accidents associated with external collaboration can be reduced.

[0014] Furthermore, by using acknowledgments and confirmations to control execution when a confidential request is detected, it is possible to deter inappropriate sharing or access to confidential information.

[0015] In addition, since the reliability information can reflect social engineering signs, inappropriate execution caused by misrepresentation of urgency, indirect instructions, and the like can be suppressed.

[0016] In addition, by including context such as the reason for the decision in the notification information, an authorized person can make a quick judgment, thereby reducing the operational load required for approval and handover. [BRIEF DESCRIPTION OF THE DRAWINGS]

[0017] [Figure 1] It is a configuration diagram of the present invention. [Figure 2] It is a functional block diagram of the present invention. [Figure 3] It is a flowchart of the present invention. [Figure 4] It is a sequence diagram of the present invention. [MODE FOR CARRYING OUT THE INVENTION]

[0018] <Details of the Embodiment> Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0019] Hereinafter, a mode for carrying out the present invention will be described with reference to the drawings. It should be noted that the following embodiments do not limit the present invention, and a person skilled in the art can modify and implement them as appropriate.

[0020] As shown in FIG. 1, the information processing system 1000 is configured to be able to communicate with a requester terminal 2000, and can execute an external action via an application programming interface of an external service 4000. The information processing system 1000 can transmit a notification or a confirmation request to an authorized person terminal 3000 and a confirmation destination terminal 3100.

[0021] The information processing system 1000 may be configured as, for example, a server device, and may be arranged in a cloud environment or an on-premises environment. Furthermore, part or all of the functions may be executed on the requester terminal 2000 side.

[0022] The information processing system 1000 can access the authorization information storage unit 5000, which stores authorization information 5100, and the history information storage unit 6000, which stores history information 6100. Furthermore, it can cooperate with the authentication infrastructure 7000 for identity verification.

[0023] As shown in Figure 2, the information processing system 1000 may include an acquisition unit 1100, a processing identification unit 1200, an urgency generation unit 1300, an authority determination unit 1400, a reliability generation unit 1500, a decision unit 1600, an execution unit 1700, a notification unit 1800, and an identity verification unit 1900.

[0024] The acquisition unit 1100 may receive input messages from the client terminal 2000, or acquire them as email, form input, voice recognition results, etc.

[0025] The processing identification unit 1200 identifies a candidate process to be executed based on the input message. Here, "process" is a concept that includes information acquisition, information transmission, state change, recording, or operations on external services 4000.

[0026] The urgency generation unit 1300 generates urgency information based on the input message and processing. The urgency information may be a score expressed on a predetermined scale, a deadline, or a set of features indicating urgency.

[0027] The authorization determination unit 1400 determines whether the processing falls within the predetermined scope of authorization based on the authorization information 5100. The authorization information 5100 may include the client's attributes, role, scope for the external service 4000, destination constraints, or permission conditions for each type of processing.

[0028] The authorization determination unit 1400 may determine whether the process is an external action involving transmission to an external device or acquisition of information, transmission of information, or change of state via the application programming interface of the external service 4000. Furthermore, if prior authorization has not been granted, the determination unit 1600 may output determination information regarding the external action so that it can select escalation.

[0029] The identity verification unit 1900 verifies the identity of the requester based on the requester identification information associated with the input message. Identity verification may be performed using any method, such as authentication token verification, signature verification, device verification, or referencing multi-factor authentication results.

[0030] The reliability generation unit 1500 generates reliability information indicating the reliability of the requester and the instructions contained in the input message. The reliability information may be generated based on the verification results by the identity verification unit 1900, the attributes of the requester, the contextual characteristics of the input message, or the history information 6100, etc.

[0031] The reliability generation unit 1500 may extract contextual features that may constitute social engineering, and expressions indicating indirect instructions based on third-party authorization. Based on the extraction results, reliability information may be reduced, or the selection of additional verification or escalation may be controlled to be more likely.

[0032] The decision unit 1600 decides whether to proceed with autonomous execution or escalation based on urgency information, authorization determination results, and reliability information. The decision unit 1600 may calculate an overall risk index based on these and decide to escalate if the overall risk index exceeds a predetermined threshold. The threshold may be changed depending on the client attributes or processing type.

[0033] The notification unit 1800 generates notification information and sends it to the recipient when escalation is decided. The notification information may include at least part of the input message, processing, and reason for the decision. The notification unit 1800 waits for a response from the recipient for a predetermined period and may resend the notification or notify an alternative recipient if there is no response.

[0034] The execution unit 1700 executes processing when autonomous execution is determined. If the processing is an external action, the execution unit 1700 may perform information acquisition, information transmission, or state change via the application programming interface of the external service 4000.

[0035] At least one of the urgency generation unit 1300, the authority determination unit 1400, the reliability generation unit 1500, and the notification unit 1800 may analyze the input message using a trained large-scale language model 1950 and extract features such as time constraints, the presence or absence of external actions, requests for money or contracts, confidentiality requests, or signs of social engineering.

[0036] The urgency generation unit 1300 may search for cases similar to the input message from the history information 6100 and correct the urgency information based on the search results. Similarity search may be implemented by keyword matching, a classifier, or vector representation, etc.

[0037] The reliability generation unit 1500 may detect expressions in the input message that request the sharing of confidential information or access to confidential information. In this case, the information processing system 1000 may send a confirmation request to a confirmation destination associated with the confidential information or an authorized person regarding the handling of said confidential information, and control autonomous execution or escalation depending on whether the confirmation response is affirmative or negative.

[0038] The notification unit 1800 may, during escalation, generate a summary of the input message, urgency information, authorization determination result, reliability information, and reason for the decision using a pre-trained large-scale language model 1950.

[0039] Next, with reference to Figure 3, an example of the processing procedure of the information processing system 1000 will be described. The processing procedure shown in Figure 3 is just one example, and the order of processing can be rearranged, or parts can be omitted or added.

[0040] The acquisition unit 1100 acquires the input message (step S101). Next, the processing identification unit 1200 identifies a candidate processing based on the input message (step S102). Subsequently, the urgency generation unit 1300 generates urgency information based on the input message and the processing (step S103).

[0041] Next, the authorization determination unit 1400 determines whether the processing is within the predetermined scope of authorization based on the authorization information 5100 (step S104). Next, the identity verification unit 1900 verifies the identity of the requester based on the requester identification information (step S105). Next, the reliability generation unit 1500 generates reliability information based on the identity verification result and contextual characteristics, etc. (step S106).

[0042] Next, if the reliability generation unit 1500 detects a confidentiality request, a confirmation request is sent to the confirmation destination terminal 3100 (step S108). Subsequently, if the confirmation response is positive, autonomous execution of the process is permitted, and if the confirmation response is negative or if an confirmation response is not obtained within a predetermined period, escalation may be selected (steps S109, S110).

[0043] Next, the decision unit 1600 decides whether to proceed with autonomous execution or escalation based on the urgency information, authorization determination result, and reliability information (step S111). If autonomous execution is decided, the execution unit 1700 executes the process (step S112).

[0044] On the other hand, if escalation is decided, the notification unit 1800 generates notification information and sends it to the authorized terminal 3000 (step S114). The notification unit 1800 waits for a response from the authorized terminal 3000 for a predetermined period, and if there is no response, it may resend the notification or notify an alternative notification destination (step S115).

[0045] Figure 4 is a sequence diagram illustrating the interaction between the requester terminal 2000, the information processing system 1000, the verification terminal 3100, the external service 4000, and the authorized terminal 3000, and corresponds to steps S101 to S115 in Figure 3.

[0046] Next, let's describe an example of a specific use case. For example, if a requester terminal 2000 inputs a message indicating that it will retrieve information stored in an external service 4000 and send it to a third party, and that information may be confidential, the information processing system 1000 can detect the confidential request, send a confirmation request to the confirmation terminal 3100, and control execution or escalation based on the confirmation response.

[0047] Furthermore, if the input message contains expressions indicating a deadline or urgency, the urgency generation unit 1300 can extract such expressions and reflect them in the urgency information. However, the decision unit 1600 makes a comprehensive judgment in conjunction with the authorization judgment results and reliability information, thereby suppressing autonomous execution even in cases of false urgency.

[0048] Furthermore, if the input message includes an indirect instruction that "a third party has authorized it," the reliability generation unit 1500 can extract that expression and prompt the user to choose to send a confirmation request or escalate the process, thereby preventing inappropriate execution due to unconfirmed instructions.

[0049] <Modification 1> The representation of the authorization information 5100 is not limited to role-based representation, but may also be attribute-based or scope-based representation for external services 4000, and the authorization determination unit 1400 can determine whether or not it is within the scope of authorization according to the representation. In other words, even if the method of describing the authorization information is different, a configuration can be realized that controls the autonomous execution or escalation of processing in response to input messages.

[0050] <Modification 2> The calculation of the overall risk index by the decision unit 1600 is not limited to rule-based calculations, but may also be implemented using a trained model or weighted calculation, and the thresholds may be dynamically changed according to the client attributes and processing type. In other words, even if the risk calculation method is different, execution controls based on multi-axis evaluation can be realized.

[0051] <Modification 3> The notification destination of the notification unit 1800 is not limited to a single authorized terminal 3000, but may be selected from multiple candidates depending on the processing type or risk level, and a hierarchical approval route may be used. In other words, even if the method of selecting the notification destination is different, human intervention can be achieved through escalation notifications.

[0052] <Modification 4> Confirmation of confidentiality requests is not limited to the same channel, but may also be achieved through confirmation via a different route, time-limited approval, or confirmation using a token. In other words, even if the confirmation method is different, it is possible to control whether execution is permitted or not based on the acceptance or rejection of the confirmation response.

[0053] <Modification 5> The use of the pre-trained large-scale language model 1950 is not limited to extraction and summarization, but may also be applied to preprocessing of similarity searches based on historical information 6100, or to verbalizing the reasons for decisions, while definitive information such as authorization information 5100 and authentication results may be guaranteed by a separate verification system. In other words, even if the scope of application of the language model differs, execution controls that balance security and operability can be achieved.

[0054] Each of the components described above may be implemented in a single device or distributed across multiple devices. Furthermore, the present invention may be realized as a program that causes a computer to function as one of the aforementioned components, and as a recording medium storing said program.

[0055] Furthermore, each embodiment and its variations can be combined in any way, and the present invention can be implemented without departing from the spirit of the invention even if certain components are omitted or replaced with other components.

[0056] <Example Hardware Configuration> Each of the functional blocks described above can be configured using hardware, a DSP (Digital Signal Processor), or software, for example, provided in a server device (terminal device). For example, when configured using software, each functional block is actually configured using a computer's CPU, RAM, ROM, etc., and is realized by the operation of programs stored on recording media such as RAM, ROM, hard disk, or semiconductor memory.

[0057] The embodiments described above are merely illustrative to facilitate understanding of the present invention and are not intended to limit its interpretation. The present invention can be modified and improved without departing from its spirit, and it goes without saying that the present invention includes equivalents thereof. [Explanation of Symbols]

[0058] 1000 Information Processing Systems 1100 Acquisition Department 1200 Processing Specification Unit 1300 Urgency generation part 1400 Authority Determination Unit 1500 Reliability Generation Unit 1600 Decision Section 1700 Execution Department 1800 Notification Department 1900 Identity Verification Department 1950 Pre-trained Large-Scale Language Models 2000 Client terminal 3000 Authorized Terminals 3100 Confirmation terminal 4000 External Services 5000 Authority Information Storage Unit 5100 Permission Information 6000 History Information Storage Unit 6100 History Information 7000 Authentication Infrastructure 180 External Skill Delivery Platform 190 Management entity 200 Large-Scale Language Models

Claims

1. An information processing system that autonomously executes or escalates processing in response to an input message, The acquisition unit acquires the aforementioned input message, A processing identification unit identifies a process that is a candidate for execution based on the input message, An urgency generation unit that generates urgency information indicating the degree of urgency based on the input message and the processing, An authority determination unit that determines whether the aforementioned process falls within a predetermined scope of authority based on authority information, A reliability generation unit generates reliability information indicating the reliability of the requester of the input message and the instructions contained in the input message, A decision unit that determines whether to autonomously execute the process or escalate the process based on the urgency information, the determination result of the authorization determination unit, and the reliability information, An execution unit that executes the process when autonomous execution is determined by the decision unit, When the decision unit determines that escalation is necessary, the notification unit generates notification information including at least part of the input message, the processing, and the reason for the decision by the decision unit, and sends it to the notification destination associated with the authorized person. An information processing system equipped with the following features.

2. The information processing system according to claim 1, The urgency generation unit is an information processing system that detects expressions indicating time constraints contained in the input message and generates urgency information based on the detection results.

3. The information processing system according to claim 1, An information processing system in which the authorization determination unit determines whether the process is an external action involving transmission to an external device or acquisition of information via an application programming interface provided to an external service, transmission of information, or change of state, and if it is an external action and prior authorization has not been granted by the authorization information, the decision unit outputs information to the decision unit for the decision unit to determine escalation.

4. The information processing system according to claim 1, An information processing system comprising: a reliability generation unit that detects an expression in the input message requesting the sharing of confidential information or access to confidential information, and when detected, sends a confirmation request regarding the request to a confirmation destination associated with the confidential information or the person authorized to handle the confidential information; and a decision unit that permits the autonomous execution of the process if the confirmation response to the confirmation request is affirmative, and decides to escalate the case if the confirmation response is negative or if no confirmation response is obtained within a predetermined period.

5. The information processing system according to claim 1, The reliability generation unit is an information processing system that extracts contextual features that may constitute social engineering and reduces the reliability information based on the extraction results.

6. The information processing system according to claim 1, An information processing system comprising at least one of the urgency generation unit, the authority determination unit, and the reliability generation unit, which uses a trained large-scale language model to analyze input natural language to extract from the input message at least a portion of time constraints, the presence or absence of external actions, requests for money or contracts, requests for sharing confidential information, and signs of social engineering.

7. The information processing system according to claim 1, The notification unit is an information processing system that generates a summary of the input message, the urgency information, the determination result of the authorization determination unit, the reliability information, and the reason for the decision of the decision unit using a trained large-scale language model, and transmits it to the notification recipient as notification information.

8. The information processing system according to claim 1, The decision unit is an information processing system that determines escalation when the overall risk index calculated based on the urgency information, the decision result of the authority determination unit, and the reliability information exceeds a predetermined threshold, and changes the threshold according to the attributes of the requester or the type of processing.

9. The information processing system according to claim 1, The notification unit is an information processing system that waits for a response from the notification recipient for a predetermined period of time, and performs follow-up control to resend the notification or notify an alternative notification recipient if no response is received.

10. The information processing system according to claim 1, An information processing system further comprising an identity verification unit that verifies the identity of the client based on client identification information associated with the input message, and a reliability generation unit that generates reliability information based on the verification results by the identity verification unit.

Images