A method and system for preventing malicious search chat robot vulnerabilities
A chatbot and malicious technology, applied in the computer field, can solve the problems of prone to loopholes, a large number of sensitive words, difficult to cover completely, etc., to achieve the effect of protecting the company or platform and avoiding the harm of public opinion
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0056] see figure 1 As shown, the present invention discloses a method for preventing malicious search chat robot loopholes, comprising the following steps:
[0057] Receive the user's dialogue request;
[0058] Extract the user's personal information from the dialog request, and store the personal information in the user identity database; the personal information includes one or more of the user's IP address, device code, WeChat ID or QQ number;
[0059] Real-time monitoring of the chat process;
[0060] Monitor whether a specific event occurs in a certain number of dialogue times;
[0061] Based on the monitoring results, a corresponding dialogue strategy is adopted for the user.
Embodiment 2
[0063] On the basis of the foregoing embodiments, embodiments of the present invention may include the following:
[0064] In the embodiment of the present invention, after real-time monitoring of the chat process, the monitoring of whether a specific event occurs in a certain number of dialogue times includes:
[0065] Count the number of occurrences of a specific feature in a certain number of conversations.
[0066] In an application scenario, the specific features of this embodiment of the present invention include seven category features. Further, the occurrence times of specific features specifically include:
[0067] The characteristics of the first category include: asking for opinions and positive expressions of preferences, such as the number of occurrences of "you like", "you love" and "do you think";
[0068] The second category features include: asking for opinions and negative expressions of preferences, such as the number of occurrences of "you hate", "you hat...
Embodiment 3
[0094] On the basis of the foregoing embodiments, embodiments of the present invention may also include the following:
[0095] In an application scenario, after collecting a data set with 7 features, the next step is to find outliers, that is, malicious customers, that is, to analyze the data set formed by 7 features in a certain number of conversations within a certain period of time, see figure 2 As shown, it may specifically include:
[0096] 1. given sample data , the feature dimension is 7, and the recorded sample is , . Dither each discrete feature of each sample separately, after dithering , which increases random number between. The dithered data set is denoted as . The purpose of jitter is to prevent data overlap. here .
[0097] 2. In Choose d features arbitrarily in Correspondingly, the original data set is divided into m subsets, denoted as ,in .
[0098] 3. In each subset , calculate each sample anomaly score of .
[0099] A. ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


