A SYSTEM THAT ENABLES REAL-TIME DETECTION OF BANK IMPOSITION AND FRAUD RISKS ON SOCIAL MEDIA.

Abstract

This invention relates to a system (1) that scans current posts on social platforms to detect accounts that imitate the bank’s corporate identity and engage in fraudulent activities, coordinated attacks, accounts sharing negative opinions, complaints and positive / negative comments, analyzes these posts with artificial intelligence large language models (LLM) to detect unusual situations, report them, visually interpret them and generate alarms.

Description

1 TARIFF IMPOSTORING BANK IDENTITIES ON SOCIAL MEDIA AND... REAL-TIME DETECTION OF FRAUD RISKS A SYSTEM THAT ENABLES IT TO BE DONE Technical Area This invention scans recent posts on social platforms to identify the bank. accounts that impersonate corporate identity and engage in fraudulent activities, coordinated attacks, accounts sharing negative opinions, complaints, and 10 The identification of positive / negative comments and the processing of these posts are largely done by artificial intelligence. By analyzing with language models (Large Language Models - LLM), unusual situations are identified. Identifying, reporting, and visually interpreting situations, and It relates to a system that enables the generation of alarms. Previous Technique Fraud detection systems in the current state of the art generally process focusing on internal data and neglecting external data sources in the form of social media. Today, fraud detection and reputation management in the banking sector are 20 Management is generally carried out using systems based on transaction data. This The systems detect fraud by analyzing transactions in customer accounts. while focusing on identifying initiatives, banks' presence on social media platforms Fraudulent activities impersonating corporate identity, coordinated attacks, Reputational risks such as negative reviews, complaints, or service interruptions are considered. 25 This is being followed. Recent posts on popular social media platforms, Fake accounts committing fraud by impersonating bank identities, users sharing complaints or attacks related to service disruptions Threats like these cannot be monitored in real time by current systems. and cannot be analyzed. This situation leads to delayed detection of threats, reputational damage. this leads to losses and limits the banks' ability to proactively intervene. 2 Furthermore, complex social media data can be presented in a meaningful way. the need for visualization and support with early warning systems Because these external risks are not met, banks remain vulnerable to them. Therefore, considering the studies and shortcomings in the current technology, 5 When these measures are taken, threats coming through social media can be intercepted. Addressing the shortcomings of traditional systems, providing early access through real-time analysis. providing warnings, detecting threats in real time, early detection of attacks or anomalies and proactive measures by banks It appears that a system is needed that allows it to move in this way. 10 U.S. patent number US2017161745A1, which is included in the prior art. the document mentions fraud on social media accounts related to payment accounts We are talking about a system structured to detect certain things. According to the samples, a fraud detection computer server detects 15 fraudulent acts at a specific time. It collects information associated with numerous social media accounts throughout the period. Social media that meets the threshold for potential fraud associated with at least one type of payment account. A subset of media accounts can be identified, and each that meets this threshold a fraud detection computer server for a social media account, in question It determines the geographical location associated with the social media account. Then, 20 Geographical information of social media accounts that meet the potential fraud threshold A map view can be created to show their locations. Brief Description of the Invention The purpose of this invention is to scan recent posts on social platforms, impersonating the bank's corporate identity and engaging in fraudulent activities accounts, coordinated attacks, accounts sharing negative opinions, complaints, and The identification of positive / negative comments and the processing of these posts are largely done by artificial intelligence. By analyzing with language models, unusual situations can be detected, 30 3 reporting, visual interpretation and alarm generation It is about implementing a system developed to provide this. Another aim of this invention is to analyze social media data with large language models. analysis of reputational risks and fraudulent activities in real time 5 a system developed to enable detection and visualization to accomplish. Detailed Description of the Invention The "Banking on Social Media" project was carried out to achieve the purpose of this invention. Real-time detection of imitation and fraud risks. A "System Providing" is shown in the attached figure; Figure 1; Schematic view of the system that is the subject of the invention. 15 The parts shown in the figure are individually numbered, and these numbers correspond to... The corresponding answers are given below. 1. System 20 2. Electronic Devices 3. Interface 4. Database 5. Server D. External Server By scanning recent posts on social media platforms, the bank's corporate profile is analyzed. Accounts impersonating individuals and engaging in fraudulent activities are acting in a coordinated manner. attacks, accounts sharing negative opinions, complaints, and positive / negative 30 the detection of comments and the analysis of these posts using artificial intelligence and large language models 4 by analyzing, detecting and reporting unusual situations, visually developed to enable it to be interpreted and alarms to be generated. The subject of the invention is the system (1); - communicating using any remote communication protocol at least one electronic device configured to be (2), 5 - is being carried out on electronic device (2) and the bank manager or the security software team detects fraud and negative reviews to enable monitoring of results, incoming reports and alerts to ensure that necessary actions are taken by evaluating the situation, analysis 10 Displaying the results in the form of reports, graphs, and dashboards at least one interface structured to (3), - storing data obtained from external server (D), raw social storing media posts and metadata in an organized manner and prepare for analysis, analysis from the large language model Storing results, fraud and negative opinion detection 15 organized and ready-to-visualize the results. at least one database structured to make it (4), - Analyzing social media data, fraud and complaints, Developed to identify negative opinions in the form of bad reviews. using natural language processing techniques, obtaining analysis results, and 20 integrating with fraud prevention units within the bank, Account suspension and customer alerts in response to detected threats. Taking automatic or manual actions, results on the interface (3) presenting in the form of reports, graphs and dashboards, from the large language model Based on the fraud detection results received, 25 fake accounts To determine this, additional analysis is needed, such as account age and follower ratio. Classifying fake accounts based on user metadata. It includes at least one server (5) configured for this purpose. The electronic device (2) in the system (1) which is the subject of the invention can be accessed remotely from any 30 Communicating using the communication protocol, connecting with the server (5) to install and run at least one interface (3) on it It is being structured. The interface (3) in the system (1) which is the subject of the invention, on the electronic device (2) is being carried out and the bank manager or the security software team, 5 to enable monitoring of fraud and negative opinion detection results, incoming reports and warnings should be evaluated and necessary actions taken. to provide and display analysis results in the form of reports, graphs, and dashboards. It is structured accordingly. The database (4) in the system (1) which is the subject of the invention is obtained via the external server (D). storing the collected data, raw social media posts and their metadata. to store and prepare for analysis in an organized manner, large language Storing analysis results from the model is considered fraud and negative feedback. Organizing and storing the detection results and making them ready for visualization. 15 It is structured to bring about this. The server (5) in the system (1) which is the subject of the invention, social media application programming interface (API) and keywords related to bank names using an external server (D) to filter social media posts. real-time data collection including user count, follower count, account age, and location. It is configured to collect metadata. The server (5) is in the database. (4) Cleaning raw data, with artificial intelligence and natural language processing algorithms To perform sentiment analysis in terms of positive, negative, and neutral, and for the large language model. 25 to extract features such as keyword frequency and sentiment score is configured. The server (5) analyzes social media data, Identifying negative opinions in the form of fraud, complaints, and bad reviews. It is structured to use advanced natural language processing techniques. Server (5) receives the analysis results and prevents fraud within the bank. integrating with units, freezing accounts against detected threats, customer 30 Taking automatic or manual actions in the form of a warning, results interface (3) 6 It is structured to present this information in the form of reports, graphs, and dashboards. Server (5) on the fraud detection results from the large language model Additional analysis is needed to identify fake accounts, including account age and follower ratio. Classifying fake accounts based on user metadata such as: It is structured accordingly. 5 Industrial application of the invention In the system in question (1) the user, bank manager or security team interface (3) monitors threats. Server (5), social media application programming 10 Using their interfaces, they can find bank names and keywords related to fraud. It collects social media posts filtered by keywords and uses this data to gather information. It records the data in the database (4). The database (4) stores the raw data. The server (5) stores the data. It cleans, analyzes sentiment using artificial intelligence algorithms and natural language processing, and Extracts features for the large language model. Server (5) with large language models 15 It detects fraud and negative opinions and stores the analysis results in a database. (4) records. The results are used in two ways: Firstly, suspicion of fraud In this case, action is taken against the threats. Secondly, the results are interfaced (3) It transmits this information to the user, allowing them to monitor threats and respond quickly. can. 20 The system in question (1) is based on artificial intelligence large language models, natural language processing and using machine learning techniques to analyze social media data, It identifies suspicious activities. The system in question (1) identifies financial activities of banks. It helps prevent losses and reputational damage. 25 Based on these fundamental concepts, the invention's theme is "Banking on Social Media". Real-time detection of imitation and fraud risks. It is possible to develop a wide variety of applications related to "A System Providing (1)". and the invention cannot be limited to the examples described here, but mainly to claims 30 as stated.

Claims

7 REQUESTS 1. By scanning recent posts on social media platforms, the bank's corporate profile is analyzed. accounts that impersonate and engage in fraudulent activities, coordinated attacks, accounts sharing negative opinions, complaints, and 5 The system identifies positive / negative comments and analyzes these posts using artificial intelligence. By analyzing with large language models, unusual situations are detected. monitoring, reporting, visual interpretation, and alarms. developed to enable its production; - communicating using any remote communication protocol 10 at least one electronic device configured to be (2), - is being carried out on electronic device (2) and the bank manager or the security software team detects fraud and negative reviews to enable monitoring of results, incoming reports and alerts to ensure that necessary actions are taken by evaluating the situation, analysis 15 displaying the results in the form of reports, graphs, and dashboards. at least one interface structured to (3), - storing data obtained from external server (D), raw social storing media posts and metadata in an organized manner and prepare for analysis, analysis from the large language model 20 storing results, detecting fraud and negative opinions. organized and ready-to-visualize the results. containing at least one database (4) structured to make it so; - Analyzing social media data, fraud and complaints, 25 advanced tools for identifying negative opinions in the form of bad reviews using natural language processing techniques, obtaining analysis results, and integrating with fraud prevention units within the bank, Account suspension and customer alerts in response to detected threats. Taking automatic or manual actions, results on the interface (3) presenting in the form of reports, graphs and dashboards, 30 from the large language model fraud detection results on fake accounts 8 To determine this, additional analysis is needed, such as account age and follower ratio. Classifying fake accounts based on user metadata. a server characterized by having at least one server (5) configured for system (1).

2. Communicating using any remote communication protocol. to be, to connect to the server (5) and to have at least an interface on it (3) is characterized by the electronic device (2) configured to operate. A system like the one in Request 1 (1).

3. It is carried out on the electronic device (2) and the bank manager or The security software team's fraud and negative opinion detection results. to ensure monitoring, evaluating incoming reports and warnings to ensure that necessary actions are taken, analysis results; reports, Interface structured to display in the form of graphs and dashboards (3) 15 as in any of the above claims characterized by system (1).

4. Storing data obtained from the external server (D), raw social Organized storage of media posts and metadata and 20 preparing for analysis, analysis from the large language model storing results, fraud and negative opinion detection results. to store in an organized manner and prepare for visualization The above is characterized by the database structured as (4) a system like any of the requests (1). 25 5. Social media application programming interface and bank names. filtered social media posts using relevant keywords Collecting real-time follower count, account data via external server (D) 30 to collect user metadata such as age and location 9 from the above requests characterized by the configured server (5) a system like any other (1).

6. Cleaning the raw data in the database (4), artificial intelligence and natural language Sentiment analysis in the form of positive, negative, and neutral using processing algorithms 5 Keyword frequency, sentiment score for making and large language model The server (5) is configured to extract features such as a system like any of the above-mentioned requests (1).

7. Analyzing social media data, fraud and complaints, bad 10s Advanced natural language for identifying negative opinions in the form of comments characterized by the server (5) configured to use processing techniques. a system like any of the above-mentioned requests (1).

8. Obtaining analysis results and preventing fraud within the bank 15 integrating with units, freezing accounts against detected threats, Taking automated or manual actions in the form of customer alerts, and the results to present reports, graphs and dashboards on the interface (3) from the above requests characterized by the configured server (5) a system like any other (1). 20 9. Fraud detection results from the large language model are based on fake information. We need to conduct additional analysis to identify accounts, including account age and follower ratio. fake accounts based on user metadata like this 25 characterized by the server (5) configured to classify a system like any of the above requests (1).

Images