AI Server Screening of SMS Authentication Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems are vulnerable to artificially inflated traffic attacks where fraudsters use bots to register fake accounts and generate large volumes of fake SMS authentication requests, leading to financial losses for servers paying for unnecessary SMS traffic.
Innovation Solution
A server equipped with processing circuitry and memory uses an artificial intelligence model to analyze features extracted from authentication requests, including single-event, multi-event, and country-event features, to identify abnormal message attacks by comparing output values against a threshold, thereby distinguishing legitimate from fraudulent traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the server processes all SMS authentication requests without detection, then the processing speed is maintained, but the server suffers financial losses from fake traffic attacks
Solution Approach 1:
The system performs preliminary detection of SMS authentication requests using an AI model before processing them. By analyzing features such as device information, telephone number patterns, and request characteristics in advance, the system identifies and blocks fake traffic before it reaches the SMS transmission stage, preventing financial losses while maintaining processing speed for legitimate requests
Solution Approach 2:
An AI detection model is introduced as an intermediary component between the SMS authentication request and the SMS transmission process. This intermediary analyzes multiple features including device information, telephone number patterns, and request characteristics to determine whether to allow or block the request, thereby protecting the system from fake traffic without affecting the speed of legitimate authentication processes
2Device complexity
If the server uses traditional detection methods, then the system complexity is low, but the detection accuracy of complex AIT attacks is insufficient
Solution Approach 1:
The system transforms the detection approach by changing parameters from simple rule-based checks to multi-dimensional feature analysis. It extracts and analyzes multiple parameters including device information, telephone number patterns, request characteristics, and temporal patterns, feeding these into an AI model that dynamically adjusts detection thresholds based on learned patterns, thereby achieving high detection accuracy for complex AIT attacks
Solution Approach 2:
The detection system combines multiple detection mechanisms into a composite approach: feature extraction from various data sources, AI model analysis, and threshold-based decision making. This composite detection system integrates different types of information (device data, communication patterns, request features) to achieve superior detection accuracy that surpasses individual detection methods
3Measurement precision
If the server implements comprehensive feature analysis, then the detection accuracy is high, but the processing time increases
Solution Approach 1:
The system implements partial feature analysis by prioritizing the extraction and analysis of the most discriminative features. Instead of analyzing all possible features equally, it focuses on key features such as device information, telephone number patterns, and request characteristics that provide the highest detection value, thereby maintaining high accuracy while reducing processing time through selective feature analysis
Data Source
AI summary
According to an embodiment, a server may include: at least one processor, comprising processing circuitry, and memory configured to store instructions, wherein the instructions are configured to, when executed by the at least one processor individually or collectively, cause the server to: receive a first message for an authentication request, identify information included in the first message, acquire at least one of a first feature acquired using information related to the authentication request among information included in the first message, a second feature acquired using information related to a device and a telephone number among information included in a plurality of first messages received during a designated time period, and/or a third feature acquired using information related to a device and a telephone number among information included in a plurality of first messages received from a designated country during a designated time period, input the at least one feature to an artificial intelligence model as an input value, and based on an output value output from the artificial intelligence model being greater than or equal to a threshold value, identify the first message as an attack of an abnormal message.