ARTIFICIALLY INFLATED TRAFFIC IN A2P MESSAGING: WINDOW-BASED DETECTION MODELS AND PREFIX-LEVEL ANALYSIS FOR FRAUD MITIGATION

Abstract

The article examines the growing problem of Artificially Inflated Traffic (AIT) in Application-to-Person (A2P) SMS delivery and introduces an interpretable, metadata-driven framework for detecting and mitigating such fraud in real time. The study highlights how AIT exploits legitimate messaging pathways, generating rapid or distributed bursts of outbound traffic toward sequential or closely grouped MSISDN ranges. These patterns impose substantial financial losses on enterprises and CPaaS providers while degrading delivery efficiency metrics.

To address these vulnerabilities, the paper proposes a prefix-level analytical approach based on fixed and rolling time-window aggregation, message-count and unique-MSISDN thresholds, and entropy-based evaluation of MSISDN distributions. The methodological toolkit includes large-scale empirical dataset analysis, comparative evaluation of detection latency and blocking efficiency, and integration of outcomes with previously introduced delivery performance metrics—the Intelligent Guaranteed Delivery Index (IGDP) and Price–Delivery Gap (PDG).

Experimental results show that fixed-window analysis achieves up to 96% mitigation for high-intensity AIT bursts, while rolling windows enable earlier detection of short-lived anomalies. The study also reveals that threshold-based detection models may misclassify legitimate marketing campaigns, emphasizing the need for auxiliary features such as sender behavior profiling, content-aware classification, and multi-resolution anomaly scoring. The paper concludes that transparent statistical models provide a robust baseline for AIT defense and outlines practical recommendations for enhancing fraud resilience in large-scale CPaaS infrastructures.