An SMS aggregator using a Burundian operator’s network tripled its verified A2P traffic—from 6,000 to 17,000 messages a month—while keeping subscribers safe from spam and fraud. EW SMS Firewall made it possible by customizing the solution for a unique Apache Kafka architecture, integrating AI, and fine-tuning blocking rules to local market realities.
Deploying a Fraud Prevention System
We were tasked with implementing a fraud prevention system for a mobile operator in Burundi. This project came through our SMS aggregator partner, following a successful collaboration in Mozambique.
The Burundi project aimed to reach 10,000 international A2P messages per day, unlocking the expected revenue for both the mobile operator and the SMS aggregator handling commercial traffic.
The implementation of the fraud prevention system in Burundi took 4 months, including all necessary customizations. Working in parallel with the client helped ensure smooth testing and on-time project acceptance
EW SMS Firewall enables SMS aggregators and mobile operators to boost A2P monetization and shield subscribers from spam and fraud. With flexible filtering logic, built-in URL scanning, and AI-powered detection, the system quickly blocks unwanted messages. Advanced analytics provide full transparency and traffic control.
Integration of EW SMS Firewall with Apache Kafka
The mobile operator in Burundi built its messaging system on Apache Kafka instead of the traditional SS7 protocol. This posed two key challenges for Eastwind:
- Cut SMS processing time from 1 second to just 250 milliseconds to maintain consistent data writing
- Enable SRI (Send Routing Information) support to track recipient routing, which Apache Kafka doesn’t offer by default
To meet these requirements, we optimized the EW SMS Firewall and adjusted several internal processes. This enabled transparent message delivery via SRI requests and improved SMS processing speed to four times the standard rate.
Enabling SRI in the Apache Kafka API meant customizing both the EW SMS Firewall and the operator’s SMSC to handle request flow.
We worked in sync with the client, and both teams completed their tasks by the time the new feature was launched. This smooth collaboration helped avoid delays during testing and acceptance.
Andrey Simonov, Head of Mobile Messaging Group,
Computer Telephony Department, Eastwind
Tested at a projected load of 2,200 TPS, EW SMS Firewall ran smoothly at 3,500 TPS and beyond, guaranteeing reliable performance even as traffic scales.
In this project, the client hosts the Apache Kafka server where messages are received. EW SMS Firewall connects via dedicated topics to pull and push SMS. This architecture requires less hardware
Configuration of Fraud Blocking Rules
We then proceeded to configure rules to block “gray” A2P traffic.
Language-Aware Filtering for Better A2P Protection
To block illegal A2P traffic more effectively, we analyzed local linguistic habits and identified patterns in how words and phrases appear in both local and international SMS messages:
- What languages subscribers typically receive messages in
- Which words and phrases are most often used
- How often OTP codes show up in international traffic
Burundi is a multilingual country where English, French, and local dialects are widely used.
Following the analysis, the managed services team set up blocking rules targeting international fraud patterns, including OTP messages from messaging apps and social media platforms.
Soon after the fraud prevention system was deployed, the managed services team identified and blocked a sender distributing high volumes of OTP codes from international platforms through local delivery routes
Configuration of Rule Exceptions for Banking Transactions
Messages from global platforms like Apple, Google, and major social networks or messengers often contain not only OTP codes but also payment information. To prevent accidental blocking, we added exception rules to the blacklists, for example, allowing messages that include a transaction amount and currency. This approach preserved service quality while effectively filtering out fraud and spam.
EW SMS Firewall allows exporting individual charts for each service when required
Separation of Blocking Rules for Anonymous A2P and P2P SMS
To boost filtering accuracy without disrupting business communication, Eastwind designed targeted rules to block anonymous messages that contain OTP codes:
- In P2P channels, EW SMS Firewall blocks any message with an OTP code, regardless of other content or brand names
- In local A2P traffic, messages are automatically blocked in two scenarios:
- When the text contains only an OTP code
- When the text includes an OTP code and the name of a global brand
This logic helps protect the network from fraud while keeping legitimate local traffic—like brand promotions—flowing smoothly.
By tailoring blocking rules to each channel, we keep filtering sharp without disrupting business communications. A local brand can send access codes to customers without issue—but the same message from a private sender won’t make it through.
Irina Timofeeva,
Customer Success Manager, Eastwind
Manual rule setup by managed services teams takes time—and that can delay the response to new threats. The AI module steps in to speed up fraud and spam detection dramatically.
Activation of AI Module for Advanced Fraud Detection
The AI-driven anti-fraud module is integrated into EW SMS Firewall to automatically analyze SMS traffic and user behavior. It enables faster identification of A2P fraud and SIM boxes in the operator’s network. The AI module analyzes 100% of network traffic, delivering sharper insights and faster results than manual checks ever could.
Powered by the XLM-RoBERTa cross-lingual model, the AI module can:
- Scan SMS messages and uncover hidden authorization codes
- Understand multiple languages and pinpoint key elements like one-time passcodes in the message text
We trained the model using Named-Entity Recognition (NER), allowing it to identify and classify messages according to defined rules. Additionally, the model was trained to understand message context. As a result, it is capable of:
- Spotting disguised OTP codes
- Classifying incoming messages
- Flagging fraudulent activity in real-time traffic
At the Burundian mobile operator, the AI module runs in post-analysis mode, generating daily SMS traffic reports for the managed services team. Each report lists messages suspected of fraud over the previous 24 hours, helping the team quickly update blocking rules and respond to threats. Together with the client, we chose this model as the most cost-effective and responsive solution for current traffic levels. Real-time analysis, while possible, would demand far greater performance and investment.
EW SMS Firewall blocks thousands of fraudulent messages daily across the African operator’s network, most of which are delivered through P2P channels using SIM boxes
The Result: Fraud Down, Legitimate A2P Traffic Up
Six months after implementation, EW SMS Firewall has helped raise commercial SMS volumes to the target level of 10,000 international A2P messages per day—and beyond. This shows the anti-fraud solution is working as intended, securing the operator’s projected revenue from A2P traffic. And because financial transactions and payment alerts are excluded from filtering, user experience remains unchanged.
The volume of commercial A2P SMS grew from 6,000 to 17,000 messages per day. This is 1.7 times higher than the client’s initial target of 10,000
By combining broad-based rules with targeted filtering scenarios, the mobile operator blocked routes used for “gray” A2P SMS, resulting in a significant increase in legitimate traffic
Launching Voice Firewall Pilot
After implementing EW SMS Firewall, we ran a Voice Firewall pilot with the mobile operator to analyze flash call volumes—those one-ring calls carrying verification codes—and see whether they require dedicated controls.
The pilot revealed that subscribers often receive flash calls. This indicates that implementing the module and blocking voice fraud could further boost A2P traffic volumes, leading to higher revenue for both the SMS aggregator and the mobile operator.
Book a free demo to explore what EW SMS Firewall can do for your network
