ICN FILKOM UB

Authored by: Rayhan Egar

Malang, 06 January 2025

In August 2024, ICN Lab kicked off a new project to develop its prototype: an integrated water potability detection system, integrating all four research areas: IoT, AI, web service and cybersecurity. This collaboration marked the beginning of a series of iteration in research and development as part of Tri-Dharma Perguruan Tinggi initiative. The system architecture consists of 5 key components: IoT for water quality monitoring, web service as the system integrator, AI/MLops for water quality classification, all-round cybersecurity measures and monitoring dashboard. The student team compositions are as follows.

IoT Developers: Tiara Calista (Informatics Eng. 2022), Rayhan Egar (Informatics Eng. 2022), Maritza Aliyya (Informatics Eng. 2022)
Backend Developers: Archie Vian (Information Tech. 2022), Gede Infra (Informatics Eng. 2022), Mirza Hilmi (Information Tech. 2022)
Network/Security Engineers: I Gusti Ngurah Ryo (Informatics Eng. 2022), Virgananta Saputra (Informatics Eng. 2022), Ahmad Nabih (Informatics Eng. 2022)
AI Engineers: Firhan Imam Haekal (Informatics Eng. 2021), Fauzan Ghaza (Informatics Eng. 2022), Evan Laksana (Informatics Eng. 2021)

IoT node is constructed using ESP32 as the computing module. pH meter sensor, Total Dissolved Solids (TDS) sensor, and Turbidity sensor. Data from the pH meter is the pH of the water in the form of numbers ranging from 0-14. Data from TDS is a number in units of ppm (parts per million). Electrical conductivity data can be converted from TDS data in units of μS / cm. Data from turbidity is in units of NTU (Nephelometric Turbidity Units). All data from the sensor will be sent every 1 minute to the server using the MQTT (Message Queuing Telemetry Transport) protocol in a secure, encrypted manner. Mosquitto was chosen as the broker for IoT MQTT communication.

The web service is built to prioritize scalability and its ability to handle IoT data streams, MLOps integration as well as data visualization. The web service handles data sent from IoT, feeding them into a ML model to make predictions about the suitability of water for drinking. In addition, the service can handle data requests from dashboards for monitoring purposes using open-source solutions like Grafana with InfluxDB for storage.

The machine learning model was created using only 3 features that match the data from the sensor in accordance with WHO standards in determining the feasibility of water for drinking. XGBoost was selected as the preferred ML model for this purpose due to its performance. After successfully creating the model, the model will be deployed as a web service using MLFlow. The model is created as a service using the Python programming language.

On the aspect of cybersecurity, all measures were taken in all aspects of the water potability detection system: from the IoT node, to the ML model and the web service. The network for this system utilizes a private network. The MQTT and other protocols are enhanced in terms of security by implementing end-to-end encryption from the node, to the broker and to the web service.

Lastly, real-time data and water quality classification results are being visualized in an intuitive dashboard using Grafana, This allows stakeholders to ensure a safe drinking water for everyone, especially in places where public water fountains are present. By developing this system in-house in collaborative manners, ICN Lab aims for an even-better, production version of the system. Stay tuned as we work on our prototype for the next iteration!

Contact and Information

For more information and inquiries about the ICN Lab, please contact us through:
Email: icn@.ub.ac.id
Address: Fakultas Ilmu Komputer, Gedung F Lt. 9.6, Jl. Veteran No. 8, Malang, 65145, Indonesia

ICN Lab Water Potability Detection System Progress Report

Contact and Information