Forest fire detection model developed using trust and machine learning

A study published in Scientific Reports by Tayyab Khan, Karan Singh, Bhoopesh Singh Bhati, Khaleel Ahmad, Amal Al-Rasheed, Masresha Getahun and Ben Othman Soufiene has outlined the development of a trust-based, machine learning-enabled system for early forest fire detection.

The research introduces a Universal Trust Model (UTM) designed to evaluate the reliability of wireless sensor networks and improve the accuracy of fire detection by validating environmental data before transmission.

The study reports that the system was tested on a dataset of 7200 samples, which included varied fire and weather conditions across multiple locations.

Intelligent sensors assess environmental changes in real time

The study explains that forest areas are segmented into clusters, with each cluster equipped with intelligent sensor nodes (ISNs) that measure variables such as temperature, humidity, light intensity and gas concentrations.

Each sensor is uniquely identified and assessed for trustworthiness using metrics that include energy levels, communication consistency and data reliability.

The authors stated that ISNs deemed unreliable or malicious are removed from the network, allowing only trusted data to influence detection outcomes.

Trust ratings improve system resilience and detection speed

The Universal Trust Model functions in two phases. At the sensor level, it calculates trust values, and at the base station, it applies machine learning to verify fire-related patterns in the incoming data.

According to the study, this structure improves detection accuracy and reduces false alarms, especially under challenging environmental conditions such as low visibility or extreme weather.

The authors noted that the model enhances decision-making by combining continuous sensor validation with algorithmic analysis.

Experimental data supports improved performance

The model was tested in diverse conditions across locations such as Sanjay Van and Jawaharlal Nehru University campus in Delhi.

Researchers gathered data during daytime and nighttime, across dry and humid environments, and simulated both high- and low-intensity fire events.

According to the findings, the proposed system achieved an accuracy rate of 98.5 percent in identifying fire presence, with a reduced rate of false positives and negatives compared to earlier models.

Machine learning and trust model improve data efficiency

The authors reported that once environmental indicators exceed threshold levels, the base station’s machine learning algorithm verifies the likelihood of fire and, if confirmed, dispatches alerts to relevant authorities.

This process is designed to reduce alert time by using recent and validated data for decision-making.

By filtering unreliable sensors before reaching the central system, the model aims to ensure more consistent and responsive fire detection.

Energy efficiency and coverage considered in deployment

The research details how sensor deployment was optimised to maintain wide area coverage while conserving battery life, aided by solar panels and rechargeable batteries.

Sensor distances were calculated to balance energy usage with detection reliability, and the trust model factored in energy levels to avoid misclassifying low-power sensors.

This architecture supports longer network lifespan and more accurate detection, especially in resource-constrained environments.

Limitations identified in real-world implementation

The authors acknowledged that sparse sensor placement may cause detection gaps, while harsh weather could distort readings.

They also noted the need for retraining machine learning models as new fire scenarios emerge, and pointed to the potential risk of network congestion in large-scale deployments.

Plans for future development include adaptive clustering, energy harvesting and the use of federated learning to update models in real time.

Trust-based model and machine learning improve forest fire detection system: Summary

A study published in Scientific Reports outlined a machine learning and trust-based system for early forest fire detection.

The Universal Trust Model (UTM) uses intelligent sensor networks to monitor environmental conditions.

Each sensor is assigned a trust score based on communication, energy and data accuracy.

Faulty or malicious sensors are excluded from the network.

Sensor data includes temperature, humidity, CO levels and other fire-related indicators.

Data is transmitted to a base station only if verified as reliable.

A machine learning model analyses the data and triggers alerts if fire indicators are confirmed.

The model was tested with 7200 samples collected under varied weather and fire conditions.

It achieved 98.5 percent accuracy in fire detection with low false alarm rates.

Sensor placement was optimised for both energy efficiency and wide coverage.

Limitations include coverage gaps, model retraining needs and risks from coordinated attacks.

Future improvements will focus on energy harvesting, real-time learning and network scalability.