DNS Tunneling Detection with New Patterns Emerging For Intelligent Agriculture: A Forest-Based Classifier with An Unknown Option

In Intelligent agriculture, securing data transmission is critical due to the vast amount of data generated by sensors. DNS tunneling poses a threat by exfiltrating data through DNS queries, a challenge for traditional classifiers, especially with 'unknown queries' not seen during training. This paper frames DNS tunneling detection as a machine learning problem involving Classification with New Patterns Emerging (CNPE). We introduce a forest-based classifier that identifies unknown patterns as a new class and accurately classifies known samples. Our model is efficient in both computation and memory. Experiments on live network data and public datasets validate its effectiveness, demonstrating its potential to enhance data security and system reliability in Intelligent agriculture.