Abstract: Data mining has a wide range of uses in a variety of industries, including banking, medicine, scientific research,
and government agencies. One of the most common tasks in data mining applications is classification. Many theoretical and
functional solutions to the classification problem have been suggested under various security models over the last decade as a
result of the growth of various privacy issues. With the recent rise in popularity of cloud computing, users can now outsource
their results, in encrypted form, as well as data mining tasks to the cloud. Current privacy-preserving classification methods
aren't applicable since the data in the cloud is encrypted. The classification issue over encrypted data is the subject of this
paper. We propose a stable k-NN classifier for encrypted data in the cloud, in particular. The proposed protocol safeguards
data confidentiality, preserves the privacy of a user's query, and conceals data access patterns. To our knowledge, this is the
first time a stable k-NN classifier has been developed over encrypted data using the semi-honest model. We also use a realworld
dataset to empirically test the efficiency of our proposed protocol with various parameter settings. The K-nearest
neighbor (KNN) classification method is widely used in data mining techniques. It is widely used in a variety of fields due to
its ease of implementation, clarity of theory, and excellent classification efficiency. When training samples are distributed
unevenly or the sample number of each class is very different, however, KNN can increase classification error rate. As a
result, this paper adopts an improved KNN classification algorithm and applies it to object-oriented classification of highresolution
remote sensing images, building on the concept of clipping-KNN. Image artefacts are first collected as sample
points by image segmentation. Second, the original KNN, clipping-KNN, and improved KNN are all implemented and used to
classify the sample points. Finally, the effects of the classification are contrasted. The improved KNN algorithm can achieve
higher precision in the classification of high-resolution remote sensing images in the same training and testing sets,
according to the experiment.
Keywords: K-Nearest Neighbor(KNN);KNN classification; object-oriented; segmentation; Data mining; privacy-preserving data
mining (PPDM) ; stable multi-party computation (SMC) ; Machine Learning (ML).