Knowledge and Information Systems Взаимодействие человека и компьютера
Efficient detection and encryption of sensitive data in cloud systems using SHRSO-K3NN framework
Аннотация
In a cloud environment, cloud service users and providers face challenges associated with security threats, data loss, availability, unauthorized access, and data confidentiality. However, the performance of various existing techniques and their security solutions are inefficient in providing cyber security or remain resource-intensive. To overcome these problems, a novel detection and classification model is proposed in this paper called Scaled Hippo Rhinopithecus Swarm Optimized K-three layer Nearest Neighbor (SHRSO-K3NN) model. The proposed model utilizes standard datasets such as UNSW-NB15, CIC-IDS-2017, CSE-CIC-IDS2018, and DARPA datasets, which are preprocessed by applying data cleaning, handling missing values, data normalization, and dimensionality reduction using Scaled Principal Component Analysis. A Hippo Rhinopithecus Swarm Optimization algorithm is employed for optimal feature selection by integrating the Hippopotamus Optimization’s global exploration capability and Rhinopithecus Swarm Optimization’s local exploitation capability. Then, the K-three-layer Nearest Neighbor classifier is utilized for the classification of network traffic into sensitive data (legitimate user data requiring confidentiality) and non-sensitive data (malicious or attack traffic) by applying the majority voting rule. To minimize encryption, storage, and computational overhead, the non-sensitive (attack) traffic is discarded after detection and only sensitive legitimate user data are encrypted using Efficient Time-oriented Data Encryption scheme and stored in the cloud. Experimental evaluations demonstrate that the SHRSO-K3NN model attained classification accuracy of 98.8% and an average security rate of 98.4% across the four benchmark datasets, outperforming existing models. By protecting sensitive data and optimizing resource utilization selectively, the SHRSO-K3NN model addresses limitations faced by cloud service providers and users. Although the approach is applicable for real-time intrusion detection, the current evaluation is limited to offline standard datasets.