A Semantic Ontology-Based Model by Ensemble Learning For Secure Attribute-Based Encryption in Fog-Enabled Smart Homes

Fog computing empowers resource-limited applications by bringing cloud computing close to the network periphery, effectively limiting latency, improving efficiency, and assuring better resource management. Security challenges, however, restrict widespread adoption in practice, necessitating cryptographic mechanisms with low computational overheads. One well-known approach for data sharing in a secure way is Ciphertext-Policy Attribute-Based Encryption (CP-ABE), which provides a way of access control based on attributes. However, the high execution time and storage requirements of CP-ABE, due to the diversity of attributes in secret keys and access structures, limit its practicality in resource-constrained environments. In response to these problems, this paper presents a hybrid semantic model consisting of an outsourced CP-ABE with attribute revocation and an optimized AES algorithm based on ensemble learning. The model employs classifiers such as GMDH, KNN, and SVM to identify attributes relevant to CP-ABE. Additionally, the Dragonfly optimization algorithm and ontology-based semantic techniques enhance the efficiency of feature selection. With experimental analysis on five smart building datasets, the prediction performance of this model outperforms existing methods. The times of encryption, decryption, and attribute revocation decreased to 2.99 ms, 2.86 ms, and 18.6 ms. The growth of storage for secret keys and access structures was reduced to 13.56 KB and 10.4 KB, which made its use more efficient and secure. Overall, the results indicate that the model improves data security and minimizes computational overhead, leading to a more feasible implementation of CP-ABE for fog computing scenarios.