An efficient scheme for secure feature location using data fusion and data mining in internet of things environment

• Published in: Software, practice & experience practice & experience, 2022-03, Vol.52 (3), p.642-657
• Main Authors: N, Balaji, S, Lakshmi, M, Anand, M, Anbarasan, P, Mathiyalagan
• Format: Article
• Language: English
• Subjects: Algorithms; artificial flora optimization; association rule mining; Caesar Cipher‐Rivest; Data mining; Data integration; Data mining; Encryption; Internet of Things; latent semantic indexing; normalized K‐means; Optimization; Security; Software; Source code
• ISSN: 0038-0644; 1097-024X
• DOI: 10.1002/spe.2805

Summary Feature location (FL) is performed to find the relationships between domain concepts and other software artifacts. One major problem in maintaining a software system is to understand how many functional features exist in a system and how these features are implemented. Also, poor security is the prime problem in the FL system. However, the existing recent FL techniques use a textual and dynamic approach, which is not found to be secure, keeping in view the changes in the description of security attacks. To overcome this drawback, this work proposed a novel secure approach for FL utilizing data fusion as well as data mining for the internet of things environment. Firstly, the repeated test cases (TC) are eradicated as of the labeled TC. Next, important attributes are selected using the artificial flora optimization algorithm from the removed labeled TC. Then, association rule mining is performed to ascertain closed attributes. Subsequently, encrypt the closed attributes utilizing Caesar Cipher‐Rivest, Shamirs, as well as Adelman algorithm. After that, the score value of the closed attributes counts was found utilizing entropy calculation. Finally, the score value is given as input to the normalized‐K‐Means (N‐[K‐Means]) algorithm, where the score value is normalized utilizing min‐max normalization and then grouped utilizing K‐Means algorithm (KMA). It proffers better results for FL in the source code. The proposed N‐(K‐Means) performance is found better in comparison to the KMA and latent semantic indexing methods. The proposed system proffered better FL results in comparison to the other prevailing methods.