Optimal Deep Belief Network Enabled Vulnerability Detection on Smart Environment

Cyborgs invoke visions of super-humans intertwined with innovative technologies able to surpass the restrictions of the human body. Conventional vulnerability detection models sustain maximum false positive rates and depend upon manual participation. Machine learning (ML) and Artificial intelligence (AI) technologies are exploited in several real-time applications, like vulnerability, malware, and software function detection, for high-quality feature representation learning. In this aspect, this study introduces a hyperparameter-optimized deep belief network-enabled vulnerability and classification (HOSDBN-VC) technique on cyborgs. The presented HOSDBN-VC model aims to detect and classify the existence of vulnerabilities. The presented HOSDBN-VC model involves a Z-score normalization approach to transform the input data into a useful format. In addition, the Hypercube Optimization Search Algorithm-based feature selection (HOS-FS) method is employed for selecting feature subsets. Moreover, a flower pollination algorithm (FPA) with a deep belief network (DBN) paradigm is applied for vulnerability and classification. The metaheuristics-based FPA is exploited to choose the hyperparameters related to the DBN paradigm appropriately. A wide-ranging experiment has been conducted to investigate the results of the HOSDBN-VD model under two databases, FFmpeg and LibPNG. The experimental outcomes implied the betterment of the HOSDBN-VD algorithm over recent approaches.