DBFS: Dragonfly Bayes Fusion System to detect the tampered JPEG image for forensic analysis

Abstract

Due to the advancement of a variety of photo editing and image processing software, image forensics analysis has become an important research topic in recent years. Numbers of research works are presented for image forensic analysis. Accordingly, this paper proposes a method named as Dragonfly Bayes Fusion System (DBFS) by integrating the Naive Bayes (NB) classifier and the Dragonfly optimization for detecting the tampered Joint Photographic Experts Group image. Initially, the input image is applied to the existing six forensic tools separately, and the classified binary map is generated. Then, the proposed DBFS fuses these decisions for generating the optimal decision. Here, the NB classifier creates the model by finding the mean and variance of every feature and this model is given as input to the Dragonfly optimization for optimally generating the probabilistic measures. Finally, the posterior probability of each feature is determined with respect to the tampered class, and the original class and the tampered image block is determined. The performance of the proposed system is evaluated with the existing methods, such as fuzzy theory based classification, rule-based classification, average method, and weighted average method for the evaluation metrics accuracy, False Positive Rate (FPR), and True Positive Rate (TPR). The experimental results show that the proposed system outperforms the existing methods by obtaining the maximum accuracy of 0.9519, minimum FPR of 0.0490, and maximum TPR of 0.8720 when compared to the existing methods.