A blind steganalysis-based predictive analytics of numeric image descriptors for digital forensics with Random Forest & SqueezeNet
Wasiu Akanji, Ogobuchi Daniel Okey, Saheed Adelanwa, Oluwafunsho Odesanya, Taiwo Olapeju Olaleye, Amusu Mary, Akinfolarin Akinrinlola, Abiodun Oladejo · 2022
Image steganalysis have been a prominent study in digital forensics and the data science use case of artificial intelligence has been widely adopted in conceptual frameworks. In existing studies, deep learners gain prominence for intrusion detection systems while other dissimilar modules are used for feature extraction. Hence, this study rather employs deep learners as image embedding networks aimed at feature extraction for a predictive analytics of image steganalysis. The extracted numeric image descriptors trains three learner algorithms for pattern recognition using a 10 fold cross-validation system. Experimental result indicates the ensemble of Random forest algorithm and SqueezeNet image embedder as the best for steganalysis in digital forensics while the size of the training set turns out to be insignificant for the supervised machine learning study.