Issues in Information Systems Volume 20, Issue 3, pp. 107-116, 2019 107 EXAMINING A DEEP LEARNING NETWORK SYSTEM FOR IMAGE IDENTIFICATION AND CLASSIFICATION FOR PREVENTING UNAUTHORIZED ACCESS FOR A SMART HOME SECURITY SYSTEM Beloved Egbedion, Georgia Southern University, be02054@georgiasouthern.edu Hayden Wimmer, Georgia Southern University, hwimmer@georgiasouthern.edu Carl M. Rebman Jr., University of San Diego, carlr@sandiego.edu Loreen M. Powell, Bloomsburg University, lpowell@bloomu.edu ABSTRACT There are many different smart home surveillance and control systems, which will need some type of visual identification and classification system. Past models of Deep Learning have had great success in visual identification and image classification particularly in the healthcare and security industries. This study reviews past architecture and applications of Deep Learning and Convolutional Neural Networks. This paper then presents the creation, process, testing, and results of a CNN model with the end objective of identifying images for determination of access rights. Evaluation outcomes show that after 50 forward and backward dataset training passes the deep learning network achieved an identification accuracy of 96.7% and a 98.0% probability of proper classification of access authorization. The results suggest that deep learning models could be successful in strengthening smart home security systems. Keywords: Deep Learning, Convolutional Neural Networks, Image Classification, Smart Home, Security INTRODUCTION Americas are increasingly become more concerned about home security For example, in a study by asecurelife.com 63% percent of Americans between the ages of 25 and 34 are more concerned about a home invasion than financial security and 53% more women feared home invasion over identity theft. Advances in smart home technologies are helping to address these fears and concerns. According to the International Energy Agency (IEA 2013) the smart home market is projected to be $26 billion in 2019 (up from $40 million in 2012). This forecast is echoed by Statista, which also predicts the smart home market to be $27 billion and $44.7 billion in 2023 (Statista 2018). There are also many risks in smart home security systems (Fernandes et al 2016). For example, smart home security systems have to be prepared to address cyberattacks that would allow criminals to monitor, steal personal information, or lock the person out of their home. According to Cate Lawrence of readwrite.com, there are three main layers of cyber defense (Lawrence, 2017). The first or perimeter layer is about preventing unauthorized access of one’s devices. The second layer is about intrusion detection and prevention. The third is anomaly detection and behavioral analysis. An area that has been used to assist with these three layers of cyber defense is Deep learning. Deep learning, a popular machine learning technique being used to establish state of the art solution for problems that ranges from Natural Language Processing (NLP) to object classification. Deep learning has recently seen rapid development and received significant attention due to its state-of-the-art performance on previously thought hard problems (Hohman et al 2017). In the last five years, the deep learning technique has achieved superior results over other classical machine learning methods in image classification, detection and segmentation in several applications (Ghazi et al 2017; Hinton et al 2012; Le et al 2012; Shu et al 2017; Yu et al 2017). Deep learning and neural networks are terms that many people use interchangeably and yet there is a difference between the two. They are both are techniques that exist under the global umbrella of artificial intelligence along with machine learning. According to Skymind (2019), it is best to consider all the techniques “to be like a set of Russian dolls nested within each other, beginning with the smallest and working out.” In other words, Deep learning is a subset of machine learning, and machine learning is a subset of AI (Dhande 2018).” Deep learning is also referred to https://doi.org/10.48009/3_iis_2019_107-116