FreeHand sketch-based authenticated security system – A comparative study on convolutional neural network, Levenshtein distance and sequence matcher procedure

2019 ◽

Vol 9 (2) ◽

pp. 3405-3411

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Gaussian Mixture ◽

Adaptive Threshold ◽

Input Image ◽

Security System ◽

Security Strategy ◽

System Database ◽

Freehand Sketch ◽

Authentication Security

An Authenticated Security System is a highly desired feature. In this paper, a FreeHand Sketch-based Authentication Security strategy is proposed for authentication purposes by allowing a user to choose one label from a collection of different labels and asking him to sketch the corresponding image for the selected label for registration to avoid mischievous registration and the sketched image gets preprocessed using adaptive threshold with Gaussian mixture and then predicted with a trained Convolutional Neural Network(CNN) data model to generate the necessary image label. The produced image label will compare with selected image label. If both are same then the details will store in the system database. The user gets login with his/her authorized details with sketch based image password. The image password gets preprocessed using adaptive threshold with Gaussian mixture and then predicted with a trained CNN model to produce the image name. The produced image name will compare with the system database for authentication. The methodology is tested with some sample input image passwords and the performance calculation is carried out using metrics like Recall and Precision. The proposed work exhibits the accuracy of approximately 85% by ensuring the authentication for the user security.

Download Full-text

DEVELOP A COMPREHENSIVE SECURITY SYSTEM

Interexpo GEO-Siberia ◽

10.33764/2618-981x-2020-7-1-140-148 ◽

2020 ◽

Vol 7 (1) ◽

pp. 140-148

Author(s):

Anton А. Yushmanov

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Image Matching ◽

Overlay Networks ◽

Security System ◽

Recognition Mechanism ◽

System A ◽

Reference Images ◽

Smoke Detectors ◽

Characteristic Features

The article describes the process of developing an integrated security system based on a single-board NVIDIA Jetson Nano computer. The main objective of the system is to provide profiled access to the room for trusted users. As a result of the work, a face recognition mechanism based on the convolutional neural network was designed and implemented. Face detection is implemented using histograms of directional gradients. Then, thе face is centered and processed by the trained convolutional neural network, out the characteristic features of the image supplied to the image classifier, which gives the percentage of image matching with the reference images. In the process of implementing the system, a framework was developed that allows implementing a complex security system, regardless of hardware components. To ensure the protection of the room, smoke detectors, opened doors and windows were added to the system. The scientific novelty of tht work is the encapsulation of all transmitted system traffic to overlay networks to ensure confidentiality.

Download Full-text