Professor Rashmi Agrawal
Professor, Manavrachna International Institute of Research and Studies, Faridabad, India
Satya Murthy Sasubilli
Solution Architect, Huntington National Bank
Srinivasa Rao Swarna
Program Manager/Senior Data Architect, Tata Consultancy Services
1
Biometric Identification Using Deep Learning for Advance Cloud Security
Navani Siroya1* and Manju Mandot2
1 MDS University Ajmer, India
2 Computer Science, JRN Rajasthan Vidyapeeth University, Udaipur, India
* Corresponding author: [email protected]
Abstract
A few decades ago, biometric identification was a staple technology of highly advanced security systems in movies, but today, it exists all around us. Biometric technologies have the potential to revolutionize approaches to identity verification worldwide.
This chapter discusses the prevailing Biometric modalities, their classification, and their working. It goes on to discuss the various approaches used for Facial Biometric Identification such as feature selection, extraction, face marking, and the Nearest Neighbor Approach.
Here, we propose a system that compares an input image with that of the database in order to detect the presence of any similarities. Moreover, we use fiducially point analysis to extract facial landmarks and compare them with the database using data mining and use the Nearest Neighbor Approach for identifying similar images.
The chapter ends with deliberations on the future extent of Biometric technologies and the need to put in ample safeguards for data protection and privacy.
Keywords: Biometric, feature extraction, facial recognition, nearest neighbor approach
1.1 Introduction
Biometric authentication is a security process that relies on the unique biological characteristics of a person in order to affirm their identity. Biometric verification frameworks compare biometric data with existing original datasets that are stored. Examples of biometric characteristics include iris, palm print, retina, fingerprint, face, and voice signature. In recent years, deep learning-based models have helped accomplish best in class results in machine vision, audio recognition, and natural language processing tasks. These models appear to be a characteristic fit for dealing with the everexpanding size of biometric acknowledgment issues, from phone verification to air terminal security frameworks. Thus, application of machine learning techniques to biometric security arrangements has become a trend [1].
Classification of Biometric Data:
• Behavioral Biometrics: gestures, vocal recognition, handwritten texts, walking patterns, etc.
• Physical Biometrics: fingerprints, iris, vein, facial recognition, DNA, etc.
Data science consultants can use machine learning’s capacity to mine, look, and examine huge datasets for improving the execution of security frameworks and their reliability.
In light of its exceptional capacity to recognize people, biometric innovation has quickly become a way to help forestall shams and discovered its place in today’s standard advancements. Consequently, it turns out to be more reliable than the customary validation frameworks that utilize passwords and documents for verification shown in Figure 1.1.
Figure 1.1 Biometric modalities [2].
Physical modalities like fingerprints, voice, faces, veins, iris, hand geometry, and tongue print are unique and provide robust advancements in the field of cyber security [2]. They are useful compared to names, ID numbers, passwords, etc. because they are extraordinary, hard to reproduce, and are more significantly and genuinely bound to the individual.
A computing model which gives on-demand services like information stockpiling, computer power, and infrastructure to associations in the IT industry is termed to be “cloud computing” [3]. Despite the fact that cloud offers a ton of advantages, it slacks in giving security which is an issue for most clients. Cloud clients are hesitant to put classified information up because of looming threats to security.
1.2 Techniques of Biometric Identification
1.2.1 Fingerprint Identification
An automated technique for recognizing or affirming the identity of an individual dependent on the examination of two fingerprints is termed as Fingerprint Recognition. Human fingerprints are not easy to manipulate and are nearly unique and durable over a person’s lifetime. They are unique, permanent, easy to acquire, and are a universally acceptable mode of identification [4].
Human fingerprints are difficult to control but remain sturdy over the life of an individual, making them suitable as long stretch markers of human character.
WORKING OF DIFFERENT TYPES OF FINGERPRINT READERS
1 1. Optical Readers’ sensors work using a 2D image of the fingerprint. Algorithms can be utilized to discover novel patterns of lines and edges spread across lighter and hazier zones of the picture
2 2. Capacitive Readers use electrical signals to form the image of fingerprints. As the charges differ in the air gap between the ridges and lines in the finger set over the capacitive plate, it causes a difference in the fingerprint patterns.
3 3. Ultrasound Readers use high frequency sound waves to infiltrate the external layer of the skin which is used to capture a 3D depiction of the fingerprint. It involves the use of ultrasonic pulses using ultrasonic transmitters and receivers.
4 4. Thermal Readers sense the temperature difference between fingerprint valleys and ridges on making a contact. Higher power consumption and a performance reliant on the surrounding temperature are impediments for these readers.
1.2.2 Iris Recognition
The iris is a shaded, flimsy, roundabout structure of the eye which controls light entering the retina by regulating the diameter and size of the pupil. It doesn’t change its appearance over a range of an individual’s lifetime except if harmed by external components [5]. Hereditarily
