Biometrics for Forensic Identification in Web Applications and Social Platforms Using Deep Learning

Introduction

As the popularity of the Internet has increased, there is a growing need to identify people who access sensitive content on the Web. Most of the web applications ensure the security of their services by using passwords (Duic, Cvrtila & Ivanjko, 2017). However, it is easy to intercept passwords of people. Moreover, intrusion into personal accounts has become common, thus leading to cybercrimes. Therefore, the concept of using biometric authentication is gaining popularity in web systems because biometric authentication is reliable and is an important technique (Saini & Kumar Kapoor, 2016) to identify criminals by using physical traits (such as fingerprints, facial patterns and iris scans) and behavioural traits (such as signature and voice).

The choice of a biometric trait depends on the following factors:

• •

  Universality: Every individual should possess the biometric trait.

• •

  Uniqueness: Each should have a unique trait.

• •

  Permanence: The trait should not change over time.

• •

  Performance: Given the limitations of an individual, the traits should be recognized with high accuracy.

Thus, a biometric system acquires data from an individual, extracts features from the available data, and compares these features with a set of templates that are already input in the system database (Mayron 2015). Biometric systems can be divided into two categories based on the mode of application:

• •

  Identification: A biometric system searches among all the existing templates stored in its database and compares the features to prove a person’s identity. Thus, the system performs a 1:N comparison, and this process is slower than verification because the template is matched with all other templates in the database. The response validates the identity of the individual.

• •

  Verification: This is a one-to-one comparison technique in which an individual’s information is compared with his template in the database. For example, an individual’s fingerprint is captured digitally through an application. The system then compares the fingerprints to verify if the fingerprint belongs to this individual.

Figure 1 displays the working of a biometric matcher in a web application. In a biometric matcher, features are either extracted and identified or verified through templates that are already stored in the database to allow access to a user.

Figure 1.

Biometric authentication

Biometric technologies are sufficient to perform forensic analysis and detect crimes in web applications because they overcome the limitations of human abilities and increase the effectiveness of an investigation. By applying various techniques, a considerable amount of data can be easily analysed on a scientific basis in crime investigations, which would not have been possible otherwise. Moreover, the analysis results obtained can be easily visualised and documented.

The performance of a biometric system can be analysed using various metrics such as false match rate or false accept rate, which finds the percentage of unauthorised access in a system, and false rejection rate or false non-match rate, which analyses the number of authorised inputs falsely rejected by a system. Thus, there is a need to find a balance point at which the false match rate is equal to the false non-match rate, also known as equal error rate (Marasco & Ross 2014). The representation of the rates is presented in Figure 2.

Figure 2.

Performance metrics in a biometric system