Table of Contents

Among the research topics of the Security Group, we are investigating the use of behavioral biometrics for user authentication on new-generation devices. Existing authentication schemes, e.g., PIN/passwords, face and iris recognition, etc., have shown to be less secure and less usable and users are reluctant to enable them on their devices. Consequently, the research has been diverted to find usable alternatives.

Human biological data, due to its permanence and uniqueness, can be used as a means of identification, authentication and access control. The use of biological data for the purpose of identity management is termed as biometric recognition or simply biometrics. Physical (based on the physical characteristics) and behavioral (based on behaviors) biometrics are the most popular types. Physical characteristics include fingerprints, hand geometry, iris or retina scans, etc., and behavioral characteristics include keystrokes, gait, signature, voice, etc. Other biometrics use chemical features (based on events that happen in a person's body, measured by e.g., odor or temperature) and cognitive features (based on brain responses to specific stimuli, e.g. odor or sound).

Biometrics has got all the potential to completely replace PINs and passwords because they can be stolen, forgotten, and shared. Biometric authentication has been studied for a long time. Large-scale commercial deployments already exist, such as the fingerprint sensors on laptops and smartphones. However, these deployments are based on physical biometrics, which essentially requires explicit user action, hence result in annoying users. As a result, most of the research about transparent, implicit and unobservable authentication for smartphone's security and access control is based on behavioral biometrics.

Behavioral biometrics offer many advantages over physiological traits. One of the main advantages is that the behavioral patterns can be collected transparently or sometimes even without the user's knowledge. More importantly, data collection does not require any special dedicated hardware. However, most of the behaviors are not unique enough to provide accurate user identification but have shown promising results in user verification. Various behavior-based authentication solutions have been tested and evaluated but are yet to be deployed at large scale. One reason is that the performance of many of these schemes is not yet at the same level as physical biometrics. Another reason is that not much attention has been paid to the performance of biometrics under differing or difficult circumstances. For example, gait authentication is typically evaluated by having subjects walk along flat surfaces of corridors in buildings.

We have been developing behavioral-biometric-based solutions that authenticate the users with either minimal or no cooperation from the users. We are designing, prototyping and testing the proposed authentication based on our identified behaviors, i.e., how a person holds her phone, moves her phone, or interacts with its touchscreen. We are also performing the comparative evaluation, based on accuracy, performance, and usability, with the state-of-the-art behavioral-biometric-based solutions. All of our solutions exploit the existing hardware (avoiding additional hardware requirements) and hence can be implemented on most of the smartphones available in the market today.

Themes

Within the mainstream project we covered a number of themes.

People

Team members

we are reachable via email @ name.surname@unitn.it

Projects

This activity was supported by a number of projects

Datasets

Instructions to access to our datasets

  1. Send the signed license agreement by email as per the instructions mentioned below.
  2. Send an email to security(AT)disi(DOT)unitn(DOT)it, as follows:
  3. Subject: [DATABASE download: DISI Security Lab Datasets]
  4. Body: Your name, e-mail, telephone number, organization, postal mail, the purpose for which you will use the database, time and date at which you sent the email with the signed license agreement.
  5. Once the email (preferred you to use your company/institute/university email id) along with the license agreement has been received, we will send the requested database.

Datasets Available

  1. 41 users micro hand-movements dataset (21263 observations) collected in the wild using smartphone.
  2. 95 users touch and hold-movements data (smartphones)
  3. 86 users swipe, pickup, and voice data (smartphones)
  4. 40 users hold and digital signature data (tablets)
  5. 40 users hand-movements data for Smarthandle

Solutions

Talks and Tutorials

Publications

2020

2019

2018:

2017:

2016:

2015: