ECML-PKDD17 OUTLINE 1 Problem 2 Methodology 3 Experiments 4 - - PowerPoint PPT Presentation

▶
ecml pkdd17 outline
SMART_READER_LITE
LIVE PREVIEW

ECML-PKDD17 OUTLINE 1 Problem 2 Methodology 3 Experiments 4 - - PowerPoint PPT Presentation

Mar 01, 2024 35 likes •268 views

Sequential Keystroke Behavioral Biometrics for Mobile User Identification via Multi-view Deep Learning Lichao Sun 1 , Yuqi Wang 2 , Bokai Cao 1 , Philip S. Yu 1,3 , Witawas Srisa-an 4 , and Alex D Leow 1 1 University of Illinois at Chicago, 2 Hong

slide-1
SLIDE 1

Sequential Keystroke Behavioral Biometrics for Mobile User Identification via Multi-view Deep Learning

Lichao Sun1, Yuqi Wang2, Bokai Cao1, Philip S. Yu1,3, Witawas Srisa-an4, and Alex D Leow1

1University of Illinois at Chicago, 2Hong Kong Polytechnic University, 3Tsinghua University, 4University of Nebraska Lincoln

ECML-PKDD17

slide-2
SLIDE 2

2

Problem

1

Methodology

2

Experiments

3

Conclusions

4

OUTLINE

slide-3
SLIDE 3

3

Backgrounds

Problem Statement

Our task

slide-4
SLIDE 4

4

Authorization Identification

Problem Statement

System System

Owner or Not ? Sam/John/Bob is using Our task

slide-5
SLIDE 5

5

Problem Statement

  • Stolen Phone
  • Using the Phone

without Owner’s Permission

  • Recommendation
  • Auto Personal

Setting Changing

Authorization vs Identification

slide-6
SLIDE 6

6

Traditional Identification

Problem Statement

Weakness:

  • Not Convenient
  • Security Issues

Account

+

Passward

slide-7
SLIDE 7

7

Major Challenges……

  • 3. Data Privacy
  • 1. High Identification

Performance

  • 2. Data Features

Design

Problem Statement

slide-8
SLIDE 8

8

Problem Statement

Feature Design & Selection Authorization vs Identification

Accelerometer Gyroscope Magnetometer Raw touch event Tap gesture Scale gesture Scroll gesture Fling gestur Key press on virtual keyboard … Accelerometer Tap gesture Key press on virtual keyboard

slide-9
SLIDE 9

9

Problem Statement

Solution I : Single-view Traditional Learning

Multi-class Traditional Learning: Support Vector Machine Decision Tree Random Forest Logistic Regression

slide-10
SLIDE 10

10

Solution II : Single-view Deep Learning

Problem Statement

slide-11
SLIDE 11

11

Problem Statement

Solution III : Multi-view Deep Learning

slide-12
SLIDE 12

12

OUTLINE

Problem

1

Methodology

2

Experiments

3

Conclusions

4

slide-13
SLIDE 13

13

Multi-view Multi-class Deep Learning

Step I : Auto-encoder for Each View

A GRU is formulated:

Inputs of Each View Representation of Each View

slide-14
SLIDE 14

14

Multi-view Multi-class Deep Learning

Step II : Concatenate Representations

  • f Each View
slide-15
SLIDE 15

15

Multi-view Multi-class Deep Learning

Step III : Softmax & Output

Softmax Function Multi-class Output:

[0,0,0,1,0,…,0]: single one value Result: Index of 1 is the multi-class

slide-16
SLIDE 16

16

Problem

1

Methodology

2

Experiments

3

Conclusions

4

OUTLINE

slide-17
SLIDE 17

17

Experiments

Datasets

  • 40 Volunteers
  • 26 of 40 Active Users (17 females and 9 males)
  • 8 Weeks
  • 11 – 63 years old
  • Minimum: 29 Maximum: 4702 Times Usage of the

Phone

slide-18
SLIDE 18

18

Pattern Analysis

Experiments

slide-19
SLIDE 19

19

Experiments

Results

slide-20
SLIDE 20

20

slide-21
SLIDE 21

21

Problem

1

Methodology

2

Experiments

3

Conclusions

4

OUTLINE

slide-22
SLIDE 22

22

We have shown that DEEPSERVICE can be used effectively to identify multiple users. Even though we only use the accelerometer in this work, our results show that more views of dataset can improve the identification performance.

Conclusions

– DeepService is the first system for mobile user identification – DeepService is the best model for multi-view multi-class dataset – DeepService takes about 0.657 ms per session which shows its feasibility of real-world usage

slide-23
SLIDE 23

Thank you！