Continuous Authentication Using Keystroke Dynamics

1

• Continuous Authentication Using Keystroke Dynamics

by Bruno Guerreiro
2
BIOMETRICS

• Biometrics is the science of people
  identification.
• This technology verifies the identity of an
  individual based on his/her distinct
  characteristics.
• There are two biometric authentication methods
• Physiological authentication
• Behavioral authentication

3
KEYSTROKE DYNAMICS

• Biometric behavioral method
• Checks the identity of an individual
• Monitored keyboard
• Typing characteristics are recorded in a form of
  timings

4
CHARACTERISTICS (TIMINGS)

• KEY DOWN 5,69105626553095 I
• KEY UP 5,805169372085 I
• KEY DOWN 6,4678342689764 N
• KEY UP 8,5487345268945 N

5
AUTHENTICATION

• When a person gives us something that is supposed
  to identify him/her and we check if it is true or
  not.
• There are 2 main authentication techniques
• Static Just at once at the start and it is
  valid till the end of the session
• Continuous Made continuously during the session

6
RESEARCH QUESTIONS

• Is it possible to continuously authenticate a
  user during normal typing with the same results
  as static authentication ?
• Which are the advantages and limitations?

7
EXPERIMENTS

• Experiment 1 (Enrollment phase) Data collection
  while typing a predefined text
• Experiment 2 (Analysis) Data collection while
  the user uses his computer free for some days to
  compare to the previous template afterwards

8
EXPERIMENT 1

• Features
• Single keys
• E, I, N, O, S, T, L
• Space, Backspace, Comma, Backspace
• Double keys
• MA, IN, IC, IS, BY, OF, TE, BE, NG, AN, AT, CO,
  OR, LL, TT, SpaceM
• Timings stored in one text file

9
EXPERIMENT 1
10
TEMPLATE (1)

• General
• Personalized

11
EXPERIMENT 2

• Ran in background
• Capture just some features
• Less controlled environment
• During a few days

• Timings stored in 2 text files
• singleKeys.txt
• E-4,6469324,742566dayX
• doubleKeys.txt
• AN4,9445495,02426855,03962545,1030760dayX

12
DISTANCE METRIC

• Simple
• Gives the distance between two typing patterns
• Normal or not

13
DECISION RULE 1

• Single key
• Double key

14
ANALYSIS (1)

• Template VS experiment 2 data

0,061071127 0,089815773
New templates needed !
15
ANALYSIS (3)

• With decision 0/1

• 7825 collected features
• 1276 mistakes
• 24 missing features in the template
• 0 below

16
ANALYSIS (4)

• With decision 0/1

• 10212 collected features
• 1712 mistakes
• 9 missing features in the template
• 3 below

17
DECISION RULE 2

• Single key
• Double key

18
ANALYSIS (5)

• With new decision rule

• 10212 collected features
• 1712 mistakes
• 9 missing features in the template
• 0 below

19
RESULTS

• With decision 0/1
• FAR 22
• FRR 0
• With decision Distance Threshold
• FAR 16
• FRR 0

20
COST FUNCTION

• Initial condition
• When mistake
• When correct
• Every second

YES or NO
21
EXPERIMENTS SUMMARY

• Hard to find participants
• Some participants quitted
• Can I use CtrlC CtrlV ?
• Some problems with the Splitter line
• SPLITERUsername09/05/2007 103455
• Big gap between single and double key number

22
CONCLUSION

• Results shown a good potential.
• Duration timings shown to be more consistent than
  latency timings.
• Highly sensitive to the used language to type.
• Typing characteristics depend on the environment.

23
FUTURE WORK

• More participants and longer period of time
• Better experiments setup
• Personalized choice of features
• Improve distance metric / decision rule
• Include other feature types, time of the day
• Add key pressure
• Include mouse movements