1 / 28

Brain W ave B ased A uthentication

Brain W ave B ased A uthentication. Kennet Fladby 2008. Outline. 1. Introduction 2. Research questions 3. Experimental work 4. Results 5. Conclusion 6. Further work. 1-1. Brain waves. The brain contains about 100 billion neurons.

Download Presentation

Brain W ave B ased A uthentication

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. BrainWaveBasedAuthentication Kennet Fladby 2008

  2. Outline 1. Introduction 2. Research questions 3. Experimentalwork 4. Results 5. Conclusion 6. Furtherwork

  3. 1-1. Brainwaves • The braincontainsabout 100 billion neurons. • Neurons generates and leads electrical signals. • The sum oftheseelectrical signals generates an electricfield. • Fluctuations in theelectricfieldcan be measured. • Electroencephalographic (EEG)

  4. 1-2. 10-20 System

  5. 1-3. EEG signal: 20 seconds, 128Hz

  6. 2. Research questions • Is it possible to authenticate by means of brain waves with only one EEG sensor? • What feature should be extracted from the signals? • Do we have to authenticate based on a person’s thoughts or can we use the brain waves as a biometric directly? • Will a distance metric approach work? • What is the best FMR and FNMR we can achieve?

  7. 3-1. Tasks

  8. 3-2. Setup • 10 participants • 3 sessions, 3 recordingsofeachtask per session • Eachrecording lasts 20 seconds (2560 samples) • Eyes closed • Numberofrecordings • 72 per participant ( 24 minutes ) • 720 total (4 hours )

  9. 3-3. Physicalmovementanomalies

  10. 3-4. Initialization problem

  11. 3-5. Frequencydomain • The brain operates at low frequencies usually divided into six frequency bands:

  12. 3-6. Fast fouriertransform

  13. 3-7. Feature extraction • Time domainfeatures • Meansamplevalue • Zero crossing rate • Valuesabove zero • Frequencydomainfeatures • Peakfrequency • Peakfrequency magnitude • Signal power in eachfrequency band • Pdelta, Ptheta, Palpha, PbetaLow, PbetaHigh, Pgamma • Mean band power • Meanphase angle

  14. 3-8. Statistics • Chi-squaregoodness-of-fittest • Samples and features do not follow normal distribution. • Correlation • HighcorrelationbetweenPbetaLow and PbetaHigh(8 out 10 participants).

  15. 3-9. Distancemetric d = d(signal1,signal2) : X = signal1 Y = signal2 d1 = |X.PbetaLow / X.PbetaHigh - Y.PbetaLow / Y.PbetaHigh| d2 = |X.PbetaLow / Y.PbetaLow - Y.PbetaHigh /X.PbetaHigh| d3 = |X.Palpha / X.PbetaLow - Y.Palpha / Y.PbetaLow| d4 = |X.Palpha/ Y.Palpha - Y.PbetaLow / X.PbetaLow| d= d1 + d2 + d3 + d4

  16. 4-1. Distancecomputation 1 • Computation: All vs All • Genuine attempts: • d(signal1,signal2) from the same participant • Fraudulent attempts • d(signal1,signal2) from differentparticipants • Requirement: • d(signal1,signal2)must be from the same task

  17. 4-2. DET-Curve 1 EER = 30.28%

  18. 4-3. Distancecomputation 2 • Computation: All vs All • Genuine attempts: • d(signal1, signal2) from the same participant • Fraudulent attempts • d(signal1, signal2) from differentparticipants • Requirement: • d(signal1, signal2) must be from the same task AND the same session.

  19. 4-4. DET-Curve 2 EER = 23.26%

  20. 4-5. Task selection

  21. 4-6. Distancecomputation 3 • Computation: Task selection • Genuine attempts • d(signal1,signal2) from the same participant • Fraudulent attempts • d(signal1,signal2) from differentparticipants • Requirement • d(signal1,signal2) must be from theselectedsession 1 task.

  22. 4-7. DET-Curve 3 EER = 21.46%

  23. 4-8. Distancecomputation 4 • Computation: Task selection • Genuine attempts • d(signal1,signal2) from the same participant • Fraudulent attempts • d(signal1,signal2) from differentparticipants • Requirement • d(signal1,signal2) must be from theselectedsession 1 task AND the same session.

  24. 4-9. DET-Curve 4 EER = 17.08%

  25. 4-10. DET-Curve 1-4 EER = 30.28% EER = 23.26% EER = 21.46% EER = 17.08%

  26. 5. Conclusion • Similiaritiesaresessionbased • Two consequtive signals areverysimilar • Equipmentdependant • Signal getsbetter over time • Capturestoomuchphysicalmovement • One sensor is not enough • Limited information • Lowsample rate

  27. 6. Furtherwork • Better distancemetric • Identify more feature relations • Trydifferent feature combinations • Better selectionoftasks • Tasks designed for the Fp1 location • New equipment • Better filtering • Increasedsamplefrequency • More sensors • Different sensor locations

  28. Thankyoufor listening! Questions?

More Related