Biometrics via IR Spectroscopy of the Epidermis: Potential and Difficulties

Transcription

1 Biometrics via IR Spectroscopy of the Epidermis: Potential and Difficulties by David M. Mackie ARL-TR-0413 September 2012 A reprint from the Proc. SPIE: Biometric Technology for Human Identification IX Approved for public release; distribution unlimited.

2 NOTICES Disclaimers The findings in this report are not to be construed as an official Department of the Army position unless so designated by other authorized documents. Citation of manufacturer s or trade names does not constitute an official endorsement or approval of the use thereof. Destroy this report when it is no longer needed. Do not return it to the originator.

3 Army Research Laboratory Adelphi, MD ARL-TR-0413 September 2012 Biometrics via IR Spectroscopy of the Epidermis: Potential and Difficulties David M. Mackie Sensors and Electron Devices Directorate, ARL A reprint from the Proc. SPIE: Biometric Technology for Human Identification IX Approved for public release; distribution unlimited.

4 REPORT DOCUMENTATION PAGE Form Approved OMB No Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing the burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports ( ), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. 1. REPORT DATE (DD-MM-YYYY) September REPORT TYPE 3. DATES COVERED (From - To) 4. TITLE AND SUBTITLE Biometrics via IR Spectroscopy of the Epidermis: Potential and Difficulties 5a. CONTRACT NUMBER 5b. GRANT NUMBER 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) David M. Mackie 5d. PROJECT NUMBER 5e. TASK NUMBER 5f. WORK UNIT NUMBER 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) U.S. Army Research Laboratory ATTN: RDRL-SEE-E 2800 Powder Mill Road Adelphi, MD PERFORMING ORGANIZATION REPORT NUMBER ARL-TR SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR S ACRONYM(S) 11. SPONSOR/MONITOR'S REPORT NUMBER(S) 12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited. 13. SUPPLEMENTARY NOTES 14. ABSTRACT We discuss the potential and difficulties of using infrared (IR) spectroscopy of the human epidermis as a biometric. We present preliminary data on the fingerpads of 9 individuals demonstrating the potential for uniqueness and stability. We also present data on the challenges presented by complications such as sebum changes, intra-individual location variability, and skin care products. 15. SUBJECT TERMS biometrics, IR, infrared, spectroscopy, epidermis, skin, FTIR 16. SECURITY CLASSIFICATION OF: a. REPORT UNCLASSIFIED b. ABSTRACT UNCLASSIFIED c. THIS PAGE UNCLASSIFIED 17. LIMITATION OF ABSTRACT UU 18. NUMBER OF PAGES 13 19a. NAME OF RESPONSIBLE PERSON David M. Mackie 19b. TELEPHONE NUMBER (Include area code) (301) Standard Form 298 (Rev. 8/98) Prescribed by ANSI Std. Z39.18 ii

5 Biometrics via IR Spectroscopy of the Epidermis: Potential and Difficulties David M. Mackie* U.S. Army Research Laboratory, 2800 Powder Mill Road, Adelphi, MD, USA ABSTRACT We discuss the potential and difficulties of using infrared (IR) spectroscopy of the human epidermis as a biometric. We present preliminary data on the fingerpads of 9 individuals demonstrating the potential for uniqueness and stability. We also present data on the challenges presented by complications such as sebum changes, intra-individual location variability, and skin care products. Keywords: biometrics, IR, infrared, spectroscopy, epidermis, skin, FTIR 1. INTRODUCTION A number of papers in the literature discuss using infrared (IR) spectroscopy of human skin, in conjunction with microscopy, to differentiate between healthy skin and skin with various diseases. 1 The regions of diseased skin have altered structures or compositions that give rise to tell-tale differences in the IR spectra for that region, compared to normal skin. Other papers discuss using (destructive) biochemical tests to discover the compositional differences in the epidermis of individuals. 2,3 It is shown that individuals have distinct proportions of various epidermal proteins. Lastly, a patent discusses the possibility of visible-light spectroscopy as a biometric (without much supporting data). 4 In the research presented here, we combine these ideas. Specifically, we discuss using infrared (IR) spectroscopy of the human epidermis as a biometric. This biometric could be, in principle, non-contact and fast. It is also easily combined with other biometrics, such as fingerprints or facial recognition. The initial data are necessarily preliminary, with a sample size of only 9 individuals, but seem promising. From the start we took it as a given that most biometric applications would not be conducive to special preparations of the skin (such as careful cleaning with alcohol or a special soap). This introduces additional challenges, some of which we address. 2. METHODOLOGY Although a non-contact spectroscopy system would be ideal, we tested the idea first using available instrumentation. We used a typical laboratory Fourier-transform infrared (FTIR) spectroscopy instrument (Nicolet 6700 FT-IR from ThermoScientific), which had a liquid nitrogen cooled deuterated L-alanine doped triglycine sulfate (DTGS) detector. The data presented here was taken with a diamond attenuated total reflection (ATR) attachment (GladiATR from Pike Technologies) to the FTIR. The GladiATR has a pressure mechanism to insure good contact between the sample and the diamond surface. This pressure mechanism was used for the measurements on epidermal components. Our experience with epidermal components was that the amount of pressure didn t matter beyond a certain point. (This was consistent with the GladiATR manual.) Consequently, unregulated pressing was used for the human measurements. The pressure was controlled by the individual being tested, with no feedback or special training, other than an admonition to apply steady pressure and hold still for 30 seconds. Each test comprised 16 complete scans from 400 to 4000 cm -1, and these were averaged. Nevertheless, fidgeting may account for the (slight) variability seen in repeated tests of the same individual. The GladiATR attachment was not strictly necessary for the epidermal measurements. We also achieved good results with a germanium trough ATR attachment. This had the advantage of measuring an average of all four nonthumb fingerpads at once. However, with no pressure mechanism it gave poor results for the epidermal components, due to contact problems, exacerbated by the high refractive index of the germanium (and the attendant small penetration depth of the total internal reflections). Thus, all of the results reported here are with the GladiATR attachment. *david.m.mackie.civ@mail.mil; phone ; fax ; Legal notice: Mention of any product or company in this paper does not constitute and should not be construed as a recommendation or endorsement by the author, his employer, or the U.S. government.

6 Transmittance (%) 3.1 Demonstration of principle 3. RESULTS In Figure 1 we show a typical IR spectrum from a human fingerpad, together with IR spectra from several of the more common components of human epidermis. (By fingerpad, we mean the fleshy area of the fingertip containing the fingerprints.) Data was taken between 400 and 4000 cm -1, but the region shown from 600 to 1800 cm -1 was especially rich in features. By comparing peak locations and strengths, one can see how the components of human epidermis combine to give epidermal spectra. Not all epidermal components are shown (and not all have yet been tested), so there are some features of the epidermal IR spectrum that aren t explainable from the component IR spectra shown. In interpreting this data, one should keep in mind that the spectra shown are not normalized in any way, and the components don t occur in equal amounts in the epidermis. Most of the components have distinctive water features in their spectra, which may be due to moisture in the samples. The melanin sample is the exception, perhaps because it was the only component that was synthesized rather than extracted from biological sources. The overall similarity between the epidermal spectrum and the keratin spectrum should not be surprising. Keratinocytes are of course the main cells of the epidermis, and they become increasingly filled with keratin (keratinized) as they migrate outward. 100 Comparison of Epidermal Components and Fingerpad Epidermis Caucasian (light-skinned) fingerpad Keratin powder Melanin powder Collagen fiber Hyaluronic acid Water (deionized) Wavenumber (cm -1 ) Figure 1: A typical IR spectrum from a human fingerpad, together with spectra from several of the more common components of human epidermis. The other components of the epidermis add their distinctive features to the dominant keratin trend. It is clear that if the relative abundance of epidermal components varies between individuals, then different epidermal IR spectra will result.

7 Transmittance (%) This is in fact what occurs, and it is shown in Figure 2 for 9 individuals. These are typical IR spectra, unnormalized. The distinguishing features are subtle, but real, as we will show in the next section. Normalizing these spectra in various ways does not eliminate distinguishing features. 100 Comparison of Epidermal Spectra for Individuals Hispanic female (lt. med. cmplxn.) South Asian male (med. cmplxn.) African-American male #1 (med. cmplxn.) African-American male #2 (med. cmplxn.) Western European male #1 Western European male #2 Western European male #3 East Asian male #1 (light cmplxn.) East Asian male #2 (light cmplxn.) Wavenumber (cm -1 ) Figure 2: Typical IR spectra from 9 human fingerpads, unnormalized. Normalizing does not eliminate distinguishing features. 3.2 Potential of the biometric Before proceeding to the topic of identification, we wish to point out the most obvious use of this biometric: confirming human skin. There is a complicated overall pattern for epidermal IR spectra that is consistent for human epidermis (for every individual tested to date), independent of gender, age, ethnic background, or pigmentation. (Compare Fig. 2.) This application by itself would be useful in combination with automated fingerprint or facial recognition. It would eliminate spoofing of those biometrics via latex masks. Figure 3 shows the effect of latex on epidermal IR spectra. Latex has two very strong peaks: a broad asymmetric one centered near 1430 cm -1, and a narrow one centered near 875 cm -1. Note that the latex spectrum has been shrunk by a factor of 4 for display purposes. A fingerpad covered in a latex medical examination glove shows up only as latex. A fingerpad newly-removed from the glove still shows the effect of the latex remnants. The peak near 875 cm -1, which never exists in normal fingerpad spectra, can be used to estimate the latex contribution and subtract it away (not shown). This would be useful in laboratories and hospitals where people commonly wear latex gloves as part of their job.

8 Transmittance (%) Effect of Latex Latex glove, shrunk by 4 Fingerpad in latex glove Fingerpad out of latex glove Fingerpad, normal Wavenumber (cm -1 ) Figure 3: IR spectroscopy is very sensitive to the presence of latex. Of course, the main point of biometrics is to tell people apart. Our results in this area are preliminary but promising. Our data set is still small (9 people), and the spectral differences are subtle. However, they are reproducible (Figs. 4-6). Figure 4 shows six fingerpad IR spectra (two sets of three) for two male Caucasians with similar ages, ethnic backgrounds, and skin coloration. The red, dark blue, and bright purple curves belong to one man; the green, cyan blue, and dark purple curves belong to the other man. The spectra have been normalized using Omnic software so that the transmittance maxima and minima are the same within the chosen wavenumber region. The characteristic difference between the two spectra then becomes clear: one man has large dips at 1120 and 1040, compared to 1060; the other man does not. As another example, Figure 5 shows six fingerpad IR spectra (two sets of three) for two male African- Americans with similar ages, ethnic backgrounds, and skin coloration. The red, dark blue, and gold curves belong to one man; the green, cyan blue, and bright purple curves belong to the other. The spectra have been normalized as above. Figure 4: Fingerpad IR spectra for two male Caucasians with similar ages, ethnic backgrounds, and skin coloration. The red, dark blue, and bright purple curves belong to one man; the green, cyan blue, and dark purple curves belong to the other.

9 Figure 5: Fingerpad IR spectra for two male African-Americans with similar ages, ethnic backgrounds, and skin coloration. The red, dark blue, and gold curves belong to one man; the green, cyan blue, and bright purple curves belong to the other. Figure 6: Fingerpad IR spectra for two male East Asians with similar ages, ethnic backgrounds, and skin coloration. The red, dark green, and light green curves belong to one man; the dark purple and bright purple curves belong to the other. Upon normalization, the characteristic differences in Fig. 5 become apparent. As a last example, Figure 6 shows five fingerpad IR spectra (one set of three, one set of two) for two male East Asians (specifically, Han Chinese) with similar ages, ethnic backgrounds, and skin coloration. The red, dark green, and light green curves belong to one man; the dark purple and bright purple curves belong to the other man. The spectra have been normalized as above, and again the

10 characteristic differences are apparent. We emphasize that we have not cherry-picked, i.e., only presented the best comparisons. Every individual in our data set can be easily distinguished from all the others, clearly and reliably. Obviously, though, large data sets would require some sort of automated chemometric analysis. 3.3 Difficulties with the biometric A biometric, to be useful for identification (ID) and/or confirmation, must be not merely specific to individuals, but also stable for each individual. Unfortunately, epidermal IR spectroscopy encounters great difficulties in that area. As an example, Figure 7 shows fingerpad IR spectra for a single individual (male Caucasian #3) taken on four different days (red, blue, green, and gold curves), together with spectra for four other individuals (light blue curves). It is apparent that the variation in a single individual over time is just as large as the variation among most individuals! The question then becomes: is this variation intrinsic and inherent to the human epidermis, or is the variation caused by an external factor? If the former, then epidermal IR spectroscopy may perhaps have some medical use, but the situation for ID seems hopeless. If the latter, then rescue of the biometric may be possible. Figure 7: Fingerpad IR spectra for a single individual (male Caucasian #3) taken on four different days (red, blue, green, and gold curves), together with spectra for four other individuals (light blue curves). We begin with the observation that epidermal IR spectra vary widely, depending on body location. Figure 8 shows epidermal IR spectra for male Caucasian #3, measured at the following locations: terminal fingerpad of right index finger (normal location in this paper), right cheek, center of chin (anterior fleshy part), middle of inner right forearm, middle of outer right forearm, middle of forehead, and tip of nose. (It was difficult to get good contact with the ATR using the inner forearm, so that spectrum is rather noisy.) Also measured, but not shown, were the following: end of index finger; second, third, and fourth pads and terminal knuckle of index finger; palm at base of thumb; outer palm; back of hand; and middle of outer forearm (all on the right hand/arm). It is, of course, possible that variations in the structure and/or composition of the epidermis account for some of this IR spectral variation. However, in every case, spectra that diverged most strongly from the fingerpad spectrum corresponded to parts of the body that naturally tend to be moist and/or oily. Human skin oil (sebum) is known to be a complex mixture of lipids, with the major constituents being triglycerides, cholesterol esters, free fatty acids, wax esters, and squalene. 5 It is secreted to varying degrees by sebaceous glands on every area of human skin except the palmar and plantar surfaces (i.e., front of hands and bottom of feet). It is easily and commonly transferred to fingerpads by touching the face, hair, or other body parts. Figure 9 shows the effect of this sebum on the epidermal IR spectra of the right index fingerpad of Caucasian #3. The spectra shown are as follows: no special treatment (green curve, before cleaning ); after rubbing against the nose and forehead to pick up sebum, then rubbing against the thumb to simulate activity (purple curve, after oiling ); and after careful cleaning with an alcohol

11 Figure 8: Variation of epidermal IR spectra with body location, for a single individual. Figure 9: Effect of skin oil (sebum) on epidermal IR spectrum of fingerpad. wipe (blue curve, after cleaning ). Also shown in Fig. 9 is the IR spectrum of the sebum alone, swiped from the same freshly-oiled fingerpad (red curve). Visual inspection of the curves leads one to believe that the after oiling and before cleaning spectra could be made to line up closely with the after cleaning spectrum by subtraction of the right percentage of the sebum spectrum. Calculations bear this out. The right percentage can be obtained by eliminating any traces of the two sebum peaks between 1700 and 1800 cm -1. The sebum peak near 700 cm -1 could also be used, but it is less pronounced. We have performed similar analyses to account for moisture variations (due to perspiration and/or hydration level) and for the presence of a specific moisturizer. Their effects can likewise be subtracted away, provided one knows the

12 spectrum of the moisture (not a problem) and the moisturizer (problematic). Sun tan lotion and makeup are also likely to interfere with ID using epidermal IR spectra. Since there are many possible interferents, it is not obvious that one could account for all of them without a priori knowledge of the subject who one is trying to ID. It may be possible to pick interferents from a database by their distinctive spectral features. Alternatively, it may be possible to take epidermal IR spectra from various parts of the subject s body (e.g., fingerpad, arm, forehead, and nose). Comparing unexpected spectral features present in some body locations but not others may enable a program to automatically estimate the innate spectrum of the subject s epidermis. However, it seems clear that successfully implementing epidermal IR spectroscopy as a biometric for ID purposes would not be straightforward. It would require either cooperation of the subjects or else clever post-processing. 4. CONCLUSIONS Aside from the difficulties of naturally occurring interferents (sebum, moisture) and artificial ones (lotions and creams), there still remains the need to test for specificity on a larger data set and for long-term stability against aging, tan level, and health and hormonal changes. Thus, applying this biometric to the problem of ID will require much more work. However, applying this biometric to defeat spoofing of automated fingerprint recognition systems via latex fingerprint masks is a reasonable near-term application. Although IR spectroscopy tests for human epidermis, it does not test directly for liveness, so in theory it might be defeated with an amputated or transplanted finger, if only fingers were being tested. However, an identification station that measured epidermal spectra at several body points simultaneously would be extremely difficult to trick in this manner. With current technology, epidermal IR spectroscopy could be made cost-effective for facility access control. It seems unlikely to be put into individual workstations or room doors for access control. However, testing for anomalous IR absorption at specific wavelengths may be useful in such cases to combat fingerprint spoofing. The FTIR instrument used (manufactured in 2007) could scan once from 400 to 4000 cm-1 in about 1 s. It was not set up for fast scans. Judging by the literature available online from FTIR manufacturers, a newly-built machine optimized for speed could scan the region of interest from 500 to 1800 cm-1 a hundred times in that same 1 s. We found that averaging only 16 scans gave sufficient SNR. The data processing necessary for ID is tricky but not computationally burdensome. We therefore estimate that, for a properly-designed system, identity verification via epidermal IR spectra would require less than 1 s per person. ACKNOWLEDGEMENT We wish to thank Troy Alexander for suggesting this area of research and obtaining internal funding. REFERENCES [1] Wang, L. and Mizaikoff, B., "Application of multivariate data-analysis techniques to biomedical diagnostics based on mid-infrared spectroscopy," Anal. Bioanal. Chem. 391, (2008). [2] Legrain, V., Michel, S., Ortonne, J. P., and Reichert, U., "Intra- and inter-individual variations in cornified envelope peptide composition in normal and psoriatic skin," Arch. Dermatol. Res. 283, (1991). [3] Tong, L., Corrales, R. M., Chen, Z., Villarreal, A. L., De Paiva, C. S., Beuerman, R., Li, D.-Q., and Pflugfelder, S. C., "Expression and regulation of cornified envelope proteins in human corneal epithelium," Invest. Ophthal. & Vis. Sci. 47(5), (2006). [4] Rowe, R. K., Miller, W. A., Ge, N., Robinson, M. R., "Apparatus and method for identification of individuals by near-infrared spectrum," U.S. Patent 2004/ A1 (2004). [5] Stefaniak, A. B., Harvey, C. J., and Wertz, P. W., "Formulation and stability of a novel artificial sebum under conditions of storage and use," Intl. J. Cosmetic Sci. 32, (2010).

13 NO OF COPIES ORGANIZATION 1 PDF DEFENSE TECH INFO CTR ATTN DTIC OCA 8725 JOHN J KINGMAN RD STE 0944 FT BELVOIR VA US ARMY RSRCH LAB ATTN IMAL HRA MAIL & RECORDS MGMT ATTN RDRL CIO LL TECHL LIB ATTN RDRL CIO LT TECHL PUB ATTN RDRL SEE B D STRATIS-CULLUM ATTN RDRL SEE G WOOD ATTN RDRL SEE B D MACKIE ATTN RDRL SES J EICKE ADELPHI MD