Scholarly article on topic 'Biometric Knowledge Extraction for Multi-factor Authentication and Key Exchange'

Biometric Knowledge Extraction for Multi-factor Authentication and Key Exchange Academic research paper on "Computer and information sciences"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Computer Science
OECD Field of science
Keywords
{"authenticated key exchange" / biometrics / "knowledge extraction"}

Abstract of research paper on Computer and information sciences, author of scientific article — Phillip H. Griffin

Abstract This paper describes a method for achieving strong, multi-factor and mutual authentication from a biometrics-based protocol for authenticated key exchange (B-AKE). Operation of the protocol relies on knowledge shared by communicating parties, extracted from data collected by biometric sensors. A Diffie-Hellman key-agreement scheme creates a symmetric encryption key using a weak secret, the extracted something-you-know data. This key protects the confidentiality of user credentials and other message data transferred during operation of the B-AKE protocol. If the message recipient possesses the same something-you-know information as the sender, a key is created, the message decrypted, and mutual authentication achieved. Biometric match data recovered from the encrypted message provides a second something-you-are authentication factor. The B-AKE protocol ensures users never reveal their knowledge or biometric credentials to imposter recipients or man-in-the-middle observers. Diffie-Hellman key establishment provides forward secrecy, a highly desirable protocol property, when participants choose fresh random values each time they operate the protocol.

Academic research paper on topic "Biometric Knowledge Extraction for Multi-factor Authentication and Key Exchange"

(8)

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedía Computer Science 61 (2015) 66-71

Complex Adaptive Systems, Publication 5 Cihan H. Dagli, Editor in Chief Conference Organized by Missouri University of Science and Technology

2015-San Jose, CA

Biometric Knowledge Extraction for Multi-Factor Authentication

and Key Exchange

Phillip H. Griffin*

Griffin Information Security, 1625 Glenwood Avenue, Raleigh, NC 27608 USA

Abstract

This paper describes a method for achieving strong, multi-factor and mutual authentication from a biometrics-based protocol for authenticated key exchange (B-AKE). Operation of the protocol relies on knowledge shared by communicating parties, extracted from data collected by biometric sensors. A Diffie-Hellman key-agreement scheme creates a symmetric encryption key using a weak secret, the extracted something-you-know data. This key protects the confidentiality of user credentials and other message data transferred during operation of the B-AKE protocol. If the message recipient possesses the same something-you-know information as the sender, a key is created, the message decrypted, and mutual authentication achieved. Biometric match data recovered from the encrypted message provides a second something-you-are authentication factor. The B-AKE protocol ensures users never reveal their knowledge or biometric credentials to imposter recipients or man-in-the-middle observers. Diffie-Hellman key establishment provides forward secrecy, a highly desirable protocol property, when participants choose fresh random values each time they operate the protocol.

© 2015TheAuthors.PublishedbyElsevierB.V.This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of scientific committee of Missouri University of Science and Technology Keywords: authenticated key exchange; biometrics; knowledge extraction

1. Introduction

Distinctive biometric traits can uniquely identify an individual. Biometrics is the automatic recognition of individuals based on these unique traits, the biological (e.g., fingerprint, iris, hand geometry) or behavioral (e.g., gait, gesture, keystroke dynamics) characteristics that reliably distinguish one person from another [1]. In an access control system, biometrics is the well-known something-you-are factor used for identification and authentication.

* Corresponding author. Tel.: +1-919-291-0019. E-mail address: phil@phillipgriffin.com

1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of scientific committee of Missouri University of Science and Technology doi:10.1016/j.procs.2015.09.150

Biometrics can be coupled with other categories of factors, such as something-you-have and something-you-know, to achieve two and three-factor authentication when greater assurance is required than a single factor can provide [1].

Traditionally, a distinct source of values has been associated with each type of authentication category. Access control systems typically collect values for each authentication factor category separately, perhaps relying on a token reader, a keyboard, and a biometric sensor for three-factor authentication. However, binary data collected from biometric sensors contains rich information content not limited to only the physiological and behavioral characteristics needed to support biometric matching.

Sensor collected data can contain human knowledge, something-you-know information. Knowledge extraction techniques applied to this data can reveal weak secrets expressed by an individual. These secrets are termed 'weak' because they are something a person can easily memorize, traditionally a pass phrase, a password, or a personal identification number (PIN). Biometric-based weak secrets extend these traditional secrets to also include a sequence of footsteps or the "the finger positions and hand postures" used in a gesture or "during communication of hand sign languages" [2].

A password is a something-you-know authenticator, a string of characters that have an equivalent representation shared by communicating parties. This equivalency makes passwords useful as weak secrets in cryptographic keyagreement schemes, which require both parties to know exactly the same secret in order to establish a shared key to ensure secure communications. Though knowledge extracted from biometric sensors can have this useful equivalency attribute of passwords, often it does not.

When an authenticator constructed from knowledge mined from biometric sensor data is suitable for use in Diffie-Hellman based authenticated-key exchange (AKE) protocols [3] it can serve in place of a password string. However, for the successful operation of these protocols, it is necessary for participants in the exchange to represent the knowledge information in a consistent and unambiguous format, such as a canonical encoding based on Abstract Syntax Notation One (ASN. 1) [4]. Doing so allows protocol participants to share precisely the same secret knowledge required to operate the protocol.

2. Authenticated-Key Exchange

Server authentication mechanisms, such as the widely deployed Transport Layer Security (TLS) protocol, rely on strong asymmetric cryptography supported by a resource intensive Public Key Infrastructure (PKI). However, achieving mutual authentication using TLS is not so common. It is more likely for client authentication to rely on user passwords, since most users lack the personal digital certificates needed to leverage the mutual authentication option of TLS [5].

Passwords and other shareable knowledge-based authentication values are "typically used for client-side authentication only" [6], with TLS serving to authenticate the server and protect client passwords in transit. Failures in TLS server authentication and user errors have led to widespread phishing by attackers impersonating legitimate servers to capture user credentials [5]. Authenticated Key Exchange (AKE) protocols can help defeat phishing attacks, protect passwords using strong cryptography, and provide mutual authentication without support of a PKI. The AKE family of protocols provides a strong, lightweight alternative to TLS.

Password-Authenticated Key Exchange (PAKE) protocols "offer password-based mutual authentication" through the "the marriage of password authentication with cryptographic key exchange protocols" [6]. PAKE provides communicating parties assurance they know each other (e.g., mutual authentication) and establishes a shared secret, a symmetric cryptographic key for secure communications known only to them. With PAKE, user "authentication is accomplished implicitly through the capability of establishing an authenticated session key" [6]. PAKE provides a strong key exchange mechanism, even when using weak passwords, so long as users protect their passwords from exposure to attackers.

The Biometric-Authenticated Key Exchange (B-AKE) protocols proposed in this paper extend the use of PAKE beyond the limitations of character string passwords to more general knowledge representations. B-AKE authenticators may be in either character or binary string format, and they may contain human-readable markup or other structured content. For protocol processing purposes, B-AKE and PAKE authenticators are both opaque strings with no distinguishable internal format or semantics that must be pre-established before protocol operation.

Fig. 1 defines the schema for representing a user knowledge authenticator in a B-AKE protocol. Type PW is an 'open type', and can contain a value of any ASN.1 type in its encoded form. The type of an encoded value can be simple or structured. However, for processing purposes in a B-AKE protocol, an encoded value of type PW is an opaque string, a series of octets that are independent of hardware, operating system, or programming language considerations. This serialized format is ideal for information exchange between communicating parties that have different computing environments.

PW ::= KNOWLEDGE.SType -- Implementation constrained

4id OBJECT IDENTIFIER UNIQUE, WITH SYNTAX { KNOWLEDGE Sid [ DATA iType ] ( Fig. 1. B-AKE knowledge authentication string.

Unlike passwords, biometric matching data is not a shareable authenticator, since biometric reference data and biometric matching data are not equivalent. A user first enrolls their biometric data to create a reference template used later for authentication against the users provided biometric matching data [1]. Even when two biometric samples "A and B" are close, "their cryptographic hashes h(A) and h(B) are very different, making h(A) not appropriate" for use in a PAKE protocol as a shared authenticator [6].

Modeled on the international PAKE standard [7], B-AKE is not restricted to only the use of password character strings for the value of PW in the key exchange processing defined in the ITU-T X.1035 recommendation. B-AKE uses the same processing as the international standard, but allows the use of more general, pre-shared something-you-know biometric knowledge in both client-side and server-side authentication. Like the mutual authentication option of TLS, but without the overhead of a PKI, B-AKE protocols ensure "mutual authentication of both parties in the act of establishing a symmetric cryptographic key via Diffie-Hellman exchange" [7]. Fig. 2 illustrates B-AKE protocol operation.

1. user provides biomettrc input to [eleblometrlc sensor

2. ilgenr extracts knowledge Si matching factors from input Symmetric key Is created usl ng secret f! nowledgefactor

4. Key encrypts biometric matching factor (otherdata} Client sends encrypted authentication message to server

Client

6- rnr rypted authentication message received from client

7. Svm metric Lev is crea ted using shared secret knowledge

H. Key decrypts biometric matching factor {other data) a nd user is authenticated with knowfed&e factor

biometric matching factor further authenticates usee

10 server sends encrypted authentication proof te client

Fig. 2. B-AKE protocol operation.

By using a Diffie-Hellman key exchange, the B-AKE protocol ensures "perfect forward secrecy - a property of a key establishment protocol. This property guarantees that compromise of a session key or long-term private key after a given session does not cause the compromise of any earlier session" [7] when fresh random values are used to operate the protocol. Encryption based on a user knowledge credential during operation of the protocol protects

the key exchange operation "from the man-in-the-middle attack", and leaves the weak secret unrevealed "to an eavesdropper preventing an off-line dictionary attack" [7].

Biometric data may contain embedded user knowledge, such as a passphrase or the spelled out letters in a password. When knowledge extraction can mine something-you-know information from biometric sensor data, this data can provide both biometric matching data and the shared knowledge required to operate a B-AKE protocol. In the case of speaker recognition, a user voiceprint may include a series of words or letters whose binary values can be extracted using speech recognition techniques, and then transformed into a character string.

Biometric collection of user gestures as binary video images can capture similar embedded knowledge. Gestures can be in the form of a passphrase or password, or they can represent movements completely unrelated to any language. Unlike passwords entered using conventional keyboard devices, user knowledge extracted from biometric sensor data requires additional transformation processing before being suitable for use as an authenticator, or as input to a B-AKE protocol.

3. Gestures

Fong, Zhuang, and Fister [2] describe an image-based, biometric authentication model for hand gestures captured by video recordings. The authors demonstrate that a "sequence of hand signs", such as the those representing the letters "'i' , 'l' , 'o' , 'v' , 'e' , and 'u" can be encoded as a series of gesture images and used to authenticate the claimed identity of an individual [2]. The gestures serve as what the authors refer to as a "biometric password" and provide context for biometric feature extraction and biometric matching based on the "hand shape and the postures in doing those signs" [2].

Their approach achieves two-factor authentication. A something-you-are factor is obtained by matching the "slight but unique behavioral characteristics in sign language" presented by the individual to their previously enrolled biometric reference data. A something-you-know factor relies on extracting the alphabetic letters associated with this sequence of hand signs from images, as shown in Fig. 3.(a), and matching these letters to a previously registered user secret [2].

I/Ill fWi9

uri'i ijmH

f ¥ / ' # * y" §

a vf a >1 ill

Fig. 3. (a) Finger-spelled Alphabet [10] (b) Numbered Finger-spelled Alphabet.

The process derives two authentication factors from the same biometric sensor data. However, the extracted knowledge is not yet in a form suitable for use in an authenticated key exchange protocol. Successful operation of an AKE requires the user secret to be in an identical format on the client and server, so that both parties can establish the same symmetric key. In the Fong, Zhuang, and Fister model, each hand sign is associated with one character, a letter of the alphabet. While their approach allows input of alphabetic and numeric letters, it prohibits the use of hand signs for whole words, phrases, concepts, or gestures not tied to an alphabet or language.

To create a less restrictive collection of hand signs, each letter in the hand sign alphabet can be assigned a number (See Fig.3(b)) based on its position in the alphabet (e.g., A=1, B-2 ... Z=26), followed by the ten numeric letters (e.g., 0=27, 1=28 ... 9=36). To extend the collection to include more complex hand signs, such as the clinched fist, the hand-over-heart salute, and the benediction gesture, each additional gesture must become a unique numeric value.

The hand signed letters "'i', 'l', 'o', 'v', 'e', and 'u" represented as a sequence of integers, can be the value of an ASN. 1 relative object identifier type "9.12.15.22.5.21" [4]. This value can be transferred unambiguously in an instance of communication in a compact binary format, encoded as the 8-byte hexadecimal value "0D06090C 0F160515" using the Distinguished Encoding Rules (DER) of ASN.1 [4]. Encoded integer values also can represent complex gestures, such as the clinched fist gesture, that are not associated with a signing alphabet in the same way.

Sae-Bae, Memon, Isbister, and Ahmed investigate "user-defined gestural passwords", multi-touch gestures for user authentication on touch-sensitive devices [8]. Their research leverages the capabilities found frequently on mobile devices, such as smart phones and tablet computers. These devices "contain a rich array of sensors that can support many authentication methods beyond passwords and PIN codes, including biometrics" [9]. The authors focus on gestures using biometric technology that measures "variations in hand geometry and muscle behavior" [8].

The gesture images collected by their biometric sensors are not associated with any alphabet or language. Instead, "a set of canonical gestures were identified by creating a gestural taxonomy based on movement of the palm, the fingertips, and the number of fingertips involved in the gesture" [8]. The authors define a "multi-touch gesture" as a "time series of the set of x-y coordinates of finger touch points captured as the gesture is being performed" [8].

Table 1 provides a partial set of the canonical gesture types defined by the authors. Their type definitions are in order by numeric gesture identifier. Each identifier names a unique to a set of annotation text, palm movement, fingertip movement, and dynamic fingertip values that define a gesture in a gestural taxonomy table.

Table 1. Ordered gesture descriptions.

Gesture identifier Annotation Palm movement Fingertip movement Dynamic fingertips

1 'CCW' Static Circular(CCW) All

2 'CW' Static Circular(CW) All

3 'Pinch' Static Close All

4 'Drag' Dynamic(j) Parallel All

5 'FPO' Static Open Fixed pinky

6 'FTCW' Static Circular(CW) Fixed thumb

7 'FTP' Static Parallel^) Fixed thumb

8 'Opened' Static Open All

9 'Swipe' Dynamic(^) Parallel All

10 'User-defined' Dynamic Parallel All

The authors note that for the purposes of biometric matching, the "recognition performance of the proposed system" results in error rates that are "still too high for authentication systems" [8]. However, a series of canonical gestures "'CCW', 'Pinch', 'Drag', 'Swipe', 'CW', 'User-defined', and 'FTP", knowledge extracted from biometric sensor data and represented as a sequence of integers can be the value of an ASN.1 relative object identifier type "1.3.4.9.2.10.7" [4]. This knowledge value extracted from biometric information can be DER encoded as the 9-byte hexadecimal value "0D070103 04090A07" and used as the PW input to B-AKE. Even when a biometric technology type is too weak for use as an authenticator, knowledge data collected from biometric sensors can achieve strong mutual authentication and create a shared cryptographic key to ensure confidentiality between communicating parties.

4. Conclusions and Future Research

An authenticated key exchange (B-AKE) protocol can achieve mutual authentication and strong, two-factor user authentication with biometric matching data and user knowledge extracted from biometrics sensor data. Neither TLS nor the overhead of a PKI is required to achieve mutual authentication or to establish a shared cryptographic key for secure communications. B-AKE ensures the confidentiality of user authentication credentials from phishing and

man-in-the-middle attacks, and proper operation of the protocol makes forward secrecy possible.

Future research and development will inform definition of a complete ASN.1 message schema to support B-AKE protocol implementations using biometric technologies of any type. This schema will include structured knowledge authenticator values that can serve as inputs to the ITU-T X.1035 recommendation. These inputs will be in the form of canonical encodings of complex values represented as opaque strings to replace the simple character string password inputs defined in the international standard.

For some biometric technology types, a contrived context for biometric matching can provide the opportunity for collecting additional knowledge information from biometric sensor data. From an access control perspective, a context used to provide an additional authentication factor can be considered a layer of defence. Yun [11] provides such a context for knowledge collection, a "floor-based system, UbiFloorII, which consists of a large number of photo interrupter sensors in wooden tiles" used to study gait biometric identification [11].

The Yun system collects "both the walking and stepping patterns from the walking samples" [11] for gait biometric matching using a two-dimensional biometric sensor grid. The order and identities of tiles encountered as participants traverse the grid could convey user knowledge through a stepping pattern. A series of identifiers mapped to user-selected tiles forming a memorized stepping pattern could serve as the user's weak secret in a BAKE protocol.

Gait is appealing as a biometric identifier, since sample collection is "unobtrusive and typifies the motion characteristics specific to an individual" [11]. Gait measurement does not require user cooperation and the characteristics of an individual "can be detected and measured at both a low resolution and a long distance" [11]. These attributes make gait biometrics ideal for smart home applications and useful for vehicle and building access control.

Future research will explore the similarities between gait biometrics knowledge extraction using a 'smart floor' and gesture biometrics knowledge extraction using smart phone touch screens. Both technology types support the capture of biometric matching data using a similar ordered grid context useful for collecting knowledge information needed to operate a B-AKE protocol.

References

1. Griffin, Phillip. H. (2014). Telebiometric Authentication Objects. Complex Adaptive Systems 2014 Proceedings. Procedia Computer Science,

36, 393-400. Retrieved June 12, 2015, from http://www.sciencedirect.com/science/article/pii/S1877050914012605

2. Fong, S., Zhuang, Y., & Fister, I. (2013). A biometric authentication model using hand gesture images. Biomedical engineering online, 12(1),

111. Retrieved June 12, 2015, from http://www.biomedical-engineering-online.com/content/12/1/111/.

3. Abdalla, M., Bresson, E., Chevassut, O., Moller, B. and Pointcheval, D. (2007). 'Strong password-based authentication in TLS using the three-

party group Diffie-Hellman protocol', Int. J. Security and Networks, Vol. 2, Nos. 3/4, pp.284-296. Retrieved June 12, 2015, from http://www.ssi.gouv.fr/uploads/IMG/pdf/AbdBreCheMol_07.pdf.

4. Larmouth, John L. (2000). ASN.1 Complete. Morgan Kaufmann. Retrieved June 12, 2015, from http://www.oss.com/asn1/resources/books-

whitepapers-pubs/larmouth-asn1-book.pdf

5. lsaid, A., & Mitchell, C. J. (2006). Preventing phishing attacks using trusted computing technology. In Proceedings of the 6th International Network Conference (INC'06) (pp. 221-228). Retrieved June 12, 2015, from https://www.cs.auckland.ac.nz/research/groups/ssg/pastbib/ pastpapers/alsaid06preventing.pdf

6. Wang, X., & Lin, H. (2010). Cryptography-Based Authentication for Protecting Cyber Systems. Applied Cryptography for Cyber Security and

Defense: Information Encryption and Cyphering: Information Encryption and Cyphering, 32.

7. ITU-T X.1035: Password-authenticated key exchange (PAK) protocol (2007). Retrieved June 12, 2015, from http://www.itu.int/rec/T-REC-

X.1035-200702-I/en

8. Sae-Bae, N., Memon, N., Isbister, K., & Ahmed, K. (2014). Multitouch Gesture-Based Authentication. IEEE Transactions on Information Forensics and Security, 9(4), 568-582.

9. Griffin, Phillip H. (2014). Web Services Security for All. Information Systems Security Association Journal, Vol. 12, No. 9. Retrieved June 12,

2015, from http://phillipgriffin.com/whitepapers/WebServices_ISSA0914.pdf

10. Vicars, W. (2011). American Sign Language (ASL). Retrieved June 12, 2015, from http://.www.lifeprint.com

11. Yun, J. (2011). User identification using gait patterns on UbiFloorII. Sensors, 11(3), 2611-2639, doi:10.3390/s110302611. Retrieved June 12, 2015, from http://www.mdpi.com/! 424-8220/11/3/2611/htm