JILT 1998 (3) - Matthew Ford
Identity Authentication and 'E-Commerce' Matthew D. Ford Identity authentication is the process whereby some chosen attribute of a real-world entity ('the distinguishing character or personality of an individual') is demonstrated to belong to that entity. As ever larger numbers of organisations start to transact business over the Internet and other open networks, it becomes increasingly important to achieve secure identity authentication. A variety of identity authentication systems are available: including passwords, physical tokens, biometrics and digital signatures. Digital certificates attesting to a relationship between a documented attribute and a public key provide an extremely secure, affordable and flexible means of identity authentication. The procedures for establishing the veracity of a subscriber's claims are varied and depend upon both the type of certificate being sought and the level of assurance required. A taxonomy of commercially available digital certificates in relation to their procedures for identity authentication reveals some interesting differences and ambiguities in the terminology adopted by the organisations providing them. Above all, the paucity of information currently made available by such organisations indicates the embryonic state of these systems and their associated procedures. Authority authentication is highlighted as an area in particular need of more research. This is a Refereed Article published on 30 October 1998. Citation: Ford M, 'Identity Authentication and 'E-Commerce'', 1998 (3) The Journal of Information, Law and Technology (JILT). <http://elj.warwick.ac.uk/jilt/98-3/ford.html>. New citation as at 1/1/04: <http://www2.warwick.ac.uk/fac/soc/law/elj/jilt/1998_3/ford/> 1. Introduction Identity authentication systems and procedures are rapidly becoming one of the central issues in the practice and study of information systems development and security. The need to make electronic transactions secure is now a standard caveat attached to projections of e-commerce growth. While the global reach of the Internet and its potential for dramatically reducing transaction costs make it a very attractive medium for transacting business - to organisations and individuals alike - the inherent insecurities of an open network make this problematic. Modern asymmetric cryptography techniques are gaining currency as a practical technical solution to the problems posed by open networks. However, these techniques cannot solve the real-world human issues of confidentiality, identity and authority. This paper will focus on currently documented procedures for identity authentication and how they relate to the real world. Specifically, what are the mechanisms for confirming or authenticating a real-world identity? What levels of assurance can we attach to these different mechanisms? After refining the terms 'e-commerce' and 'identity authentication', the paper will provide an overview of the technical systems available for authenticating identity, before analysing the procedures available for identity authentication. An incomplete taxonomy of commercial solutions is presented and, finally, some comments on the increasingly important area of authority authentication are provided. N.B. This paper does not discuss the legal validity of identity authentication techniques. Regional and national jurisdictions are beginning to issue legislation concerning digital signatures and there is, currently, no internationally accepted standard. Readers should be aware that, whatever the identity authentication techniques involved, local legal stipulations need to be understood in order that the techniques are productively implemented. Cross-border recognition of digital certificates and other identity authentication methods is an issue that has only recently begun to be addressed. Before discussing identity authentication, it is necessary to be clear about what is meant by e-commerce. Electronic commerce has been a feature of information systems in large organisations for many years now. The implementation of Electronic Data Interchange (EDI) systems, beginning in the early 1980s, facilitated electronic ordering, stock control, invoicing and all the features of a modern business transaction (excluding, until fairly recently, payment). Ensuring the security of these systems has been a relatively insignificant problem as they are constructed around closed networks between long-established business partners with prior, real-world, business relationships. Issues of trust, identity and confidentiality are, therefore, relatively unproblematic in this context. More recently, the exponential growth of the Internet has created both the potential for, and a desire to conduct, these same business-to-business transactions, but also, critically, business-to-consumer transactions, across open networks. This development has given rise to some significant concerns. The openness of the network creates problems of confidentiality, and, in combination with the absence of a prior real-world relationship between the transacting parties, problems of identification and trust. We should not focus exclusively on the commercial potential of modern computer communications networks however. Wherever two parties, be they individuals or organisations, public or private, wish to conduct a transaction across an open network such as the Internet, concerns as to confidentiality and identity will arise. Financial transactions across the Internet are usually cited as the best example of the need for confidentiality and identity authentication, for obvious reasons. However, governments and other public bodies also need to ensure rigorous identity authentication procedures are implemented before they can undertake electronic transactions with private individuals. The negative consequences of issuing a bogus passport are, surely, greater than those incurred from non-payment of an invoice. The following analysis of identity authentication issues, systems and procedures will, therefore, have relevance to any individual or organisation, whether public or private, wishing to transact across an open network for profit, or otherwise. Identity n 2 the distinguishing character or personality of an individual Identity authentication is the process whereby some chosen attribute of a real-world entity ('the distinguishing character or personality of an individual') is demonstrated to belong to that entity. For example, if I am standing in front of you (provided that you are not visually impaired) the unique topography of my face is demonstrated to belong to me because you can see that it is attached to the rest of my body and that it displays all the subtleties of expression that you would expect in a human face. Thus, seeing my face permits identity authentication to take place. Note that I do not need to tell you my name. This basic instance of identity authentication serves little purpose, except that it would permit a connection to be made between this instance of seeing my face and another instance of seeing my face in the future. In that case you could be reasonably certain that this was the same person you had seen before. This is an example of a very weak identity. A stronger identity is created, for example, when I tell you my name. Provided that you trust me not to lie about my name, this newly authenticated identity is more useful: you can find my phone number in a directory, or my office in a building. Note that trust is also implicit in the example of weak identity above: I could be a polymorphic extra-terrestrial. In the field of information systems, identity authentication is increasingly important as more and more social and business transactions take place on-line. The complete absence of a physical presence coupled with the ever-present risks of communications interception demand that strong, easily verifiable and unique identity tokens are used during on-line transactions. These identity tokens come in various forms and are detailed in the next section. It is important to be aware that the meaning of 'identity' is contingent upon the purpose for which its verification is being sought. In the definition of identity authentication given above, the phrase 'chosen attribute' reflects this. In the dictionary definition of identity, what makes a characteristic 'distinguishing'? For some applications it may be a name, for others an age or a gender. In many e-commerce applications, the distinguishing characteristic is that of credit-worthiness, often coupled with a name and address for marketing purposes. This identity contingency gives rise to a need for flexibility in identity authentication systems if they are to be widely implemented. These issues of attribute selection, trust, and contingency are central to the debate surrounding modern identity authentication systems, particularly in conjunction with issues of privacy and data protection, and would benefit from further research. Having refined our terms, it is possible to examine the various identity authentication systems available and their respective merits and demerits. 4. Identity Authentication Systems Authentication 'is the most essential of all the security services because reliable authentication is needed to enforce access control, to determine who is authorised to receive or modify information, to enforce accountability, and to achieve non-repudiation.' ( Ford and Baum (1997) , p.126.) There are five methods of authenticating an identity principal: 1. Something the claimant knows; Depending exclusively on any of methods 1-4 is generally inadequate and multi-token authentication systems are the norm. For example, bank ATM systems use a combination of methods 1 and 2 in the form of passwords (PINs) and bankcards. In addition, there are two contexts in which authentication can take place: entity authentication; and data origin authentication. Entity authentication may be either unilateral or mutual. Mutual authentication is preferable in every instance to avoid the potential for 'spoofing' attacks. This paper is concerned exclusively with entity authentication techniques used to facilitate electronic transactions involving individuals and/or organisations and, as such, will not consider authentication protocols used in secure communications systems, or in secure computer operating systems. The different forms of identity tokens used to authenticate identity will now be examined. 4.1 Passwords, Physical Tokens and Biometrics Passwords (in combination with user names) have been the mainstay of identity authentication systems since multi-user information systems came into being. Unfortunately, they are also one of the least secure techniques available and 'constitute one of the major vulnerabilities of electronic commerce systems.' ( Ford and Baum (1997) , p.127.) Threats to password-based authentication are:
Password authentication systems are so notoriously insecure that they are nearly always combined with other methods of identity authentication. Physical tokens are frequently used to enhance the security of identity authentication systems. Physical storage tokens are used by banks to corroborate an account number (held on a magnetic stripe card) with a password (PIN). More recently, tamper-proof integrated circuit cards (ICCs) allow the physical token to interact dynamically with the verification hardware. Examples of this interaction include:
Using these ICC tokens affords greater protection against reverse-engineering, forgeries and replay attacks to which stripe cards are susceptible.[ 3 ] Biometric techniques include fingerprint recognition, retinal scanning, hand-geometry scanning, and handwriting or voice recognition. These techniques are all, currently, extremely expensive to implement effectively and are therefore only worth considering in big-budget, high-security applications. As the drawbacks (including high cost, poor ergonomics, reliability, speed, and data storage requirements) are mitigated by improvements in technology, biometric techniques could emerge as the most secure method of automated identity authentication.[ 4 ] Strong asymmetric cryptography combined with secure hash functions allows the creation of a digital signature. Digital signatures verify that the sender of a message is in possession of a unique 'private' key and also verify that the message has not been altered in transit. An examination of the technical details of this procedure is beyond the scope of this paper.[ 5 ] 4.3 Certification Authorities and Digital Certificates As stated above, digital signatures verify that the sender of a message is in possession of a unique private key. In order to verify a message signed with a digital signature, the recipient needs a copy of the sender's public key. If the recipient knows the sender personally and can meet with them to exchange public keys, then a secure identity authentication system can be established. Electronic transactions between people or organisations in close physical proximity, with prior identity-trust relationships, are in the minority however, as has been discussed. It is therefore necessary to trust someone to provide recipients with public keys and to guarantee the association of the public key with some 'chosen attribute'. This function is performed by a Certification Authority (CA) which issues digital certificates attesting to the connection between attribute and public key. Thus, 'a public key is a surrogate presence in cyberspace for some entity in physical space. It acts directly in cyberspace, just as the associated entity can act in physical space.' ( Ellison (1996) )
Figure 1: Generating a digital certificate [ 6 ] Certification authorities eliminate the problems of scalability and physical separation associated with digital signatures, and introduce the problem of accepting a third party as suitably trustworthy. Several organisations have established themselves as CAs and are now beginning to issue digital certificates. Their procedures for establishing the 'physical-space' identity of an entity are examined below, after some comments on the importance of private key protection. Assuming that the person wishing to authenticate a cyberspace identity trusts his CA and assuming that the CA is operating securely and correctly, the only problem with public key infrastructure (PKI) based systems arises when the private key of a certified private-public key-pair becomes compromised. Private keys must be generated randomly, and stored and transmitted securely in order for the trust placed in the associated digital certificate to be deserved. Once again, the general point is made to draw the reader's attention to an area of interest, without going into lengthy technical detail about the merits of the various solutions available. It is important to be aware that, currently, a password (or pass-phrase) is generally used to protect the private keys associated with web browser certificates (used for sending digitally signed e-mail and engaging in Internet-based electronic transactions). This is like buying the most secure door lock on the market and leaving the key under the doormat. The inherent weaknesses of password-based access control systems are not mitigated by utilising them to control access to more sophisticated identity authentication systems. On the contrary, the strength of the PKI fosters an entirely false sense of security when this method of private key protection is adopted. Storing the private key on a physical token (an ICC card or a PCMCIA card) is an alternative protection method, and, when combined with a password, is a reasonably secure one. A secure biometric system would provide the maximum private key protection, at a price. 5. Identity Authentication Procedures
It is an unproven matter of faith among computer security specialists that private keys and digital certificates can be used to establish identity. ( Garfinkel and Spafford (1997) , p.126.) As has been previously stated, digital certificate verification does not prove that the certificate owner possesses the attribute attested to by the certificate. It only proves that they are in possession of a secret cryptographic key that has been signed by an appropriate CA. Michael Baum, who works for VeriSign, the largest and most experienced CA in business, describes digital certificates as providing 'probative evidence' useful in making a determination of identity in court. However, as previously stated, this facility depends upon:
Nobody is suggesting that CA-backed digital certificates are a final solution to the problems of identity authentication, simply that they are significantly more secure than earlier techniques. Despite their observations above, Garfinkel & Spafford (1997) accept that, 'digital certificates are a substantially more secure way of having people identify themselves on the Internet than the alternative: usernames and passwords.' ( Garfinkel and Spafford (1997) , p.133.) With these comments in mind, let us turn to an examination of the procedures used by CAs when issuing digital certificates. 5.1 Types of Digital Certificate There are basically 4 types of digital certificate:[ 7 ]
The following analysis will focus on the authentication procedures adopted by CAs to verify the identity of an entity prior to issuing a personal certificate. 5.2 Certification Authority Procedures As a component of the operating procedures of a CA, identification and authentication is divisible into four sections:
Sub-components of the initial registration section (which will include all the sub-components of any other section) are as follows:
Ideally, the taxonomy below would present an entry under each of these sub-component headings, for each CA. As is mentioned in the introduction to the taxonomy, the paucity of detailed information currently made available by commercial CAs makes this impossible. We will now focus on the authentication requirements for establishing organisational and individual identity as mentioned above. On-line procedures involve transmission of identity-related information across the same network for which a secure form of identity authentication is being sought. These procedures include:
Given that the CA needs assurance that both the public key value and the subscriber information originate from the named subscriber and have not been tampered with during transmission, the CA may verify submitted information with an on-line third party database (typically a consumer credit database, like Equifax in the USA). There still remain severe 'limitations to the assurances available using a purely online registration system' however. ( Ford and Baum (1997) , p.208.) Transferring some information through traditional off-line channels helps to mitigate these limitations. Off-line procedures allow real-world verification of subject identity attributes to take place. These include:
Identity authentication procedures, which employ both on-line and off-line techniques, are relatively trustworthy and form the basis of current commercial CA high-assurance certificate registration practice. However, off-line procedures create problems for CAs desiring to support a large, geographically disperse subscriber population. Solutions to these problems involve establishing local registration authorities (LRAs) or converting existing organisations with the necessary office infrastructure already in place (e.g. banks or post offices) into CAs.[ 15 ] 5.5 Local Registration Authorities LRAs are organisations to which the CA delegates responsibility for authenticating the identity attributes of some real-world entity. LRAs provide local support to a set of CA subscribers. Examples of LRAs include:
LRA functions might include:
Organisations wishing to issue digital certificates to their employees may establish themselves as LRAs related to an independent commercial CA. Large organisations may wish to establish themselves as fully functioning CAs with control over the technical and operational procedures, and several commercial organisations are now marketing business solutions to this end.[ 16 ] Note that, regardless of who is issuing the certificates, unless subject authentication and private key protection procedures are rigorously defined and maintained, digital certificates will not be worth the silicon they are stored on. 'The extent to which identity must be confirmed is dependent upon the level of assurances that a particular class or type of certificate purports to provide – this is typically expressed within the applicable certification practice statement or user agreement.' ( Ford and Baum (1997) , pp.208-209.) As will become apparent from the taxonomy below, commercial CAs tend to issue different classes of certificate to cater for different entities and for different assurance requirements. Certificate classes differ with regard to:
A simple example of how some of these factors might be defined is given below: Table 1: U.S. PKI Technical Working Group Certificate Assurance Levels This paper is principally concerned with the identity authentication requirements of digital certificates. How do these requirements differ by the class of certificate involved? An attempt to create a specification for certificate-class identity authentication requirements is presented below. (IA stands for Issuing Authority meaning the organisation responsible for issuing the certificates, i.e. not an LRA.) Table 2: Required Certificate Application Information To date, CAs have principally concerned themselves with issuing identity certificates. There is a need to develop procedures allowing CAs to issue certificates attesting to other attributes of a real-world entity (like age, sex, authority, privileges, etc.). The current paradigm for digital certificates is the X.509v3 standard (see appendix) which does not permit selective disclosure. This means that entities wishing to present digital credentials attesting to a particular attribute must reveal all other attributes contained in their digital certificate. Thus, digital certificates could provide unwelcome opportunities for data surveillance. Some solutions to the problem of attribute certification will be analysed in the final section, which follows the taxonomy of commercial CA certificates below. The following taxonomy of commercial CA certificates, with regard to their identity authentication procedures, is not comprehensive. Every effort has been taken to ensure the accuracy of the information presented. Many certification authorities have established web-sites advertising their products, but provide scant detail about their operating procedures. Where information was unavailable at the time of writing, the relevant entry has been marked '?'. Note that some CAs use extremely vague terms (e.g. 'moderate ID authentication') which are essentially meaningless. Ambiguity is also present in the use of the term 'e-commerce' which VeriSign seem to understand to mean high-risk inter-organisational transactions, while other CAs imply lower-risk individual transactions. Table 3: A taxonomy of commercially-available digital certificates with regard to their identity authentication procedures [ 17 ] The notion of authority authentication is becoming increasingly important for practitioners and researchers in the field of digital certification. While the fundamental purpose of public-key certificates is to bind a public-key to some entity by means of unique identifiers included in the certificate, certificate verifiers may require other information before the signature can be trusted. For example, does the certificate holder have authority to:
Authority authentication 'represents an important area of electronic commerce technology which is yet to be fully explored or developed,' and as such will be briefly dwelt upon in this paper. ( Ford and Baum (1997) , p.253.) The X.509v3 standard for digital certificates certifying public-keys could be used to concurrently certify authorisation information. However, it is to be expected that the authority most suitable for issuing public-key digital certificates (namely, the commercial CAs currently in existence) will not be the authority most suitable for issuing certificates of authority. Therefore, identifying appropriate authorities is an important first step in this process. Another problem with tying authority information to public-key certificates is apparent when one considers that the time-period validity of public-key certificates may not match with the required time-period validity of an authority certificate. Attribute certificates have been developed by the US financial industry's ANSI X9 committee in response to these problems with public-key certificates. Anyone can define and register attribute types and an attribute authority digitally signs certificates. This standard also permits attaching attributes to documents, which can then be transmitted along with the signature. If the attributes of the document require a level of authority not attested to by the attributes of the signature, the verifier can deduce a lack of proper authorisation.[ 18 ] Finally, another possible solution to the problems encountered when linking identity authentication with authority authentication is the concept of a role register. Role registers provide a means of mapping identities to authorities or privileges. Keeping the identity authentication process separate from the record of associated privileges enhances both flexibility and security. Securely authenticating the identities and privileges of cyberspace entities is essential if the widely predicted explosion in the growth of e-commerce is to become a reality. Digital certificates issued and managed by certification authorities potentially provide a scaleable, flexible, and, above all, secure system by which to implement identity authentication procedures. It is critically important to realise that private-key protection and rigorous, real-world identity authentication procedures are the bedrock of this system. Therefore, information systems security practitioners need to be clear about what attributes of an entity they are concerned with authenticating. They need to be clear about what a particular digital certificate is actually attesting to, and, most importantly, they need to satisfy themselves that the certification authority procedures are adequate for the level of assurance they desire. Information detailing the procedures employed by individual CAs to authenticate entity attributes (in the form of Certification Practice Statements) is notable by its absence from many CA web-sites. This indicates the embryonic state of the technology and, perhaps, the difficulties that have to be surmounted to define workable, secure solutions. Garfinkel & Spafford (1997) are right to burst the bubble by stating: 'The people who are actually setting up [public key infrastructure] systems ... don't want a system that is perfect, just one that is better than today's paper-based identification systems. Unfortunately, it's not clear whether public-key technology even gives that kind of assurance about identity.' ( Garfinkel and Spafford (1997) , p.126.) Only if these techniques and procedures are implemented on a wide-scale will their practicability and reliability be tested. The same qualification holds true for more recently developed attribute certificates, which rely on the same technological principles. Biometric identity authentication systems combined with rigorous operating procedures have the potential to provide maximally secure identity authentication for electronic transactions. They are unlikely to be widely implemented for reasons of cost, data storage, processing time, and ergonomics. There are also serious ethical issues associated with biometric identification systems. ( Greenleaf and Clarke (1997) .)
Figure 2: X.509 Version 3 Digital Certificate References Ellison C (1996) Establishing Identity Without Certification Policies < www.clark.net/pub/cme/usenix.html > Ford W & Baum M S (1997) Secure Electronic Commerce: Building the infrastructure for digital signatures and encryption (Prentice Hall, Inc., New Jersey) Garfinkel S & Spafford G (1997) Web Security & Commerce (O'Reilly & Associates, Inc., Cambridge, USA) Grant G L (1998) Understanding Digital Signatures: Establishing trust over the Internet and other networks (McGraw-Hill, Inc., New York) Greenleaf G & Clarke R (1997) Privacy Implications of Digital Signatures Invited Address, IBC Conference on Digital Signatures, Sydney, 12 March 1997, < http://www.anu.edu.au/people/Roger.Clarke/DV/DigSig.html > BOOKS, JOURNAL ARTICLES & INTERNET WHITE PAPERS Anderson R & Kuhn M (1996) Tamper Resistance – a Cautionary Note The Second USENIX Workshop on Electronic Commerce Proceedings, Oakland, California, November 18-21, 1996, pp 1-11 < http://www.cl.cam.ac.uk/~mgk25/tamper.html > Chokhani S & Ford W (1996) Certificate Policy and Certification Practice Statement Framework (Draft) NIST, < http://csrc.ncsl.nist.gov/pki/docs/fmk03nov.doc > Computer Security Technologies (1996) Certification Policies in the COST Certification Infrastructure Sweden < www.cost.se/cost_pol.htm > Electronic Commerce Promotion Council of Japan (ECOM) (1997) Certification Authority Guidelines (Alpha version) < www.ecom.or.jp/eng/output/ca/eng-guideline.htm > Ford W & Chokhani S (1997) Internet Public Key Infrastructure Part IV: Certificate Policy and Certification Practices Framework Internet Engineering Task Force, <ftp://ietf.org/internet-drafts/draft-ietf-pkix-ipki-part4-02.txt> [no longer available] Ford W & Leech M (1996) Public Key Infrastructure Certificate Policy and Certification Practice Statement Framework Version 1.2b Policy Management Authority Committee of the Government of Canada Public Key Infrastructure Garfinkel S & Spafford G (1996) Practical Unix and Internet Security, Second Edition (O'Reilly & Associates, Inc., Cambridge, USA) KeyPOST (1997) Certification Practice Statement v.1.1 , Australian Postal Corporation, < www.auspost.com.au/keypost > Kim H J (1995) Biometrics – Is it a viable proposition for identity authentication and access-control Computers & Security, Vol. 14, No. 3, pp.205-214 VeriSign (1997) VeriSign Certification Practice Statement version 1.2 VeriSign, Inc., California, < www.verisign.com/repository/CPS1.2/intro.html > COMMERCIAL CERTIFICATION AUTHORITY WEB-SITES Baltimore Technologies, Eire www.baltimore.ie Footnotes 1 A claimant might be an individual, an organisation, a piece of hardware or software. 2 For an entertaining and true account which illustrates each one of these threats (with the possible exception of replay attacks), see Clifford Stoll (1991) The Cuckoo's Egg (Macmillan Publishers Ltd., London). 3 See Anderson & Kuhn (1996) for details of the limitations of current 'tamper-proof' technologies. 4 American company Digital Persona have recently released a new fingerprint recognition package which, they claim, is cheap, easy to use and reliable. See also Kim H J (1995) for an analysis of biometrics' viability. 5 For an excellent primer on every aspect of modern cryptographic techniques see Schneier B (1995) Applied Cryptography (2nd Edition, Wiley, New York). 6 Based on a diagram in Grant (1998), p.37. 7 Certificates currently issued by commercial CAs are based on the X.509v3 standard detailed in the appendix. Other forms of certificate are being developed to improve flexibility and security. 8 VeriSign uses the VeriSign Public Primary Certification Authority (PCA) to sign its CA public keys. 9 See http://www.commerce.net/news/press/pilot.html for news of the world's first cross-certification pilot between JapanNet and CommerceNet. 10 This division and the following sub-components are drawn from the Internet Engineering Task Force Public Key Infrastructure Working Group paper on certificate policy and certification practices (Ford & Chokhani, 1997). 11 Examples include X.500 distinguished name, Internet e-mail address, and URL. 12 The term `meaningful´ means that the name form has commonly understood semantics to determine identity of the person and/or organisation. Directory names may be more or less meaningful. 13 Examples of proof include the issuing CA generating the key, or requiring the subject to send an electronically signed request or to sign a challenge. 14 Ford & Baum (1997), p.210. Garfinkel & Spafford (1997) also highlight one of these risks when they draw particular attention to the relationship between the time-period validity of the documents used to validate an identity and the time-period validity given to the certificate based on those documents. 15 Postal services in Canada, Scandinavia, Australia, US and the UK have all realised this opportunity. 16 E.g. Cybertrust GTE (USA), Baltimore Technologies (Ireland), and others. 17 Source material for this table was obtained from the web-sites of the certification authorities. Internet addresses used are detailed in the end-matter. 18 See Ford & Baum (1997) for more details of this emerging standard and other authority authentication procedures like SDSI and SPKI. The author welcomes comments and/or questions and can be contacted at: pp9 MLB1 Tel: +44 1473 605354 |
