ON THE USE OF “QUALIFIED” DIGITAL SIGNATURES

Maurizio Adriano Strangio

Department of Mathematics, University of Rome “Roma Tre”, Italy

Keywords:

Qualiﬁed Digital Signatures, Key Agreement Protocols.

Abstract:

The European Commission Directive 1999/99/EC aims to provide a community-wide framework for the use

of electronic signatures and thus for promoting electronic trade and communication among the member states.

The directive introduces the notion of “qualiﬁed” digital certiﬁcates as a means to maintain legal effects of

digital data that are analogous to those of paper-based signatures. To this end, proofs of (physical) identity and

possession (of the private key) are fundamental requirements that must be fulﬁlled by the requester during the

public key enrollment process.

Digital signatures are often employed as secure buildings blocks in key agreement protocols that use public

key authentication. The need for the rigorous analysis of such protocols has recently emerged; there are

currently several formal models of distributed computing that may serve for this purpose. However, we point

out these models employ rather trivial or unpractical approaches in the modeling of the procedures and policies

employed by certiﬁcation authorities.

We believe that usage of qualiﬁed certiﬁcates not only should represent the standard practice for CAs in order

to sustain secure electronic commerce (and in general all forms of secure communication) but also represents

the ﬁrst step towards the domain of a global PKI.

1 INTRODUCTION

Symmetric key algorithms have a troublesome short-

coming — the need to generate and distribute a se-

cret key to the parties who wish to privately commu-

nicate. Furthermore, keys must be exchanged prior to

any communication by making use of a (out-of-band)

secret channel and not via a public network.

With public key cryptography, in a network of n

principals, each party needs only to distribute a public

key to her n − 1 peers so the total number of keys to

manage is O(n), rather than O(n

However, with asymmetric key algorithms public

keys can be easily exposed (since they are delivered

in unencrypted form) therefore a principal using the

wrong key may be totally unaware that a malicious

party is posing as her intended peer. As the result,

public keys must be certiﬁed in a separate process

prior to any communication, possibly by a trusted

third party; for the parties willing to communicate

this is logically equivalent to having an authenticated

channel available to distribute the public keys in order

to receive assurance of each others purported identity.

In practice, this service is provided by a Public

Key Infrastructure (PKI) which is an arrangement that

entails a trusted third party (a.k.a. certiﬁcation author-

ity — CA) fulﬁlling a notary function by binding the

public key (and a set of informative attributes) to the

owner of the corresponding private key. The outcome

of the process is a digital certiﬁcate (e.g. X.509v3),

signed by the authority itself, that provides undis-

putable evidence of the true identity of the principal

holding the private key/public key pair.

While this approach is technically straightfor-

ward, the management (e.g. operation and funding)

of a PKI can become quite intricate (see (Ellison and

Schneier, 2000) for an overview of the difﬁculties that

may be encountered) so that the public key validation

process may be almost as difﬁcult as distributing se-

cret keys (after all symmetric key cryptography has

been proﬁtably employed since the era of Ceasar).

Whether a CA may be regarded as trustworthy

(which ultimately is a decision of the relying party)

is a major concern and is not just a matter of con-

sidering certain distinguishing elements such as the

reputation or the nationality of the CA. It turns out

that there are other requirements that must be satis-

ﬁed and often these are not clearly stated in the CA

334

Adriano Strangio M. (2008).

ON THE USE OF “QUALIFIED” DIGITAL SIGNATURES.

In Proceedings of the International Conference on e-Business, pages 334-338

DOI: 10.5220/0001909203340338

 SciTePress

policies (e.g. in the Certiﬁcation Practice Statements

— CPS) or, even worse, not even completely under-

stood. For this reason, in some countries (e.g. Italy)

there are government agencies entitled to perform the

ﬁnal accreditation of CAs (which are also subjected

to periodic inspections to verify full compliance with

the law).

In fact, the main thesis of this paper is

that a “physical proof-of-identity” and “proof-of-

possession” are both mandatory to maintain the legal

status of digital signatures. With a physical proof-

of-identity the certiﬁcate requester must provide cor-

roborative evidence of her identity (e.g. an id-card)

in a “face-to-face” registration procedure at the reg-

istration authority (RA). With a proof-of-possession

the subscriber must prove he exerts (exclusive) con-

trol over the private key corresponding to the public

key undergoing the certiﬁcation.

Although the “physical” identiﬁcation process is

often perceived as an unduly burden by most sub-

scribers (although it is essentially identical to apply-

ing for a driving license) it is beneﬁcial for a number

of reasons:

- it discourages from making use of PKIs for crim-

inal activity;

- facilitates law enforcement agencies in the

prosecution of the crimes related to signa-

ture/encryption schemes (e.g. identity theft,

fraud);

- is also valuable for assessing the trustworthiness

of individuals.

On the one hand, the proof-of-possession require-

ment has been the subject of many discussions (see

(Asokan et al., 2003; Lauter and Mityagin, 2006))

with the majority of PKI standards ﬁrmly encourag-

ing its use (observe that when the CA generates the

public/private key pairs this requirement is immedi-

ately fulﬁlled). On the other hand, many CAs will

issue certiﬁcates offering different levels of identity

assurance; a physical proof-of-identity is required for

the certiﬁcates that are used in applications that need

to establish the purported identity of the subscriber

with certainty (e.g. commercial transactions); this is

opposed to certiﬁcates issued with email addresses

used as identiﬁcation credentials (which are suitable

for non critical applications).

However, often an explicit indication of how the

subscriber was identiﬁed is omitted from the certiﬁ-

cate attributes thus opening subtle vulnerabilities in

its use (the relying party may unwillingly omit to per-

form any validity checks).

We stress that a proof-of-possession merely im-

plies that the applicant has access to the private key

corresponding to the target public key; therefore a

physical proof-of-identity is essential for the legal

recognition of digital signatures.

CA policies often refer to the PKCS#10 standard

(PCKS#10v1.7, 2000) which describes a syntax for

certiﬁcation requests; an entity applying for a cer-

tiﬁcate may be prompted by the RA to digitally sign

an electronic “certiﬁcate request” that will be subse-

quently conveyed to the CA for the ﬁnal certiﬁcate

handout. Such an arrangement subsumes a proof-of-

possession. Surprisingly, the standard also includes

the provision for requests using paper forms; this im-

plies that CAs should deﬁne alternative means in their

policies for the requester to prove knowledge of the

private key.

2 DIGITAL CERTIFICATES IN

THE EUROPEAN COMMUNITY

The European Commission Directive 1999/99/EC

(European-Parliament, 1999) aims to provide a com-

munity framework for the use of electronic signatures

and thus for promoting electronic trade and commu-

nication among the member states.

In particular, the legal recognition of electronic

signatures entails the use of advanced electronic

signatures based on “qualiﬁed certiﬁcates”, which

are created by secure (signature-creation) devices

(e.g. tamper-proof standard-compliant devices such

as smart cards holding the private key), to ensure they:

(a) satisfy the same legal requirements of hand-

written signatures on paper-based data (refer to

the deﬁnition of advanced electronic signature in

the Directive);

(b) are admissible as evidence in a court of law.

The majority of European member countries have al-

ready embodied the directive in their legal systems

(civil laws).

Qualiﬁed certiﬁcates are certiﬁcates that com-

ply with the rules contained in Annexes I and II of

the aforementioned Directive and appear to satisfy

both the proof-of-possession (Annex I, letter (e)) and

proof-of-identity (Annex I, letter (c)) requirements as

we discussed in the preceding section.

Under such arrangements we call the resulting

certiﬁcates “strong qualiﬁed certiﬁcates”; as opposed

to “weak qualiﬁed certiﬁcates” which may be sup-

plied by CAs that do not strictly comply with both

the above requirements (e.g., consider a CA perform-

ing a proof-of-identity by means of an on-line zero-

knowledge identiﬁcation protocol).

ON THE USE OF “QUALIFIED” DIGITAL SIGNATURES

335

3 KEY ESTABLISHMENT

PROTOCOLS THAT USE

DIGITAL SIGNATURES

Digital signature schemes (and other public key en-

cryption schemes) are often employed as secure

building blocks in key establishment protocols that

use public key authentication for the purpose of en-

suring the parties involved in the communication that

the session key was established with the intended peer

and not with an impostor. Once a session key is avail-

able, the data exchanged is authenticated and undis-

closed against third parties thus allowing secure com-

munications over an insecure network.

Traditionally, cryptographic protocol security has

been a matter of perceived conﬁdence supported by

heuristic proof arguments and by the protocol surviv-

ing many years of public scrutiny. More recently, the

approach has changed and the priority has switched

to developing formal proof frameworks. In this con-

text, a main line of research employs complexity-

theoretic models for distributed network computing

and is dedicated to key establishment protocols (Bel-

lare and Rogaway, 1993; Blake-Wilson and Menezes,

1998; Shoup, 1999; Bellare et al., 2000; Canetti and

Krawczyk, 2001; LaMacchia et al., 2006; Difﬁe et al.,

1992). Such models employ both private and public

key cryptographic techniques.

In the above models most of the ﬁne level details

of PKIs are abstracted away; while in principle this

is a reasonable approach (we already mentioned that

PKIs are an intricate subject on their own) we often

see that important topics such as the key registration

procedures and policies employed by CAs are disre-

garded by many authors. We believe that such issues

are not only tightly related to the correct operation

of the protocol but may also eventually lead to (legal)

disputes among the participants and therefore must be

appropriately settled.

As a starting point for our discussion, let us recall

three public key registration procedures commonly

encountered in the literature (LaMacchia et al., 2006):

a. Honest key registration. All parties (including

those controlled by the adversary) follow the key

generation procedures honestly and register the

resulting public keys before engaging in any com-

munications. The adversary can corrupt parties

only after key registration has completed;

b. Proof-of-possession. An authority performs some

validity check upon public key registration. In

particular, a party is required to prove knowledge

of the corresponding secret key. The adversary

can register public keys for corrupted parties at

any time;

c. Arbitrary key registration. Parties can register ar-

bitrary public keys (even the same key as some

other party) without any validity checks. The ad-

versary can register public key for corrupted par-

ties at any time.

We comment that case c. is unrealistic since it is ex-

tremely unlikely that a CA will ever accept to enroll

public keys for which the requester has not provided

any validity check; the minimum requirement is veri-

fying the identity of the applicant.

Item a. exempliﬁes the behavior one would nor-

mally expect from honest principals holding valid cer-

tiﬁcates. The case of interest here is the compromise

of the private key of a principal whether or not he is

aware of this fact (the adversary may be able to subtly

obtain a copy of the private key).

The proof-of-possession requirement invoked by

item b. is not sufﬁcient to achieve adequate levels

of security (as discussed above); we have already

pointed out that without a proof-of-identity CAs will

not issue certiﬁcates. We nowelaborate further on this

point.

Lauter and Mityagin (Lauter and Mityagin, 2006)

have recently presented protocol KEA+ that, as op-

posed to the original version (KEA), is resilient to un-

known key share (UKS) attacks; as a countermeasure,

they suggest to include the identities of the protocol

participants as arguments of the key derivation func-

tion (KDF) and also claim that the above countermea-

sure avoids the need for a proof-of-possession.

Recall that a UKS attack involves a (man-in-the-

middle) adversary, posing as a legitimate party (say

C) in a protocol run between honest parties (say A

and B), that is able to convince one party (e.g. A) to

accept her identity (C) while the peer (B) is unaware

of this fact (i.e. B thinks he’s interacting with A). This

vulnerability is regarded of interest whether or not the

adversary is able to have the attacked party accept a

chosen session key.

We believe there are two issues that are over-

looked in their arguments. Firstly, the adversary must

provide a valid proof-of-identity;therefore, unless she

is willing to reveal her true identity (otherwise she

must be able to perfectly disguise herself and also of-

fer a counterfeit id card — indeed a non trivial task),

she may be liable of being legally pursued for her ac-

tions in a court of law. Observe also that the (physical)

proof-of-identity policy rules out the well known on-

line UKS attack described by Kaliski (Kaliski, 2001)

against the MQV protocol (Law et al., 2003) (notice

that in this attack the adversary is able to have the tar-

get party accept a chosen session key).

Secondly, the lack of a proof-of-possession (albeit

ICE-B 2008 - International Conference on e-Business

336

not being necessary for a protocol to resist UKS at-

tacks) implies loss of the non-repudiation property for

digital signatures; at a later time the signer can even-

tually deny having participated in the protocol run

(and therefore having established a particular session

key with a peer).

We may conclude that the KEA+ protocol (and

many others alike) are not suitable for electronic trade

and commerce and at best can be used within the

realm of a corporate domain for the exchange of data

between employees (the proof-of-possession would

be satisﬁed anyway since the keys are generated and

managed in house).

From the above remarks we see that the primary

sources of concern derive from the legal implica-

tions involved in electronic data processing facilities

(which constitute the underlying principles of the Eu-

ropean Directive).

The lessons learned are twofold:

- strong qualiﬁed certiﬁcates are necessary to pre-

vent legal disputes;

- if legality is not a concern peers may use weak

qualiﬁed certiﬁcates.

Therefore, relying parties must be able to verify cer-

tiﬁcate policies of the CA to determine whether the

certiﬁcates are suitable and trustworthy for a particu-

lar application.

As a consequence, CAs should also have clearly

stated policies for appropriately setting certiﬁcate at-

tributes; for example, in strong qualiﬁed certiﬁcates

the X.509v3 Key Usage extension (RFC3280, 2000)

should always specify only one use for the public key

(namely for signature validation) and have the non re-

pudiation bit asserted.

We mention the recent attempt of Boldyreva et al

(Boldyreva et al., 2007) to set out rigorous models of

the public key enrollment process with a CA. In par-

ticular, they show that a simple challenge-response

interactive protocol sufﬁces to fulﬁll the proof-of-

possession requirement. Notice that a fundamental

assumption in their work is the existence of an authen-

tic channel between the user and the CA; although (as

the authors admit) it would be otherwise difﬁcult to

establish any reasonable security claim this hypothe-

sis enormously reduces the practical usefulness of the

registration protocols (e.g. running it on an open net-

work environment is risky — similar arguments apply

to the proof-of-possessionprotocolsfound in the tech-

nical literature for the Internet (RFC2875, 2000)). As

a possible enhancement to the registration protocols,

one may consider the CA assigning a unique trans-

action key TK to the subscriber when the physical

proof-of-identity is performed; the TK may be used

only once either as a shared session key (provided it

is cryptographicallystrong enough — say, at least 128

bits) or as an access token to be included as an argu-

ment of the registration protocol on the user side (and

veriﬁed by the CA) thus providing an authenticated

communication channel.

4 CONCLUSIONS

In this paper we have (re)afﬁrmed the need for CAs

to require mandatory proofs of identity and posses-

sion (of the private key) from subscribers requesting

digital certiﬁcates for public keys in order to preserve

the legal status of the associated signature schemes.

We have also reviewed the notion of qualiﬁed certiﬁ-

cates as introduced by the European community di-

rective 1999/99/EC (European-Parliament, 1999). As

far as electronic documents are concerned (e.g. con-

tracts) the directive says that all member states should

ensure that advanced electronic signatures (i.e. signa-

tures that are linked to qualiﬁed certiﬁcates) must be

given the same legal effect as paper-based signatures

and are thus form valid evidence in court trials.

We have also pointed out that formal models of

distributed computing, where the security of key es-

tablishment protocols using public key authentica-

tion is evaluated, often adopt a rather trivial approach

when considering public key enrollment procedures.

Our main thesis is that large-scale CAs (offering

national or international cross-borderservices) should

only offer qualiﬁed certiﬁcates (with the appropriate

proofs established) whether they are relative to digi-

tal signatures used to sign documents or employed as

building blocks in key establishment or other crypto-

graphic protocols. Other types of certiﬁcates, which

we have broadly classiﬁed as weak qualiﬁed certiﬁ-

cates, only grant a limited degree of security (assur-

ance) with no legal effects and can be eventually used

in particular domains (e.g. the PKI used in a home-

banking system).

Although (in principle) one could choose the

appropriate certiﬁcate depending on the application

(among those offered by a CA) there would be signif-

icant beneﬁts if all the players involved could assume

that certiﬁcates delivered an equivalent degree of as-

surance rather than having to rely on non-standard

policies. To this end, it would be desirable that strong

qualiﬁed certiﬁcates be adopted as a universal stan-

dard for digital signatures, thus delivering further im-

pulse to the development of interoperable world-wide

PKIs.

ON THE USE OF “QUALIFIED” DIGITAL SIGNATURES

337

ACKNOWLEDGEMENTS

The author is grateful to the reviewers for their helpful

comments and suggestions.

REFERENCES

Asokan, N., Niemi, V., and Laitinen, P. (2003). On the Use-

fulness of Proof-Of-Possession. Proceedings of the

2nd Annual PKI Research Workshop, pages 122–127.

Bellare, M., Pointcheval, D., and Rogaway, P. (2000). Au-

thenticated Key Exchange Secure Against Dictionary

Attack. In Proceedings of EUROCRYPT 2000, LNCS

1807:139–155.

Bellare, M. and Rogaway, P. (1993). Entity Authentica-

tion and Key Distribution. In Proceedings of CRYPTO

1993, LNCS 773:232–249.

Blake-Wilson, S. and Menezes, A. (1998). Entity authen-

tication and Authenticated Key Transport Protocols

Employing Asymmetric Techniques. Security Proto-

cols - 5th International Workshop, LNCS 1361:137–

158.

Boldyreva, A., Fischlin, M., Palacio, A., and Warinschi,

B. (2007). A Closer Look at PKI: Security and Ef-

ﬁciency. Proceedings of PKC 2007, LNCS 4450.

Canetti, R. and Krawczyk, H. (2001). Analysis of Key Ex-

change Protocols and Their Use for Building Secure

Channels. Advances in Cryptology-EUROCRYPT

2001, LNCS 2045:453–474.

Difﬁe, W., van Oorschot, P., and Wiener, M. (1992). Au-

thentication and Authenticated Key Exchange. De-

signs, Codes and Cryptography, 2:107–125.

Ellison, G. and Schneier, B. (2000). Ten risks of PKI: What

you’re not being told about Public Key Infrastructure.

Computer Security Journal, 16(1):1–7.

European-Parliament (1999). Directive 1999/99/EC of the

European Parliament and of the Council of 13 Decem-

ber 1999 on a Community Framework for Electronic

Signatures.

Kaliski, B. (2001). An Unknown Key Share Attack on the

MQV Key Agreement Protocol. ACM Transactions

on Information and System Security, pages 36–49.

LaMacchia, B., Lauter, K., and Mityagin, A. (2006).

Stronger Ssecurity of Authenticated Key Exchange.

http://eprint.iacr.org/2006/073.

Lauter, K. and Mityagin, A. (2006). Security Analysis of

KEA Authenticated Key Exchange Protocol. Pro-

ceedings of PKC’06, LNCS 3958:378–394.

Law, L., Menezes, A., Qu, M., Solinas, J., and Vanstone,

S. (2003). An Efﬁcient Protocol for Authenticated

Key Agreement. Designs, Codes and Cryptography,

28:119–134.

PCKS#10v1.7 (2000). Certiﬁcate Request Syntax Standard.

RSA Laboratories.

RFC2875 (2000). Difﬁe-Hellman Proof-of-Possession Al-

gorithms. Network Working Group.

RFC3280 (2000). Internet X.509 Public Key Infrastructure

Certiﬁcate and Certiﬁcate Revocation List CRL Pro-

ﬁle. Network Working Group.

Shoup, V. (1999). On Formal Models for Secure Key Ex-

change. Technical Report RZ 3120, IBM Research.

ICE-B 2008 - International Conference on e-Business

338