ON THE NEED TO DIVIDE THE SIGNATURE CREATION
ENVIRONMENT
Jorge L. Hernandez-Ardieta, Ana I. Gonzalez-Tablas, Benjamin Ramos and Arturo Ribagorda
SeTI Research Group, Department of Computer Science, University Carlos III of Madrid
Av. Universidad 30, 28911 Leganes (Madrid), Spain
Keywords:
Non-repudiation, Digital evidence, Electronic signature, Vulnerabilities.
Abstract:
Electronic signatures have been legally recognized as the key element for boosting e-commerce under secure
conditions. Several legislations throughout the world establish electronic signatures as legally equivalent to
hand-written signatures, assigning them the property of evidence in legal proceedings. In addition, interna-
tional standards define electronic signatures as non-repudiation evidence respecting the signed information.
Bearing this in mind, it is obvious that the reliability of electronic signatures is paramount. However, the re-
sults show that several attacks on signature creation environments are feasible and easy to perform. As a result,
the reliability of evidence is drastically undermined. We claim that the division of the environment becomes
the most effective solution to counteract current threats. The formal proofs that support this statement are
given along with an overview of the legal background and a summary of main potential threats on signature
creation environments.
1 INTRODUCTION
Many efforts have been made to boost e-commerce,
especially those related to the enhancement of e-
commerce security. Maybe the most remarkable one
has been the support given to the electronic signa-
ture (e-signature) by Governments and the IT Indus-
try. An e-signature is generally considered as data in
electronic form which are attached to or logically as-
sociated with other electronic data and which serve
as a method of authentication (European Directive,
1999). Furthermore, several legislations throughout
the world set the e-signature as legally equivalent to
hand-written signatures, as well as assign them the
property of evidence in legal proceedings (European
Directive, 1999; Federal Trade Commission, 2001;
Government of Canada, 2000).
International standards also consider e-signatures
as non-repudiation evidence in electronic transac-
tions, either when applied to e-commerce or other
contexts (ISO/IEC DIS 13888-1, 1996). This scenario
clearly places the e-signature as a key technology,and
therefore its security is a critical issue which should
be carefully analyzed.
However, several results show that the security of
the vast majority of e-signature applications is not as
trustworthy as it should be. The feasibility of some
published attacks should make the reader consider
if the property of non-repudiation and legal equiva-
lence of e-signatures should be maintained under cur-
rent conditions. If no secure means can be provided
to the end user, a signatory should not be held li-
able for the commitment made in a signed message
(i.e. a purchase order, a contract, an email). The
signatory could allege that the corresponding private
key was compromised or the document she intended
to sign had been modified before generating the e-
signature. If the case went to court, the signatory
could easily proveon the balance of probabilities (in a
civil action) or beyond reasonable doubt (in a criminal
action) (McCullagh, 2000) that any potential attack
could have been carried out to successfully forge the
signature. However, the mere possibility of having to
deal with the consequences arisen from forged signa-
tures obviously undermines the user’s confidence in
technology, especially once legislations have legally
reversed the onus of proof for e-signatures. The proof
is now set on the signatory instead of on the verifier,
contrary to traditional commerce.
We claim that an untrustworthy environment can-
not generate reliable signatures, and thus cannot as-
sure the non-repudiation of evidence. As every new
proposal is always welcomed with a new attack, a
completely different approach must be taken in or-
375
L. Hernandez-Ardieta J., I. Gonzalez-Tablas A., Ramos B. and Ribagorda A. (2009).
ON THE NEED TO DIVIDE THE SIGNATURE CREATION ENVIRONMENT.
In Proceedings of the International Conference on Security and Cryptography, pages 375-380
DOI: 10.5220/0002224103750380
Copyright
c
SciTePress
der to enforce the reliability of evidence. We con-
sider that dividing the signature creation environment
can be the most effective solution to counteract cur-
rent threats. By imposing the usage of different en-
vironments to create the legally binding evidence, the
probability of a successful attack is highly decreased,
even when those environments are untrustworthy on
their own.
The article is organized as follows. The next sec-
tion provides an overview of the legal background
on e-signatures. Section 3 introduces the security
problem for PKI-based e-signatures. The proofs that
demonstrate the benefits of the division of the signa-
ture creation environment are given in section 4. We
conclude the article in section 5.
2 LEGAL BACKGROUND
The European Directive on a Community framework
for e-signatures (European Directive, 1999), the US
e-Sign Act (Federal Trade Commission, 2001), the
Canadian Personal Information Protection and Elec-
tronic Documents Act (Government of Canada, 2000)
and the UNCITRAL Model Law on Electronic Signa-
tures (United Nations, 2001) recognize electronic sig-
natures as the means for promoting e-commerce un-
der secure conditions.
After the approval of these laws, e-signatures are
regarded as legally equivalent to hand-written signa-
tures only if the requirements for hand-written sig-
natures are fulfilled. In a nutshell, these requirements
imply the signatory’sintention of signing and the abil-
ity to uniquely identify the signatory based on her sig-
nature mark, from which a reasonable difficulty to
forge the signature is derived. These requirements
are transposed in a technology-neutral viewpoint by
the laws. However, if we take a look at these re-
quirements, it will be easily noticed that, taking into
account the current state-of-technology, only digital
signatures based on Public Key Cryptography (PKC)
and a Public Key Infrastructure (PKI) satisfy them. In
fact, current legislations implicitly consider PKC and
PKI as the underlying technologies. For instance, the
UNCITRAL Model Law establishes PKI and PKC as
an example of implementing technologies for compli-
ant e-signatures.
On the other hand, it is obvious that sensitive in-
formation is normally exchanged in an Internet trans-
action between the participant parties. In order to pro-
tect the parties against the other’s misbehavior, digi-
tal evidence is generated during the transaction. Ev-
idence is information that, either by itself or when
used in conjunction with other information, is used
to establish proof about an event or action. Zhou
et al. have established certain conditions to be ful-
filled by digital evidence (Zhou, 1997). The ori-
gin and the integrity of the evidence must be veri-
fiable by a third party and the validity of the evi-
dence must be undeniable. When evidence is acting as
non-repudiation evidence, the party can not later suc-
cessfully repudiate having participated in the transac-
tion. ISO defines non-repudiation service as a ser-
vice that protects the parties involved in a transac-
tion against the other party denying that a particular
event or action took place. In particular, ISO speci-
fies non-repudiation services by means of asymmetric
cryptography (ISO/IEC 13888-3, 1997). Therefore,
PKC is considered again as the key technology for
creating non-repudiation evidence in the form of dig-
ital signatures. When we say digital signatures, it can
be automatically extended to the legally-supported e-
signatures.
Other important issues are the differences between
physical and digital realms. Although obvious, some-
times it seems that these differences have not been
taken into account when writing previous laws. Gov-
ernments throughout the world have directly trans-
posed the legal validity of physical evidence like a
hand-written signature to digital evidence like an e-
signature. In fact, the legal consequences are exactly
the same. However, the environment where the ev-
idence is produced is completely different, as well
as the threats that may arise. Furthermore, the envi-
ronment where non-repudiation evidence is generally
produced, the user’s Personal Computer, is almost al-
ways untrusted.
3 SECURITY PROBLEM
DEFINITION
Electronic signature technology used by end users is
prone to suffer from a wide variety of attacks. Some
of the most relevant threats that may subvert the secu-
rity of the signature creation process are briefly ref-
erenced in this section. The aim is to provide an
overview of practical and feasible attacks that can di-
minish the reliability of legally binding signatures.
3.1 Scenario
We consider a scenario where the user is in posses-
sion of a PKC key pair as well as a PKI digital cer-
tificate which bounds her identity with the public key.
The user can store her private key either in a software
keystore, like those managed by Web Browsers, or a
SECRYPT 2009 - International Conference on Security and Cryptography
376
hardware device, such as a smart card (named Signa-
ture Creation Device, SCDev). The application that
uses that private key for signing purposes is called
the Signature Creation Application (SCA). The Sig-
nature Creation Environment (SCE) is the user’s Per-
sonal Computer (PC), the most common environment
used nowadays to communicate and purchase through
the Internet.
The e-signature can be calculated over a local doc-
ument (i.e. Microsoft Word document, PDF docu-
ment, XML document, etc.), a Web content (i.e. data
of a web form) or any other kind of information (i.e.
local database information, raw data, etc.).
The context where the signature is created can be
an e-commerce transaction, a contract signing or any
other context which implies a legal commitmentmade
by the signatory on the signed information. There-
fore, the e-signature has some sort of legal effective-
ness which cannot be repudiated by the signatory.
In this scenario, two assumptions are made:
The user’s PC is untrusted. It is not possible to
obtain an certain level of assurance on the secu-
rity of the PC since the user normally has no tech-
nical knowledge, the applied protection measures
on the PC tend to be null (e.g. installation and
periodical update of anti-virus and anti-malware
programs, checking the SSL certificates when ac-
cessing to e-commerce Web sites, installation of
the operating system security patches, etc.) and,
although applied, these measures are not com-
pletely protective taking into account the number
of Trojan horses, viruses and any kind of malware
that can potentially infect the computer.
The attacker has knowledge, resources and mo-
tivation enough for successfully carrying out the
potential threats identified in section 3.2, provided
that they are technically feasible according to the
current state-of-technology.
3.2 Potential Threats
An attacker will always try to compromise the secu-
rity of a system by focusing on the weakest element.
If the attacker wants to subvert the security of an e-
signature process, like those developed to support e-
commerce transactions, the weakest point is the end
user. The reason is twofold: on one hand, the security
measures implemented in the end user’s PC are gen-
erally low compared to those implemented by server
systems. On the other hand, the user usually lacks of
any sort of security knowledge, what remarkably ag-
gravates the situation. Therefore, threats herein men-
tioned are only focused on undermining the security
of the end user’s PC where the signature creation pro-
cess is to be carried out.
In this sense, most attacks try to compromise the
cryptographic private key in order to generate forged
signatures without the user’s consent and knowledge
(Dasgupta, 2007; Girard, 2003; Marchesini, 2005).
Other attacks are focused on deceiving the user to
sign a message posing to be the original one (Spalka,
2002). In this way, it is not possible to assure a re-
liable signature if complex document formats are be-
ing used, because the WYSIWYS (What You See Is
What You Sign) property is completely undermined.
WYSIWYS (Scheibelhofer, 2001) is a security mea-
sure that provides the signatory with a last step verifi-
cation by means of a graphical representation of what
is going to be signed. Once the signatory confirms
the displayed information, it is the one supposed to
be sent to the SCDev, and therefore the information
on which the digital signature is computed. Neverthe-
less, if the document format allows the inclusion of
complex data structures, active code or hidden text,
then the semantic can vary depending on specific con-
ditions, conditions that can be manipulated by the at-
tacker. Therefore, the signatory could sign a docu-
ment with the desired semantic meaning while the
verifier could visualize a substantially different doc-
ument. Because the syntactic of the signed informa-
tion is maintained, the signature is correctly verified.
This security problem has been widely studied in the
literature (Alsaid, 2005; Jøsang, 2002; Kain, 2003),
and has become one of the most dangerous issues in
e-signatures.
Other attacks, called side-channel attacks, exploit
the information leakage from physical characteristics
of the hardware during the execution of the crypto-
graphic algorithm. The aim of these attacks is to
extract the private key from the SCDev. Depending
on the hardware characteristic analyzed, these attacks
are classified in Timing Analysis attack (Kocher,
1996; Schindler, 2000; Brumley, 2003), Power Anal-
ysis attack (Kocher, 1999; Fahn, 1999; Le, 2008),
Electromagnetic Emanation attack (Quisquater, 2001;
Gandolfi, 2001; Tanaka, 2008) or Microarchitec-
tural attack (AciiC¸mez, 2007a; AciiC¸ mez, 2007b;
AciiC¸mez, 2007c).
Finally, it is worth noting that design flaws and
code bugs in the SCA or the underlying software are
an endless source of vulnerabilities that an attacker
can exploit to break the security of the signature pro-
cess. Standard security evaluations like Common
Criteria or Federal Information Processing Standards
(FIPS) 140-2 can be applied to obtain a certain level
of assurance on the security of the system. However,
most manufactures are not keen on them due to their
ON THE NEED TO DIVIDE THE SIGNATURE CREATION ENVIRONMENT
377
high costs, and, though applied, just a level of assur-
ance, and not a level of certainty, is achieved.
4 DIVIDING THE SIGNATURE
CREATION ENVIRONMENT
It is obvious to recognize that perfect security does
not exist. There will always exist a risk. A single
signature creation environment will have a higher or
lower probability of suffering an attack, but the prob-
ability is never null, specially under the scenario de-
scribed in section 3. Our proposal consists in dras-
tically reducing the probability of an attack by us-
ing several environments, instead of one. Obviously,
there must be a trade-off between the added complex-
ity and the security improvement. Formal proofs for
our proposal are given further.
We consider that the legally binding evidence
does not consist of a single e-signature, but several.
Some protocols take into account a multi-signature
based evidence, like fair non-repudiation (Kremer,
2002; Hernandez-Ardieta, 2008) or contract signing
(Backes, 2006) protocols. Thus, these protocols are a
perfect candidate to use more than just one signature
creation environment.
4.1 Provable Benefits of Using Several
Environments
Next, the proofs of the benefits of using several en-
vironments for the generation of the digital evidence
are given. From here on, environment corresponds to
a signature creation environment (i.e. a PC, a mo-
bile device, etc.), and signatory to the end user that
needs to generate an evidence based on PKI-based e-
signatures.
Definition 1. An attack on an environment is an
attack carried out by a malicious agent (active intruder
or resident malware) which purpose is to obtain some
benefit from the signature generation capabilities of
that environment. Methods used by the malicious
agent include obtaining the signing private key and
deceiving the signatory to sign data different than the
purported one.
Definition 2. The probability of a successful
attack (PSA) on an environment depends on both
the probability of a malicious agent (attacker) to
gain access to that environment (undermine the
environment’s security measures) and the probability
of that attacker to subvert the specific security
measures implemented by the environment to protect
the signature capabilities from unauthorized usages.
Claim 1. Increasing the number of environments
needed in conjunction to generate the evidence en-
hances the reliability of the resultant evidence.
Proof 1. Suppose a set of environments Set(E) of
size n 2, being n the number of possible environ-
ments available to the signatory, each of which with
a specific PSA. The PSA on Set(E) is given by next
equation:
PSA(Set(E)) =
n
i=1
PSA(E
i
) (1)
We are considering the resultant PSA as the prob-
ability of occurrence of n independent events. How-
ever, subverting the security of a process in which
several environments are needed implies a kind of
collaborative attack from the attacker’s side. As a
consequence, the actual PSA would even be lower.
Notwithstanding, we will maintain this value of PSA
for the analysis.
Let PSA(E) be the probability of a successful at-
tack on a single environment E.
The PSA of Set(E) is always lower than the PSA
of a single environment E if an environment E
mem-
ber of the set Set(E) has a PSA lower than or equal
to the PSA of the environment E, and at least one of
the rest of the environments members of the set has a
PSA lower than 1.
PSA(Set(E)) < PSA(E), if E
Set(E) /
PSA(E
) PSA(E)
n1
i=1
PSA(E
i
) 6= 1,E
i
Set(E)
(2)
The direct consequence of Proof 1 is that adding
new environments - either equal to the former envi-
ronment, and thus with equal PSA, or different to it,
and thus with equal or different PSA - will always
improve the security of the system by decreasing the
final PSA. The assumption of adding environments to
the set with a PSA lower than 1 is reasonable, as the
signatory would never use an environment which is
known a priori to be compromised.
Another obvious conclusion that can be derived
from 1 is that if n tends to infinite, the PSA tends to 0,
providing that new environments added have a PSA
lower than 1.
lim
n
n
i=1
PSA(E
i
) = 0, PSA(E
i
) 6= 1 (3)
SECRYPT 2009 - International Conference on Security and Cryptography
378
4.2 Provable Benefits of Using
Heterogeneous Environments
Section 4.1 has proved that using several environ-
ments increase the level of security of the system.
This section analyses the impact of configuring a set
Set(E) of environments to be used for the evidence
generation in case the set consists of either homoge-
neous or heterogeneous environments.
Definition 3. We define homogeneous environ-
ments as those environments that can be attacked by
the same type of attacker. That is, their implemented
security measures and the type of potential attacker
are the same. As a result, the PSA for those environ-
ments remains the same. On the contrary, we define
heterogeneous environments as those that, either due
to their nature or the implemented security measures,
different types of attacker must be considered. In this
case, heterogeneous environments can have the same
or different PSA.
Claim 2. Replicating the same environment in the
set of environments Set(E) (homogeneous environ-
ments) always provides a higher level of security than
a configuration based on heterogeneous environments
providing that the chosen environment is the most se-
cure one among all possible environments.
Proof 2. Let PSA
hom
(E) be the resultant PSA of n
homogeneous environments:
PSA
hom
(E) = PSA(E).PSA(E)...PSA(E)
PSA
hom
(E) =
n
i=1
PSA(E) = (PSA(E))
n
(4)
Let PSA
het
(E) be the resultant PSA of n heteroge-
neous environments:
PSA
het
(E) = PSA(E
1
).PSA(E
2
)...PSA(E
n
)
PSA
het
(E) =
n
i=1
PSA(E
i
) (5)
From 4 and 5 we can deduce that:
PSA
hom
(E) < PSA
het
(E),if PSA
E
j
< PSA(E
i
),
i = 1...n and j {1...n}
(6)
Claim 3. In a more general manner, replicating the
same environment in the set of environments Set(E)
(homogeneous environments) provides a higher level
of security if the resultant PSA in (4) is lower than
that obtained from a configuration based on heteroge-
neous environments (5).
Proof 3. Basing on PSA
hom
(E) and PSA
het
(E)
given in 4 and 5 respectively, there can be a con-
figuration of homogeneous environments where:
(PSA(E
j
))
n
<
n
i=1
PSA(E
i
), j {1...n} (7)
However, in practice the signatory generally uses
her PC as the signature creation environment. Once
the PC is considered a highly risky environment (high
PSA), Claims 2 and 3 are clearly difficult to be
achieved. The conclusion is that, if the signatory uses
her PC as one of the environments, using additional
heterogeneous environments (i.e. a mobile device)
will surely provide a higher level of security.
5 CONCLUSIONS
PKI-based e-signatures are legally binding accord-
ing to International laws. Moreover, these signatures
act as non-repudiation evidence in electronic transac-
tions, preventing the parties denying that a particular
event or action took place. However, several attacks
can be easily performed on the environments where
these signatures are produced. As a consequence, the
reliability of the evidence is drastically undermined.
In this paper we have demonstrated that the divi-
sion of the signature creation environment is an ef-
fective approach that can substantially decrease the
probability of a successful attack, even though the en-
vironments used by the signatory are untrusted. As
a result, the reliability of a multi-signature based evi-
dence is clearly enhanced.
Our proposal can be applied to a wide variety of
Internet protocols, like e-commerce or contract sign-
ing protocols, where several signatures must be per-
formed by each participant. In (Hernandez-Ardieta,
2009) we propose a fair exchange protocol where the
environmentdivision principle has been applied to en-
hance to reliability of the protocol evidence.
ACKNOWLEDGEMENTS
This research has been partially supported by
the Ministry of Industry, Tourism and Trade
of Spain, in the framework of the project
CENIT-Segur@, reference CENIT-2007 2004.
(https://www.cenitsegura.es)
ON THE NEED TO DIVIDE THE SIGNATURE CREATION ENVIRONMENT
379
REFERENCES
AciiC¸ mez, O. (2007). Yet another MicroArchitectural At-
tack: Exploiting I-cache. In Proc. of the 2007 ACM
workshop on Computer security architecture.
AciiC¸ mez, O., Ko, C¸ . K., and Seifert, J.-P. (2007). On
The Power of Simple Branch Prediction Analysis.
2007 ACM Symposium on Information, Computer
and Communications Security (ASIACCS’07).
AciiC¸ mez, O., Schindler, W., and Ko, C¸ . K. (2007). Cache
Based Remote Timing Attack on the AES. Topics in
Cryptology - CT-RSA 2007 (pp. 271-286.). Springer-
Verlag, LNCS, series 4377.
Alsaid, A., and Mitchel, C. J. (2005). Dynamic content at-
tacks on digital signatures. Information Management
& Computer Security, 4 (13), 328-336.
Backes, M., Datta, A., Derek, A., Mitchell, J. C., Turuani,
M. (2006). Compositional analysis of contract-signing
protocols. Theoretical Computer Science 367, 33-56.
Brumley, D., and Boneh, D. (2003). Remote Timing At-
tacks are Practical. In Proc. of the 12th Usenix Secu-
rity Symposium.
Dasgupta, P., Chatha, K., and Gupta, S. K. S. (2007).
Vulnerabilities of PKI based Smartcards. In Proc. of
the IEEE Military Communications Conference 2007
(MILCOM 2007).
European Directive 1999/93/CE of the European Parliament
and of the Council of 13 December 1999 on a Com-
munity framework for electronic signatures.
Fahn, P., and Pearson, P. (1999). IPA: A New Class of
Power Attacks. In Proc. of CHES 1999 (pp. 173-186).
Springer-Verlag, LNCS, series 1717.
Federal Trade Commission, Department of Commerce,
United States of America. (2000). Electronic Signa-
tures in Global and National Commerce Act.
Gandolfi, K., Mourtel, C., and Olivier, F. (2001). Electro-
magnetic Analysis: Concrete Results. In Proc. of the
Cryptographic Hardware and Embedded Systems (pp.
251-261). Springer-Verlag, LNCS, 2162.
Girard, P., and Giraud, J-L. (2003). Software attacks on
smart cards. Information Security Technical Report,
8 (1), 55-66.
Government of Canada, Department of Justice. (2000). Per-
sonal Information Protection and Electronic Docu-
ments Act.
Hernandez-Ardieta, J. L., Gonzalez-Tablas, A. I., Alvarez,
B. R. (2008). An Optimistic Fair Exchange Protocol
based on Signature Policies. Computers & Security,
27 (7-8), 309 - 322. Elsevier.
Hernandez-Ardieta, J. L., Gonzalez-Tablas, A. I., Ramos,
B. (2009). Formal Validation of OFEPSP+ with
AVISPA. Joint Workshop on Automated Reasoning
for Security Protocol Analysis and Issues in the The-
ory of Security. Springer-Verlag, LNCS. (to appear)
International Organization for Standardization. (1996).
ISO/IEC DIS 13888-1. Information technology - Se-
curity techniques - Non repudiation - Part 1: General
model. ISO/IEC JTC1/SC27 N1503.
International Organization for Standardization. (1997).
ISO/IEC 13888-3 Information technology - Security
techniques - Non repudiation - Part 3: Mechanisms
Using Asymmetric Techniques.
Jøsang, A., Povey, D., and Ho, A. (2002). What You See is
Not Always What You Sign. In Proc. of the Australian
UNIX User Group. Melbourne.
Kain, K. (2003). Electronic Documents and Digital Signa-
tures. Master Thesis.
Kocher, P. C. (1996). Timing attacks on Implementations of
Diffie-Hellman, RSA, DSS and Other Systems. In Ad-
vances in Cryptology - CRYPTO ’96 (pp. 104-113).
Springer-Verlag, LNCS series 1109.
Kocher, P., Jaffe, J., and Jun, B. (1999). Differential Power
Analysis. In Proc. of CRYPTO 1999 (pp. 388-397).
Springer-Verlag, LNCS series 1666.
Kremer, S., Markowitch, O., Zhou, J. (2002). An inten-
sive survey of fair non-repudiation protocols. Com-
puter Communications, 25, 1601-1621.
Le, T-H., Canovas, C., and Clediere, J. (2008). An overview
of side channel analysis attacks. In Proc. of the 2008
ACM Symposium on Information, Computer and
Communications Security (ASIACCS 2008).
Marchesini, J., Smith, S.W., and Zhao, M. (2005). Keyjack-
ing: the surprising insecurity of client-side SSL. Com-
puters & Security, 24 (2), 109-123.
McCullagh, A., and Caelli, W. (2000). Non-repudiation in
the digital Environment. First Monday, 5 (8).
Quisquater, J.-J., and Samyde, D. (2001). ElectroMagnetic
Analysis (EMA): Measures and Counter-measures for
Smart Cards. In Proceeding of the International Con-
ference on Research in Smart Cards (pp. 200-210).
Springer-Verlag, LNCS, 2140.
Scheibelhofer, K. (2001). What You See Is What You Sign -
Trustworthy Display of XML Documents for Signing
and Verification. In Proc. of the IFIP TC6/TC11 In-
ternational Conference on Communications and Mul-
timedia Security Issues of the New Century.
Schindler, W. (2000). A Timing Attack against RSA with
the Chinese Remainder Theorem. In Proc. of the
Cryptographic Hardware and Embedded Systems (pp.
110-125). Springer-Verlag, LNCS, series 1965.
Spalka, A., Cremers, A. B., and Langweg, H. (2002). Trojan
Horse Attacks on Software for Electronic Signatures.
Informatica, 26, 191-203.
Tanaka, H. (2008). Evaluation of Information Leakage
via Electromagnetic Emanation and Effectiveness of
Tempest. IEICE Transactions on Information and Sys-
tems, 91 (5), 1439-1446.
Tiri, K. (2007). Side-channel Attack Pitfalls. In Proc. of the
44th ACM IEEE Design Automation Conference.
United Nations. (2001). UNCITRAL Model Law on Elec-
tronic Signatures with Guide to Enactment.
Zhou, J., and Gollmann, D. (1997). Evidence and Non-
repudiation. Journal of Network and Computer Appli-
cations, 20 (3), 267-281.
SECRYPT 2009 - International Conference on Security and Cryptography
380