and also requires that the user possess a secure ID-token with a small display.
Another interesting solution to secure e-commerce is proposed in [154]. Like us, Singarevelu et al. propose reducing the code size and establishing isolated environments using a microkernel. However, in contrast to our approach, Singarevelu et al. do not show how attestation techniques could be used to provide assurances that a user is still acting with a trusted environment. In addition, the approach presented by Singarevelu et al. does not provide any means of preventing man-in-the-middle attacks on TLS.
There has also been much work done to prevent phishing attacks in electronic com-merce. The objectives, motivations and attack patterns of phishers have been exten-sively studied in the literature (e.g., [72, 44, 111]) and many proposals have been made to prevent these attacks (e.g., [111, 188]).
In this context, one very interesting proposal has been made by Gajek et al. [65].
The authors proposed to augment a web-browser with an additional wallet that is responsible for performing user authentication and for authenticating websites. The wallet is strictly isolated from the web-browser and runs in a compartment established by the PERSEUS security kernel [121]; it is, thus, robust against infection from malware.
However, since the proposal does not support attestation techniques, it does not provide any means of providing a user with assurances that their used web-browser or the wallet is still trusted.
Cox et al. [40] propose the Tahoma Web browsing system. This architecture uti-lizes virtualization techniques to execute multiple isolated compartments on the Xen-hypervisor. In each compartment, an instance of a full-fledged web-browser runs and a network proxy running inside the management domain of Xen [14], controls the net-work connection. While the approach is able to restrict the access rights of one browser instance based on a specified policy, it is not able to handle and prevent man-in-the-middle attacks on TLS. In addition, there are no mechanisms present to enable a user to place trust into a web-browser running in an isolated compartment.
Other work that looks at trusted computing and e-commerce includes [12, 13, 4].
However, these approaches do not show how a secure e-commerce architecture that supports attestation could be realized. Thus, these approaches are not able to generate assurances of the trust level of a user’s platform or to transfer these assurances to the user.
9.8 Summary
Server platforms are relatively robust and, whilst capable of being compromised, are typically not the easiest target for cyber-criminals. Instead, client platforms are increas-ingly being subverted since users typically do not have the same financial motivation (as merchants) to harden their systems. Indeed, many users neither know how to harden their systems nor how to discover if their system has been subverted. The solution we have proposed is based on virtualization in combination with attestation techniques and allows a user to ensure that a particular client configuration has not been tampered
with and is trusted for the duration of the transaction. In addition, our architecture ensures that confidential data, such as authentication passwords, are not accidentally transferred to malicious servers that masquerade as authentic servers.
Chapter 10
Signature Creation with TPMs
In this chapter, we discuss the potential of the TPM and give considerations as to whether a TPM could be used as a secure signature creation device. We examine whether the TPM can be used as a secure signature creation device that conforms to the EU Electronic Signature Directive as well as to the German Electronic Signature Law. In addition, we argue that if the TPM can be used as secure signature creation device, a trusted signing software also becomes necessary. This chapter shares some material withThe Creation of Qualified Signatures with Trusted Platform Modules [162]
and Erzeugung elektronischer Signaturen mittels Trusted Platform Modules [163].
10.1 Introduction
Despite the existing shortcomings of password-based user authentication, i.e., their vul-nerability to phishing attacks [4], these techniques are currently wide spread and most of the available e-commerce applications are based on these techniques. To overcome these shortcomings qualified electronic signatures might be used. These signatures are considered to be a vital component for e-commerce transactions in the future, since they are capable of linking a particular user to a particular transaction effected over the in-ternet. In contrast to the password-based user authentication, they can act as a prima facie evidence in trial since they offer non-repudiation, which is especially relevant in the realm of e-commerce. Thus, qualified signatures are a very important component for trust-establishment.
A qualified electronic signature that confirms to article 5(1) of the EU Electronic Signature Directive (EU Directive) and § 2(3) of the Signature Act 2001 (SigG) is an advanced signature as specified by article 2(2) of the EU Directive and § 2(2) of SigG, which is based on a qualified certificate and which is created by a secure signature creation device.
However, qualified signatures are not widely used, and it is asserted that the failure to use qualified signatures thereby deprives the market for electronic commerce of an important source of potential growth [126]. A promising approach to overcome this shortcoming is to use the Trusted Platform Module (TPM) to create a qualified
signa-173
tures. The TPM is already available in approximately 200 million personal computers [178] and this technology is supported by many hardware vendors and is therefore widespread. One of the properties required of this chip is to perform the necessary cryptographic functions to create an advanced signature. However, the ability to per-form the required mathematical operations is not enough to use this device to create an advanced signature. This is due to the requirements for a compliant device for the creation of an advanced signature, as set out in article 5(1) and Annex I - III of the EU Directive and the Signature Act 2001.
Unfortunately, even if we were to assume that the TPM is able to create a qualified signature, the problem arises that the software that is responsible for communicating with the TPM might tamper on data delivered to the TPM. As a result, the owner of the TPM might sign false or malicious transaction data without noticing. This may result in the fact that the signatory made an unintentional contract which he cannot deny. As a result, attestation techniques become necessary.
In this chapter, we will firstly validate whether the TPM can be used to create qualified electronic signatures, and if so, whether such signatures could be used in e-commerce. Thereafter, this chapter considers whether the TPM fulfills the technological requirements of article 5(1) of the EU Directive and § 2(3) SigG, together with the criteria provided for in Annex III of the EU Directive and §17 Abs. 1 SigG, as well as
§ 15 Signature Decree 2001 (SigVO).