Year: 2013
Type: dissertation
Abstract: The thesis deals with open problems in the area of two-party secure communications protocols and cryptographic key protection.A secure communication protocol ensures the authenticity of both parties andprotects the confidentiality, integrity and freshness of the informationthat is transmitted. For this purpose, both parties require long termcryptographic keys that are used for mutual entity authentication and toestablish session keys. We evaluate the methods used for long-term key protection in software, and in combinations of software and hardware (e.g., smart cards). We further investigate whether these methods are sufficient to prevent key extractionwhen the protected long-term keys are combined with secure communications protocols. We provide evidence that the currently used methods are not sufficient to ensure the overall security of a system when the protected keys are used in conjunction with an arbitrary protocol, and demonstrate this withan attack on the smart card-based public key Kerberos (PKINIT) protocol. We further attempt to ensure the overall security of a system by enhancing Shoup's simulatability-based static corruptions model to account for threats related to long-term key protection. Under this model we examine the Diffie-Hellman variant of the PKINIT protocol and show that a fix to the protocol, we proposed, is secure. Furthermore, we develop a software-based long-term key protection scheme for the Linux-kernel, that is used to protect long-term keys from application compromise. It relies on the isolation mechanisms available in the operating system, and is designed to avoid all pitfalls and known attacks from previous designs. Finaly, we study a family of attacks on secure communications protocols, namely the cross-protocol or multi-protocol attacks. These attacks exploit interactions between different protocols that share the same keys. While the threat of such attacks was often highlighted, only few practical attacks exist for secure communications protocols. We survey the known attacks and countermeasures, and present a new attack on the widely deployed TLS (also known as SSL) protocol that relies on interactions between the Diffie-Hellman ciphersuites and their elliptic curve counterparts. Moreover, to mitigate this attack and related attacks we propose countermeasures for the TLS protocol.
Source: Lirias
Author: Arenberg Doctoral
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2185065272