Show description

1 Physical Attack In this section, we consider physical attacks from outside the TCP, that seek to actively penetrate or otherwise disrupt the internal device. By physically attacking a TCP, the adversary hopes to subvert its security correctness properties somehow, usually by extracting some secret the TCP was not supposed to reveal. At first glance, the natural way to achieve this goal is the direct approach: somehow bypass the TCP’s protections and read the data. As the following sections will elaborate, this direct approach can often prove rather successful.

When Paul Kocher first published his timing attacks in 1995, at least one old-timer claimed that a few older commercial accelerators also took constant time, indicating that some in the commercial world must have already known about the attack. Reproducing the RSA timing attack has made an excellent homework project here at Dartmouth. Example: Apache SSL Web Servers. In the classic instantiation of cryptographic timing attacks, the adversary has direct access to a TCP carrying out RSA, and the TCP uses some variation of the standard multiply-and-square implementation of the modular exponentiation step.

These design choices affect the attack profile. For example, a TCP that leaves computation and memory exposed (such as TCPA/TCG-based platforms) permits more attacks than one that puts physical protection around the entire unit. A single-chip TCP requires different attack and defense techniques than a larger encapsulated module. A TCP intended to be carried in a user’s pocket may need to withstand a broader range of environmental conditions, which may complicate defending against attacks that use extreme conditions.