On authenticated computing and RSA-based authentication

  • Authors:
  • Jean-Pierre Seifert

  • Affiliations:
  • Intel Corporation, Hillsboro, OR

  • Venue:
  • Proceedings of the 12th ACM conference on Computer and communications security
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a novel hardware attack against RSA-based authentication of programs. Like the seminal paper from [10] it exploits the attacker's ability to arbitrarily tamper with a computational device during its data processing. But, contrary to [10] our method targets the RSA signature verification which processes only public data. Surprisingly, we prove how to let the RSA verification process accept signatures (with high probability and assuming the Riemann Hypothesis) of arbitrary code --- signed with our own self-created private key. While our attack is so far only theoretical, its practical feasibility has been already implicitly confirmed, cf. [6]. In fact, through real experiments with tamper-resistant devices [4] discovered the following caveat being at the heart of our vulnerability: the most often observed fault during RSA-computations exposed to glitch attacks is the erroneous modification of the moduli. Moreover, our attack relies only upon the simplest and most easiest practically implementable fault induction method described by [8]. Our idea is somehow inspired by practiced methods circumventing the FLASH Boot ROM authentication mechanism of the Xbox due to F. Lehner, cf[17]. Our attack can be interpreted as the extrapolation of the mathematical foundation underlying those Xbox vulnerabilities.