Establishing end to end trust

Microsoft issued an extremely interesting white paper: Establishing end to end trust. It has been presented at RSA2008. The paper is worth reading. The main idea is that a trusted stack (encompassing hardware trust, OS trust, application trust, data trust and persona trust) and the ability to audit for accountability should make a more secure Internet.

It is interesting to note the extreme caution Microsoft takes on the topic of privacy and identity. Section IV is a fully dedicated cautionary note. Clearly, Microsoft fears that this initiative is considered as a Big Brother initiative. This is probably a sequel of the backlash on palladium.

I will focus on the notion of trusted stack. This is an addition to previous post on XBOX hack. The trusted stack is based on signature. According to the paper, there will be three categories.
“Even if code is signed, however, it will still fall into one of three buckets. There will be code that is signed by a known entity (e.g., Microsoft, Oracle, Adobe) that is trusted due to past experience, brand reputation or some other factor; there will be code that is signed but known to be malware (e.g., spyware, which can then be blocked); and there will be code signed by entities that are not known to the user.”
The paper clearly highlights the importance of the criteria to obtain the signature. If they are weak, then the trust is weak. The concept of signature relies on the fact that an authority, often called trusted third party, provides signature keys and associated certificates only to compliant and trusted principals. We expect the trusted third party do correctly its job. One of the strength of PC is the wealth of available shareware and freeware. There are thousands of small software publishers in the world. Thus, thte authority will never be possible to know if they are trust worthy. Will these publishers be allowed to sign?

To compensate, Microsoft proposes a reputation platform. Unfortunately, like in all reputation system, it has limitations. Reputation will increase only with the number of users recommending the software, i.e., the number of people taking the risk. Furthermore, many people will not check ( the same people that do not use an antivirus or do not update their software).

Furthermore, as explained in previous post, signature does not mean that the software is secure. Only peer auditing of the software before signing the application may give this assurance.

In other words, trusted stack as described will end up with the following situations:

  • Signed software that we trust because they are open source or from a publisher we trust.
  • Signed software that we do not know if we can trust.

It is still up to the user to decide if he takes the risk. In other words, we are not far from the existing situation. The only difference is that with a trusted stack based on TPM, application may trust and use secure elements of lower layers and interact with other trusted principals.

There are also many things to be said about audit. This is for another post.

Leave a Reply

Your email address will not be published. Required fields are marked *