| Internet Engineering Task Force | P. M. Hallam-Baker |
| Internet-Draft | Comodo Group Inc. |
| Intended status: Experimental Protocol | April 07, 2011 |
| Expires: October 09, 2011 |
Options for Improving PKIX Revocation
draft-hallambaker-revocation-options-00
In recent weeks a number of proposals have been made for improving the effectiveness of certificate revocation. This document is an attempt to bring these proposals atogether and analyze them with respect to a use cases and requirements framework.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on October 09, 2011.
Copyright (c) 2011 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
The following terms are used in this document:
TBS
The legacy base of browsers does not support hard fail when an OCSP responder is unavailable
The legacy base of servers does not support OCSP stapling
The legacy DNS does not provide for cryptographic authentication of responses
Prevent clients relying on mis-issued or fraudulently issued certificates
Prevent clients relying on certificates that were legitimately issued but have since been compromised (e.g. subject key is compromised)
Time spent checking certificate status in the browser
Complexity of server operations.
Certificate Authority Authorization [I-D.hallambaker-donotissue] provides a general purpose, extensible platform that allows domain name owners to express certificate issue requirements. Although currently limited to specification of trust roots, it is extensible to allow additional properties to be agreed and specified. While the security of CAA is improved with deployment of DNSSEC in the specified zone, CAA is still an effective control without DNSSEC.
Deployment of CAA does not require code at the client or server, except that the DNS server used must be recent enough to support specification of unknown RRs. It is thus compatible with [C-Legacy-Browser] and [C-Legacy-Server].
Certificate is issued for short lifetime (24-72 hours). Servers must thus ensure that they continuously update their certificate stores to download and install certificates as they are issued.
One critical issue in this approach is the synchronization of the relying party clock to that of the issuing CA (presumed to be within a few seconds of UTC). In order to avoid unnecessary rejection of certificates it is probably desirable for a server to only start using a certificate 24 hours after the notValidBefore time and to renew the certificate 24 hours before the notValidAfter time is reached. Experts are currently looking into the detailed implications of timing and short lived certificates.
By itself, [P-Short] meets criteria [R-Comp] but not [R-Fraud] since an attacker can simply request issue of a long lived certificate from a compromised CA.
[P-Short] is netural with respect to [M-Client-Efficiency] and requires a significant change to server administration and is thus negative for [M-Server-Admin] in the short term. In the long term however, it is much easier to guarantee the reliability of a certificate download that must happen every day and is thus automated, than a procedure that happens once a year and is likely to be forgotten.
Deployment of [P-Short] is fully compatible with the legacy base of deployed browsers which will all accept short lifetime certificates and refuse to rely on a certificate that has expired.
Use of [P-Short] in combination with an appropriate flag in [P-CAA] allows the requirement [R-Fraud] to be met, provided that either the CA or the client software checks the CAA record.
This proposal is identical with respect to the metrics and constraints as [P-Short]
If the lifetime of a short lived certificate is shorter than the validity period accepted for an OCSP or CRL response, a client MAY choose to rely on the short lived certificate without performing the customary revocation checking. A PKIX certificate extension MAY be specified as a means of enabling a client to determine that a certificate is intended for use as a short-lived certificate without revocation checking.
[P-Short-Optimized] allows an improvement in the metric [M-Client-Efficiency] by eliminating the need for separate revocation checking.
A certificate includes a flag that tells relying parties that it is only to be used with TLS implementations that support OCSP stapling. Should a client attempt to connect to the server that does not offer the stapled OCSP response, the connection is invalid and must be aborted.
[P-Stapling-Flag] arguably requires less administration effort than [P-Short] as it allows the site to continue to use long lived certificates and is thus better on [M-Server-Admin]. The principal disadvantage being that it requires deployment of a server that can support stapling and is thus worse on [C-Legacy-Server]. The stapling flag will only be observed by new clients and the proposal thus fails to meet [C-Legacy-Browser].
Another proposal similar to [P-Short] is for the subject to host their own OCSP responder which is in theory less likely to be vulnerable to failure. This may be the primary OCSP responder [P-OCSP-LOCAL] or merely a secondary [P-OCSP-LOCAL-2ND].
While this approach does not require server support for OCSP stapling and is thus compatible with [C-Legacy-Server], the administrative effort for the subject is considerably greater and it thus fails on [M-Server-Admin].
[TBS]
None
This draft draws on input from many contributors, in some cases the same proposal being made more than once in different contexts. The concept of short lived certificates was proposed in WAP forum. This proposal and the proposal to use local OCSP was mentioned on a Mozilla telecon.