Steven M. Bellovin
firstname.lastname@example.org 973-360-8656 AT&T Labs Research Florham Park, NJ 07932
|Ran Atkinson||Paul Krumviede|
|Fred Baker||Marcus Leech|
|Steven Bellovin||Perry Metzger|
|Bob Blakley||Keith Moore|
|Matt Blaze||Robert Moskowitz|
|Brian Carpenter||John Myers|
|Jim Ellis||Thomas Narten|
|James Galvin||Radia Perlman|
|Tim Howes||John Richardson|
|Erik Huizer||Allyn Romanow|
|Charlie Kaufman||Jeff Schiller|
|Steve Kent||Ted T’So|
- Dependency graph of protocols and security mechanisms. What are the “holes”? The duplications?
- Finite set of concrete recommendations.
- Guidelines on how to write a good “security considerations” section. Even security protocols need to document their uses and limitations.
- More use of encryption by hackers.
- User-friendly exploitation scripts.
- Increasing numbers of attacks on the network infrastructure
- Repeated use of old vulnerabilities.
- Source not needed for attacks.
- Do we choose architectural purity or let people do their own thing.
- IPsec doesn’t solve all problems.
- The IETF is made of volunteers; how do we manage the process?
- Who does the security work? The WGs? The security area?
- What about long-running WGs? How can security be incorporated into their output?
- Hippocratic Oath: First, do no harm.
- New text for Working Group charters.
- Guidelines for “security considerations” section.
- Taxonomy of attacks.
- Implementation hints and pitfalls.
- Firewall issues.
- Workshop report.
- Document the risks (what can be attacked through this mechanism?)
- Document the assumptions (topology, deployment model, etc.)
- If appropriate, recommend the use of particular security technologies.
- All WGs must address security.
- This will include existing working groups, during charter updates.
- Each WG will work with “security volunteers”.
- No “security will be added later”.
- End-to-end security is usually better.
- The infrastructure is subject to attacks.
- Even intranets cannot be presumed secure against eavesdroppers.
- The infrastructure, and not individual protocols has to provide availability, but protocols shouldn’t make this worse when under attack.
- What kinds of things do hackers do?
- What are their targets?
- How to avoid common errors (à la RFC 1750)?
- What are the global risks?
- What can they do? What can’t they do?
- When are they useful or not?
- What protocol characteristics are firewall-friendly? Firewall-hostile?
- Core — we must have it; should be used if appropriate.
- Useful but not core.
- Not widely regarded as useful.
- To be killed.
- Out of scope.
- Security Multipart
- Signed keys in the DNS (50% support)
- X.509v3 (50% support)
- TLS (some support)
Many other security protocols are useful, but not central. This does not mean they can’t be used.
Protocols in this category include those that have been superseded, or that have failed to catch on, such as PEM and MOSS, or those that are duplicative of other work.
Any protocol that relies on the transmission of unencrypted passwords is terminally broken. Any protocol that puts confidential information in public places (such as URLs) is similarly broken.
CIFS, DFS, NFS, ONC RPC, LDAP
- Object security
- Secure e-mail
- Security for external routes