Next-Steps Monkeysphere Projects: --------------------------------- Detail advantages of monkeysphere: detail the race conditions in ssh, and how the monkeysphere can help you reduce these threat vectors: threat model reduction diagrams. Determine how openssh handles multiple processes writing to known_hosts/authorized_keys files (lockfile, atomic appends?) Handle unverified monkeysphere hosts in such a way that they're not always removed from known_hosts file. Ask user to lsign the host key? Handle multiple hostnames (multiple user IDs?) when generating host keys with gen-key. Work out the details (and describe a full use case) for assigning a REVOKER during monkeysphere-server gen_key -- how is this set? How do we export it so it's available when a second-party revocation is needed? Actually enable server hostkey publication. Streamline host key generation, publication, verification. See doc/george/host-key-publication for what dkg went through on 2008-06-19 Ensure that authorized_user_ids are under as tight control as ssh expects from authorized_keys: we don't want monkeysphere to be a weak link in the filesystem. What happens when a user account has no corresponding /etc/monkeysphere/authorized_user_ids/$USER file? What gets placed in /var/cache/monkeysphere/authorized_keys/$USER? It looks currently untouched, which could mean bad things for such a user. - if authorized_user_ids is empty, then the user's authorized_keys file will be also, unless the user-controlled authorized_keys file is added. I believe this is expected, correct behavior. Consider the default permissions for /var/cache/monkeysphere/authorized_keys/* (and indeed the whole directory path leading up to that) As an administrator, how do i reverse the effect of a "monkeysphere-server trust-keys" that i later decide i should not have run? Make sure alternate ports are handled for known_hosts. Script to import private key into ssh agent. Provide a friendly interactive UI for marginal or failing client-side hostkey verifications. Handle the common cases smoothly, and provide good debugging info for the unusual cases. Make sure onak properly escapes user IDs with colons in them. Build a decent, presentable web site for documentation, evangelism, etc. Include a mention of how to report trouble or concerns. Create ssh2openpgp or convert to full-fledged keytrans. Resolve the bugs listed in openpgp2ssh(1):BUGS. Understand and document alternate trustdb models. Understand and document the output of gpg --check-trustdb: gpg: 3 marginal(s) needed, 1 complete(s) needed, PGP trust model gpg: depth: 0 valid: 2 signed: 20 trust: 0-, 0q, 0n, 0m, 0f, 2u gpg: depth: 1 valid: 20 signed: 67 trust: 15-, 0q, 1n, 3m, 1f, 0u gpg: next trustdb check due at 2008-10-09 Understand and document the numeric values between sig! and the keyid in "gpg --check-sigs $KEYID" . Compare with the details found from "gpg --with-colons --check-sigs $KEYID". This has to do with trust signatures. Fix gpg's documentation to clarify the difference between validity and ownertrust. Include better documentation for trust signatures. Make it easier to do domain-relative ssh host trust signatures with gnupg. (e.g. "i trust Jamie McClelland (keyID 76CC057D) to properly identify ssh servers in the mayfirst.org domain") See: http://tools.ietf.org/html/rfc4880#section-5.2.3.21 and grep for "tsign" in gpg(1). Fix the order of questions when user does a tsign in gpg or gpg2. File bug against ssh-keygen about how "-R" option removes comments from known_hosts file. File bug against ssh-keygen to see if we can get it to write to hash a known_hosts file to/from stdout/stdin. Add environment variables sections to man pages. Environment variable scoping. Move environment variable precedence before conf file. When using ssh-proxycommand, if only host keys found are expired or revoked, then output loud warning with prompt, or fail hard. Update monkeysphere-ssh-proxycommand man page with new keyserver checking policy info. Update monkeysphere-ssh-proxycommand man page with info about no-connect option. File bug against seahorse about how, when creating new primary keys, it presents option for "RSA (sign only)" but then creates an "esca" key. File bug against enigmail about lack of ability to create subkeys. Priviledge separation: monkeysphere user to handle authn keyring and generate authorized_keys file (which would be moved into place by root). Host keyring would be owned by root. Check permissions of authorized_user_ids file to be writable only by user and root (same as authorized_keys) Improve function that sets owner trust for keys in server keychain.