Terry it is great to re-bring these things up, as we have discussed on the Slack channel. Everyone needs to keep in mind that the DNS SRV record is called out in the spec to allow intra-org device a way to discover the internal TAXII server. Every thing else is an implementation / deployment issue.
We have public facing servers today, DNS and HTTP are the most common. At some point, some organizations will have public facing TAXII servers. So therefore the TAXII servers need to be hardened with good coding standards and have controls put around them.
Thanks,
Bret
Bret Jordan CISSP
Director of Security Architecture and Standards
Office of the CTO Blue Coat Systems
PGP Fingerprint: 63B4 FC53 680A 6B7D 1447 F2C0 74F8 ACAE 7415 0050
Without cryptography vihv vivc ce xhrnrw, however, the only thing that can not be unscrambled is an egg.
On Oct 30, 2015, at 14:02, Terry MacDonald <
> wrote:
This is a very good point. From an attackers point of view this threat intelligence will be a very juicy target.
Things we know:
·
Attackers will try and get into a TAXII server if they are aware of it
·
Security through Obscurity is not a security control
·
Automated TAXII to TAXII community will enhance the usefulness of the complete CTI stack
·
Some suggestions made on the CTI-STIX list requires the ability for TAXII servers to automatically discover and communicate with each other (STIX request/response)
This really is a question of whether the benefits of the automated discovery outweigh the possibility that someone will see the TAXII server and gain access to it. Given that the TAXII servers will need to talk to other TAXII servers in order to operate (kinda like DNS servers do) we will require some documented communication mechanism to exist between them. If its not a DNS SRV record, then it will be a particular port on a particular hostname in their domain namespace. The attacker will just look for that domain name on that port and will discover the TAXII server anyway . The only caveat to that are TAXII servers with restricted membership and some kind of restricted access group (‘iptables’/firewalling) going on, which I don’t think will be common.
The fact is that our TAXII servers will need to access other TAXII servers across the internet in order to communicate. This requires opening up a port in the firewall. This port needs to be accessible to all in order to answer queries from TAXII clients and other TAXII servers.
There is an alternative. We could do the equivalent of BGP, and manually do TAXII ‘peering’ agreements between leaf TAXII servers and upstream TAXII servers, where we use upstream TAXII servers as TAXII ‘hubs’, and they then map out the TAXII network topology to determine where to route things. But this gets pretty complex pretty quickly.
I believe a better approach would be just to accept the fact the TAXII server WILL be discovered, and to hide it behind a reverse HTTP proxy with proper authentication, ensure the TAXII server is secured, patched and monitored, and maybe even run a split TAXII server (like split DNS) where the ‘publicly’ shared STIX data sits on the externally facing TAXII server and the ‘private’ STIX data sits on the internal server – just like DNS.
Cheers
Terry MacDonald
Senior STIX Subject Matter Expert
SOLTRA An FS-ISAC and DTCC Company
+61 (407) 203 206
