DNSChanger might change the BGPSEC landscape

In early November, a sophisticated fraud was shut down and a number of people arrested. Malware from a family called “DNSChanger” had been placed on around four million machines (Macs as well as Windows machines) over several years.

The compromised users had their DNS traffic redirected to criminally operated servers. The main aim of the criminals seems to have been to redirect search queries and thereby to make money from displaying adverts.

Part of the mitigation of DNSChanger involves ISC running DNS servers for a while (so that 4 million people whose DNS servers suddenly disappear don’t simultaneously ring their ISP helpdesks complaining that the Internet is broken).

To prevent bad people running the DNS servers instead, the address blocks containing the IPs of the rogue DNS servers which used to belong to the criminals (but are now pointed at ISC) have been “locked”.

This is easy for ARIN (the organisation who looks after North American address space) to acquiesce to, because they have US legal paperwork compelling their assistance. However, the Dutch police have generated some rather less compelling paperwork and served that on RIPE; so RIPE is now asking the Dutch court to clarify the position.

Further details of the issues with the legal paperwork can be found on (or linked from) the Internet Governance Project blog. The IGP is a group of mainly but not entirely US academics working on global Internet policy issues.

As the IGP rightly point out, this is going to be an important case because it is going to draw attention to the role of the RIRs — just at the time when that role is set to become even more important.

As we move to crypto-secured BGP routing, the RIRs (ARIN, RIPE etc) will be providing cryptographic assurance of the validity of address block ownership. Which means, in effect, that we are building a system where the courts in one country (five countries in all, for five RIRs) could remove ISPs and hosting providers from the Internet… and some ISPs [and their governments] (who are beginning to think ahead) are not entirely keen on this prospect.

If, as one might expect, the Dutch courts eventually uphold the DNSChanger compulsion on RIPE (even if the Dutch police have to have a second go at making the paperwork valid) then maybe this will prove the impetus to abandon a pyramid structure for BGP security and move to a “sea of certificates” model (where one independently chooses from several overlapping roots of authority) — which more closely approximates the reality of a global system which touches a myriad set of local jurisdictions.

4 thoughts on “DNSChanger might change the BGPSEC landscape

  1. Richard, thanks for your article.

    Can you please give more information or pointers to > a “sea of certificates” model (where one independently chooses from several overlapping roots of authority)< ?


  2. I think the “Pyramid approach” is RPKI. In this scheme, the RIRs are the root of trust and can be attacked as mentioned in the article.

    The “sea of certificates” would probably go into a Web-of-Trust direction, where I trust my peers’ certificates after validating them via some out-of-band communication.

Leave a Reply

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