Everyone dealing with Information Security knows very well that SNMP (which stands for Simple Network Management Protocol and corresponds to the standard UDP protocol used to monitor servers and network elements) is considered insecure. In too many circumstances network administrators forget to change the default community strings (the strings used to “softly” authenticate the manager and the agents) from their default values which are typically “public” for read-only access and “private” for read-and-write access on the monitored device. This happens sometimes for thoughtlessness, or simply because network administrators do not consider changing the default security strings a security issues.
And even if SNMP version 3 is used (which grants encryption and mutual authentication between the manager and the agents -at least the attackers may not spoof the default community strings-) in 12 years of honorable career I never found so far the right combination between manager and agent versions: I mean when you have a network manager supporting version 3, the agents only support version 1 or 2c and vice versa if the agents support version 3 you may be sure that the manager only supports version 1 or 2c.
Now there is a reason more to consider SNMP (and its default configurations) an hazard for Information Security. This reason is four letters long and is called LOIC, the infamous tool used by Anonymous to perpetrate the well known DDoS attacks.
So far the infosec community has been divided into two opposite factions: on one side those who think that Anonymous-perpetrated DDoS attacks are successful even with a small number of “enrolled cannons” since the same Anononymous owns a Botnet which from time to time is unleashed against the target. On the other side those claiming that this kind of attacks may be successful only if a huge number of participants volunteer accomplices is enrolled.
Today an article written by Alex Holden, Cyopsis Director of Enterprise Security, offers an alternative hypothesis. The attack method Holden describes is called a Reflected Denial of Service (RDoS) and just utilizes SNMP, which is UDP-based, exploiting the weaknesses in default configurations which populate many devices composing the Internet, with devastating consequences.
The SNMP paradigm, as the name suggests, is very simple: each device (server, network device or application) which must be monitored provides some status variables to the external world. The variables may be queried by a special application called network manager. The variables are organized in different groups (or leaves), and identified by OIDs (or Obiect IDentifiers). Querying the main OID (18.104.22.168) returns all the variables (this is an operation called snmpwalk).
If the assumption of Holden is correct, suppose you are able to spoof a manager with the same address than the target of the attack, and suppose to generate continuous SNMP queries with that address, querying the main OID from all the Internet devices which are known to have standard community strings. The unaware target will be flooded by SNMP replies from those devices with a lethal amplification effect and consequently an apparently innocent misconfiguration (that is the unchanged default community string) becomes an hazard for the Internet.
Of course this is a mere speculation (I did not verify source code), but this would explain why the Anonymous claimed that LOIC traffic
is was hard to detect (but not always): the SNMP protocol is very popular and widespread on the Internet.
(Original link via Infosecisland).