Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Network Intrusion Detection, Third Edition.pdf
Скачиваний:
212
Добавлен:
15.03.2015
Размер:
2.58 Mб
Скачать

Considering that the TCP sequence number field is 32 bits long, these are very small initial sequence numbers—quite unusual. All the TCP sequence numbers from this scan were lightweight, and when the activity was dumped in hex, the reason why was discovered. The high-order 16 bits of the TCP sequence number were always 0s. This is confirmation that some kind of sequence number manipulation was performed, and it becomes a signature of this activity.

Fragments, Just Fragments

Consider this final example of an inverse mapping technique. As you have already learned, only the first fragment chunk comes with protocol information. Attackers using this technique (along with some interesting variations) were able to penetrate older firewalls and filtering routers. The firewalls would assume that this was just another segment of traffic that had already passed their access lists. Needless to say, this has been fixed in most vendors' products.

In this case, however, the prober isn't particularly interested in firewall penetration. Once again, if one of the target hosts does not exist, the router sends back an unreachable message. The attacker can then compile a list of all the hosts that do not exist and, by taking the inverse of that list, has a list of the hosts that do exist. This is why this class of techniques is called inverse mapping. Take a look:

18:32:21.050033 PROBER > 192.168.5.71: (frag 9019:480@552) 18:32:21.109287 PROBER > 192.168.5.72: (frag 9275:480@552) 18:32:21.178342 PROBER > 192.168.5.73: (frag 9531:480@552) 18:32:21.295332 PROBER > 192.168.5.74: (frag 9787:480@552) 18:32:21.344322 PROBER > 192.168.5.75: (frag 10299:480@552) 18:32:21.384284 PROBER > 192.168.5.76: (frag 10555:480@552) 18:32:21.431136 PROBER > 192.168.5.77: (frag 11067:480@552) 18:32:21.478246 PROBER > 192.168.5.78: (frag 11579:480@552) 18:32:21.522631 PROBER > 192.168.5.79: (frag 11835:480@552)

Measuring Response Time

Lately, we've seen a lot of traffic coming from all over the place directed to DNS servers, but not for the purpose of querying for DNS information or ostensibly of malicious intent. What is happening is that companies have developed software that tries to deliver the best possible response time to web requests. It has been demonstrated that most users will tolerate about an eight-second delay in receiving responses and after that they might go to a competitor site with better response time. It has become a matter of e-business survival and profitability to offer good response time, and because necessity is the mother of invention, software has been created to accomplish the mission. The patterns explained in this section are from a product known as 3DNS.

One technique is to associate the user request with an authoritative DNS server for the user's host and find the response time to the DNS server. This assumes that the authoritative DNS server and the user's hosts are geographically close, which might not always be the case. Why not just find the distance to the user's host? Indeed, this seems more logical, but many sites are well protected, and access to the user's host is not always available. They figure there is a better chance of having some kind of access to the DNS server, which may or may not be the case.

There has been a lot of hue and cry from analysts who see their IDS fired because of the traffic generated by this software. Many sites feel violated because traffic is directed to the sacred DNS server, of all hosts. And, many more sites don't understand what is happening and perceive this activity to be an attack of some sort. The final objection is that this is unauthorized information gathering, regardless of whether it benefits the end user.

Let's take a look at some of the signatures associated with this type of traffic. One thing that you should keep in mind is that many different web sites use this software and so you will see many different source IPs. Because of the unique signatures generated from multiple source IPs, this has been mistaken for some kind of coordinated attack. As you will see, however, it

really isn't.

Echo Requests

No surprise with the following TCPdump activity to measure response time to your DNS server. The echo request is issued and the response time is measured based on receipt of an ICMP echo

reply, if there is one:

10:25:44.070000 216.32.68.13 > mydns.com: icmp: echo request 10:25:44.070000 216.32.68.13 > mydns.com: icmp: echo request 10:25:44.070000 216.32.68.13 > mydns.com: icmp: echo request 10:30:01.530000 216.32.68.13 > mydns.com: icmp: echo request 10:30:01.530000 216.32.68.13 > mydns.com: icmp: echo request 10:30:01.550000 216.32.68.13 > mydns.com: icmp: echo request 10:30:25.660000 209.67.29.8 > mydns.com: icmp: echo request 10:30:25.660000 209.67.29.8 > mydns.com: icmp: echo request 10:30:25.670000 209.67.29.8 > mydns.com: icmp: echo request 10:32:12.520000 209.67.78.200 > mydns.com: icmp: echo request

As you have learned, however, many sites block ICMP echo requests because ICMP has capability to map sites both actively with a ping, and also by eliciting error messages that give away the position of hosts and routers in a site. And, if this is the case, an attacker, or even a

service provider using a tool like 3DNS might focus their reconnaissance on the DNS server.

Actual DNS Queries

If the user's DNS server didn't respond to the ICMP echo request and the server using the 3DNS probing software is configured to continue to try to make contact with the DNS server, more

activity is sent, as shown here:

216.32.68.11.3200 > mydns.com.53: 0 [0q] Type0 (Class 0)?. (36) mydns.com.53 > 216.32.68.11.3200: 0 FormErr [0q] 0/0/0 (12) DF 216.32.68.11.3201 > mydns.com.53: 256 [0q] Type0 (Class 0)? . (36) mydns.com.53 > 216.32.68.11.3201: 0 FormErr [0q] 0/0/0 (12) DF 216.32.68.11.3202 > mydns.com.53: 512 [0q] Type0 (Class 0)? . (36) mydns.com.53 > 216.32.68.11.3202: 0 FormErr [0q] 0/0/0 (12) DF

A real DNS query is not issued, but one is sent to UDP port 53 with a DNS message of all 0s. TCPdump performs some integrity checking of the DNS message and if it discovers what it considers to be noteworthy fields, it reports them. The 0q means that there were zero queries

in the DNS message, for example. Normally, for types other than inverse queries there will be at least one query. That is why TCPdump reported it and all other 0-padded fields it considers to be odd. This elicits an error response from mydns.com, which is then used to compute the

round-trip time.

Probe on UDP Port 33434

Here is yet a third type of activity directed at the DNS server if the others have failed:

209.67.78.203.3310 > mydns.com.33434: udp 36 [ttl 1] 209.67.78.203.3311 > mydns.com.33434: udp 36 (ttl 2) 216.32.68.10.3307 > mydns.com.33434: udp 36 [ttl 1] 216.32.68.10.3308 > mydns.com.33434: udp 36 (ttl 2) 216.32.68.10.3307 > mydns.com.33434: udp 36 [ttl 1] 216.32.68.10.3308 > mydns.com.33434: udp 36 (ttl 2)

Соседние файлы в предмете [НЕСОРТИРОВАННОЕ]