Home

National Security Agency, Network Shaping 101 , January 8, 2007. Top Secret.

Na

National Security Archive

May 28, 202628 min read

A 2007 NSA training deck reveals how analysts were taught to throttle foreign traffic by exploiting BGP asymmetries, using Yemen’s ISP as a sandbox.

Source: National Security Agency, Network Shaping 101 , January 8, 2007. Top Secret. Date: Jan 8, 2007 Archive: The Intercept Collection: Cyber Vault: Delegation of Authority Oct 25, 2017


Editorial Analysis

Original analysis by the DriftSeas editorial desk. The complete primary-source document, transcribed from the National Security Archive scan, appears in full below.

NSA’s ‘Network Shaping 101’ – a classroom for cyber‑warfare

The declassified slide deck dated 8 January 2007 is not a policy memorandum or an operational order; it is a training handout for NSA personnel tasked with “shaping” traffic on foreign networks. The document’s heading—TOP SECRET//COMINT and the list of allied recipients (USA, AUS, CAN, GBR, NZL)—places it squarely within the Five‑Eyes intelligence‑sharing framework that intensified after 9/11. Its provenance, NSA/CSSM 1‑52, indicates it was produced by the agency’s Central Security Service/Management (CSSM) branch, the unit responsible for signals‑intelligence network operations.

The immediate context was the early‑2000s shift from traditional “intercept‑and‑store” SIGINT to active manipulation of data flows. By 2007 the NSA was experimenting with “network shaping” – the deliberate throttling, rerouting, or black‑holing of traffic to degrade an adversary’s communications without overtly breaking diplomatic norms. The slide deck uses Yemen’s national ISP (AS12486) as a hypothetical target, mapping its upstream providers, IP blocks, and undersea cables. While the example is fictitious, the technical details (ASN numbers, CIDR ranges, specific SMW‑3/SMW‑4 fiber links) mirror real routing data that the agency would have harvested from commercial BGP monitors and satellite‑intercept feeds.

Why Yemen? In the mid‑2000s the country was a conduit for al‑Qaeda‑linked communications and a node in the broader Red Sea‑Gulf data corridor. The NSA’s focus on a single‑ASN network reflects a strategic doctrine: control the “edge” of a hostile or suspect network by influencing the outbound path (the agency can dictate which upstream provider a router uses) while exploiting the Internet’s default inbound routing to create asymmetries. The presentation’s emphasis on Layer‑3 shaping—rather than the more granular Layer‑2—reveals a realistic assessment of the agency’s access: they could manipulate BGP policies on a compromised router but lacked the physical tap points needed for Ethernet‑level throttling.

The handout also betrays an internal culture of caution. Repeated “caveats” stress that the example IPs are “made up” and should not be used for “real analysis,” a standard disclaimer that acknowledges the sensitivity of operational knowledge while still providing a realistic training scenario. The explicit “PRO TIP” about traceroute fingerprints (identifying a hop at 144.232.234.150 as Sprint’s SMW‑3 link) illustrates how analysts were taught to verify the success of shaping operations in the field.

Reading between the lines, the document hints at a broader, covert capability: the ability to force traffic out of a country via a preferred upstream provider, then let the global BGP system route return traffic through a less favorable path, effectively creating a one‑way “choke point.” This would allow the NSA to degrade bandwidth, increase latency, or inject delays without alerting the target network’s operators, who would see only normal routing fluctuations. The fact that the deck was classified for five‑eyes allies suggests that the shaping techniques were intended for joint deployment, perhaps in coordination with allied carriers that could provide the necessary upstream links.

The legacy of this training material is evident in later public revelations about NSA’s manipulation of internet traffic, from the 2013 Snowden leaks on “upstream” collection to the 2015 “Quantum” program that altered DNS responses. “Network Shaping 101” shows that the agency’s playbook long pre‑dated those disclosures, embedding technical know‑how for active cyber‑operations into routine analyst curricula. Its declassification in 2032, a full 25 years after creation, underscores how the United States continues to view network‑level influence as a core component of its national‑security toolkit, a perspective that shapes contemporary debates over internet sovereignty and the militarization of cyberspace.


Page 1

TOP SECRET//COMINT//REL TO USA, AUS, CAN, GBR, NZL//20320108

Network Shaping 101

by [illegible]

Derived From: NSA/CSSM 1-52 Dated: 20070108 Declassify On: 20320108

TOP SECRET//COMINT//REL TO USA, AUS, CAN, GBR, NZL//20320108

Page 2

TOP SECRET//COMINT//REL TO USA, AUS, CAN, GBR, NZL//20320108

This presentation is classified:

TOP SECRET//COMINT//REL TO USA, AUS, CAN, GBR, NZL

Derived From: NSA/CSSM 1-52 Dated: 20070108 Declassify On: 20320108 TOP SECRET//COMINT//REL TO USA, AUS, CAN, GBR, NZL//20320108

Page 3

What This Will Cover

  • Caveats
  • Example network we will work with
  • What shaping would look like for that network
  • Basic shaping problems
  • A bit more advanced shaping problems
Page 4
# Initial caveats

*   To understand how to do shaping, and why it does/doesn't work sometimes, you have to go back to networking basics
*   To get the most of this presentation, you should already understand how IP's, CIDR's and Autonomous Systems (ASN) work
*   Some ips/facts are just made up. This presentation uses YemenNet as our target network. This info is outdated and incomplete. Don't use any of this information for any real analysis
Page 5
# You're gonna talk about Layer 2 shaping right?

* No
* It is extremely situational and only worth talking about if you are in a position where you have the right kind of access.
* Until then, Layer 3 shaping is where it's at (in my opinion)
Page 6
# Example network - Yemen

* Yemen has 1 ASN (AS12486)
* We'll pretend it has 6 upstream providers
  - Mobily (AS35819)
  - TATA (AS6453)
  - FLAG (AS15412)
  - PCCW (AS3491)
  - STC (AS39386)
  - SPRINT (AS1239)
Page 7
# AS12486

This network owns the following IP ranges:
46.35.64.0/19, 89.189.64.0/19, 46.35.72.0/21, 109.74.32.0/20,
46.32.80.0/21, 109.74.40.0/21, 63.168.168.0/23,
109.200.160.0/19, 63.171.18.0/23, 109.200.168.0/21

So, when we reference AS12486, you can assume it includes
any IP address that falls within any of the above ranges.
Page 8

Connectivity for this network could be viewed like: 15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

Page 9

So, for traffic to get from Yemen to the rest of the Internet (or from anywhere on the Internet to get to Yemen), it HAS to go through one of those 5 upstream providers. Which could be viewed like this...

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 10

Okay, so traffic for Yemen has to go through 1 of 6 providers, so? Armed with this high-level knowledge of Yemen’s connectivity, think about what that means:

  • Yemennet has to have a router that connects it’s own network with it’s upstream providers. That router is going to have a unique interface and IP address for each connection.
  • That router has to use physical cables to connect between Yemennet and each upstream provider (think big Transnational undersea fiber cables).
  • Yemennet CAN control which upstream provider it sends data OUT of the country through (because it controls the router that’s sending the data out).
  • Yemennet CAN NOT control which provider the data comes back IN to the country through, because that is left to BGP routing tables out on the Internet.
Page 11

So, to visualize the last 2 points...

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

Page 12

So, to visualize the last 2 points...

Yemennet can control how data gets out...

The rest of the Internet response

Yemennet does NOT control how data gets back in...

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet data

Page 13

Next, let's visualize the physical connections between Yemennet and its upstream providers. You can see here which cables are used. TOP SECRET//COMINT//REL TO USA, FVEYS Yemen Connections to the World FLAG London Sprint NY FLAG NY IRAQ JORDAN AFGHANISTAN PAKISTAN EGYPT SAUDI ARABIA OMAN SUDAN ERITREA YEMEN ETHIOPIA SOMALIA FLAG/FALCON SMW3 SMW4 SAUDI (2x STM-1's, 2x STM-4's MIDDLE EAST GRAPHIC PCCW HK TOP SECRET//COMINT//REL TO USA, FVEYS

Page 14
# So to recap:

* You understand the logical connectivity of Yemenet (who it has to go through to get to the Internet)
* You grasp the physical connectivity of Yemenet (you know which fiber cables physically connect it to the rest of the world)
* You know that Yemenet can choose which provider it sends data *OUT* through
* Big Internet BGP routing tables can dynamically choose which link data comes back *IN* to Yemenet through

* There are a couple more things to know before we talk about shaping...
Page 15

About that router that connects Yemennet to it's peers...

  • Remember how I mentioned that router has a different interface and unique IP address for each upstream provider? That router will have at least 7 interfaces (one for each upstream, and one connected to the rest of it's network).
  • The connection between the router and an upstream provider has to use IPs that are in the same subnet (normally it's a /30 subnet, which consists of 2 usable IPs).
  • This means that one of the two networks will have to sacrifice an IP address to put on the other end of the connection (most of the time it's the bigger network that gives up an IP address to assign to the customer side's router).
  • So, if we were to use the connection with SPRINT for example, here's what it might look like...
Page 16

Note that the IP address on Yemen's side of it's connection with Sprint, is an IP address that belongs to SPRINT. This is because one of the two networks has to use one of their IP's on the other end of the connection.

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet Router on Yemen's end of the connection: 144.232.234.150 Router on Sprint's end of the connection: 144.232.234.149

Page 17

PRO TIP: If you do a traceroute from somewhere in the Internet to anywhere in Yemen, if one of the hops before it gets into the country is 144.232.234.150, then you can assume your trace went through SPRINT's network (and over the SMW-3 cable if you remember the map 3 slides ago) to get into the country.

trace The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Router on Yemen's end of the connection: 144.232.234.150 Router on Sprint's end of the connection: 144.232.234.149

Page 18

For the sake of completeness, adding in the rest of the routers...

The rest of the Internet

FLAG side: 62.216.145.129 Yemen side: 62.216.145.130

PCCW side: 63.218.252.185 Yemen side: 63.218.252.186

15412 FLAG 3491 PCCW 6453 TATA 39386 STC

TATA side: 66.198.126.9 Yemen side: 66.198.126.10

STC side: 84.235.108.17 Yemen side: 84.235.108.18

35819 Mobily 12486 Yemennet 1239 SPRINT

Mobily side: 86.51.2.109 Yemen side: 86.51.2.110

Sprint side: 144.232.234.149 Yemen side: 144.232.234.150

Page 19

WARNING! WARNING!

  • In the following slides when I talk about SSO collection capabilities, I am completely MAKING UP:
    • SIGADs
    • Case notations
    • Which cables are collected
    • Where SSO's collection capabilities are
  • I am MAKING UP this info for the sake of this lesson.
  • For info on what SSO's capabilities are for your own target, you will have to go talk to them yourself.
Page 20
# WARNING 2! WARNING 2!

* For the sake of this example, I am assuming that all of Yemennet's International links are equal. By that I am making the assumptions that:
  - An equal amount of traffic is going in/out each link
  - Yemennet is not doing anything to manipulate traffic going over specific links
  - All links are actually active, and are not just backups or down due to maintenance or cable breaks...
* With that out of the way...
Page 21

So now you have a good idea about Yemen's connectivity...

  • Now time to overlay it with SIGINT collect...
  • Without going into how to do this yourself, work with SSO to determine which of those links we can passively collect.
  • Let's pretend that they have capabilities to collect the Yemen-Sprint link and the Yemen-FLAG link, but have no capabilities on the rest.
  • Once again, this is only PRETEND for the sake of this lesson.
Page 22

Here is what our passive collection capabilities would look like for Yemen...

The rest of the Internet

SIGAD: US-9999 CASN: YM234500000

15412 FLAG 6453 TATA 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

SIGAD: US-8888 CASN: YM567800000

Page 23

So, remembering what we talked about in the beginning, here's what we can tell...

The rest of the Internet

6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemen.net

Page 24

If data happens to go in or out the FLAG or Sprint links for Yemen, we will collect those comms.

Data collected at US-9999 CASN: YM234500000

rest of the I

Data collected at US-8888 CASN: YM567800000

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemen.net

Page 25

If data does NOT go in or out the FLAG or Sprint links for Yemen, we will NOT collect those comms. Data that's NOT collected. ☹ Data that's NOT collected. ☹ Data that's NOT collected. ☹ 6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet Data that's NOT collected. ☹ Data that's NOT collected. ☹

Page 26

So, now you know what our passive collection posture against YemenNet is like... we may only collect about 33% (2 out of 6 links) for all of the country's traffic (this is assuming they send and receive equal amounts of traffic over each of the links).

The rest of the Internet

SIGAD: US-9999 CASN: YM234500000

6453 TATA X

15412 FLAG

3491 PCCW X

39386 STC X

35819 Mobily X

12486 YemenNet

1239 SPRINT

SIGAD: US-8888 CASN: YM567800000

Page 27

Now that you're an expert on Yemenne, let's talk about shaping

  • The purpose of "shaping" is taking traffic that wouldn't normally go through one of our passive links, and making it go through one of our passive links, so we can collect it and get it into the SIGINT system.
  • Before we talk about how to shape traffic on Yemenne, let's explore a couple different scenarios in which we would consider shaping as a solution...they will be shaping traffic OUT of Yemenne, and shaping traffic INTO Yemenne.
Page 28
# Shaping traffic OUT of YemenNet

*   For this scenario, you have an access (probably CNE) inside of YemenNet, and you want to make that access send traffic, but make sure it goes out over a link that is passively collected by SSO.
*   You need a DESTINATION on the Internet where you can send data to, where you know it will go over 1 of the 2 links we can collect.
*   Earlier I mentioned that YemenNet can control which links they send data OUT. This is true, YemenNet has that control, however, you, as an end-user on their network do *NOT* have that control.
*   So, how can you control which link your traffic will go out through?
Page 29

If you send traffic from somewhere inside Yemen.net to some random place on the Internet, you are at Yemen.net's mercy as to which link it will send the data out through.

The rest of the Internet

6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemen.net

Exfil data

Page 30

Uh oh, you tried to exfil traffic, but Lady Luck did not shine her favor upon you, and the traffic went out a link we could not passively collect...

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

Page 31

SUCCESS! The exfil will probably die somewhere in Sprint's network, but we don't care, because we accomplished our goal of collecting it along the way. Now we can go look at collect from US-8888 CASN: YM567800000 for our exfil!

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Exfil data

Page 32

So in order to shape traffic OUT of a network

  • You need to have an understanding of the network that you are starting in, who it's upstream providers are, and what the collection capabilities are against that network.
  • Then you can find a destination IP address directly on the other end of that link (by looking at any of the IP ranges in that provider's ASN).
  • From there, you have a higher probability that traffic will traverse a link you can passively collect.
Page 33

Shaping traffic INTO a network

  • This is a whole different animal, and probably more relevant to what people traditionally think of as “shaping” in the SIGINT sense.
  • There is only 1 feasible way (that I can think of) to make this work reliably.
  • But first, let’s go back and look at our Internet connectivity...
Page 34

In order to shape traffic into a network, we'll be starting from some random place on the Internet. Somewhere in here...

The rest of the Internet

6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 35

And our goal, will be to send traffic into YemenNet with the express purpose of making it go through 1 of the 2 links we can passively collect...because the entire point is getting our exfil into the SIGINT system through those passive access points.

The rest of the Internet

6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 12486 YemenNet 1239 SPRINT

Page 36

But remember, the route for traffic going into YemenNet is completely at the mercy of BGP routing tables out on the Internet. So there is no guarantee that it will go through one of the links we can collect (once again, only a 33% shot).

The rest of the Internet 15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 37

But remember, the route for traffic going into Yemenent is completely at the mercy of BGP routing tables out on the Internet. So there is no guarantee that it will go through one of the links we can collect (once again, only a 33% shot).

Exfil traffic The rest of the Internet 15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemenent

Page 38

Wow, that didn't work. ☹ So, if we are completely at the mercy of big Internet routing tables, how could we reliably send traffic from ANYWHERE in the Internet into YemenNet while guaranteeing it goes through passive collect?

The rest of the Internet

6453 TATA 15412 FLAG 3491 PCCW 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 39

There are a few options you can use to try to make that happen...

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 40

1 – You can try to tweak the BGP routing tables to make your 2 links the most attractive for inbound traffic... The rest of the Internet 15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

Page 41

That could work, but 2 bad things would result from this: 1 – It’s a pretty noisy thing to do on the Internet. People would notice bad BGP updates. 2 – You would throttle ALL Internet traffic through those 2 links, which YemenNet would probably notice.

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 YemenNet

Page 42

Or...you could blow up or cut all of the International links that we can't collect! ☺ That's fun to think about, but not very reasonable. ☹

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemennet

Page 43

There is another way that might work...we'll have to go back to one of our previous maps though...

The rest of the Internet

15412 FLAG 3491 PCCW 6453 TATA 39386 STC 35819 Mobily 1239 SPRINT 12486 Yemenet

Page 44

Remember the /30 connections between Yemen and each of the upstream providers... The rest of the Internet FLAG side: 62.216.145.129 Yemen side: 62.216.145.130 PCCW side: 63.218.252.185 Yemen side: 63.218.252.186 6453 TATA 15412 FLAG 3491 PCCW 39386 STC TATA side: 66.198.126.9 Yemen side: 66.198.126.10 STC side: 84.235.108.17 Yemen side: 84.235.108.18 35819 Mobily Mobily side: 86.51.2.109 Yemen side: 86.51.2.110 12486 Yemennet 1239 SPRINT Sprint side: 144.232.234.149 Yemen side: 144.232.234.150

Page 45

It actually looks a bit more like this... The rest of the Internet FLAG side: 62.216.145.129 PCCW side: 63.218.252.185 15412 FLAG 3491 PCCW 6453 TATA 39386 STC TATA side: 66.198.126.9 STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 YemenNet

Page 46
  • Whoa, that looks kinda messy, what do I really need to take away from this slide?
  • Good question, just keep in mind the fact that each provider has put one of their own IP addresses in Yemen for those connections. Here's why it matters...

The rest of the Internet

FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 YemenNet

Page 47

Remember, if you do a traceroute to a random IP address in AS12486, you won't know which link it will go through to get there... The rest of the Internet FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 YemenNet

Page 48

But, hypothetically, what happens if you do a traceroute to 86.51.2.110 (the Yemen side of its Mobily connection) from some random place on the Internet?

trace The rest of the Internet

FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 Mobily side: 86.51.2.109 12486 YemenNet 1239 SPRINT Sprint side: 144.232.234.149

Page 49

-Ok, it went where I expected it to go...what's so special about that?

  • First, keep in mind the IP we traced to belongs to Mobily, not Yemen net. So the traffic will first get routed to Mobily's network. From there, it happens to reside on a router belonging to Yemen net.

trace The rest of the Internet FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 Yemen net

Page 50

-AAAAAND?

  • By tracing to the Mobily IP on Yemenne's router, we forced the trace to go through Mobily to get there. What happens if we did that for FLAG's IP address?

trace The rest of the Internet FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 Yemenne

Page 51

Starting traceroute from anywhere on the Internet to 62.216.145.130...

The rest of the Internet

FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 Mobily side: 86.51.2.109 1239 SPRINT Sprint side: 144.232.234.149 12486 Yemennet

Page 52

Starting traceroute from anywhere on the Internet to 62.216.145.130... trace The rest of the Internet FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 1239 SPRINT Mobily side: 86.51.2.109 Sprint side: 144.232.234.149 12486 Yemennet

Page 53

We have just successfully forced traffic to go through FLAG's network and over a link that we can collect! So, it seems we have found a good candidate to shape traffic to from anywhere! trace The rest of the Internet FLAG side: 62.216.145.129 15412 FLAG PCCW side: 63.218.252.185 3491 PCCW 6453 TATA TATA side: 66.198.126.9 39386 STC STC side: 84.235.108.17 62.216.145.130 62.218.252.186 66.198.126.10 84.235.108.18 35819 Mobily 86.51.2.110 144.232.234.150 Mobily side: 86.51.2.109 12486 YemenNet 1239 SPRINT Sprint side: 144.232.234.149

Page 54

Scenario – “I tried shaping one time and it didn’t work...”

  • As we’ve seen earlier, there are many facets that make shaping efforts unreliable
  • It matters whether you are trying to shape traffic OUT of a network or whether you are starting at a random place on the Internet and trying to shape traffic INTO the network
  • So what steps could you take?
Page 55
# So, you might consider the following train of thought...

* First you say, “I want to do shaping through SIGAD US-9999”
* Then look at all of the links collected at that site (probably in BLACKPEARL)
* Find a World-to-Geekistan link over CASN GE010100000 (we’ll assume you also know that this is actually a Level 3-to-Geekistan link)
* Look at the IP space on the dest side of the link and say, “I will send my exfil to that IP space, and it should go through US-9999, CASN GE010100000.”
* Then you are left sorely disappointed when your exfil isn’t reliably collected.
* What went wrong? Let’s consider what we know so far...
Page 56

This isn't a bad start for network knowledge, but there's still some missing pieces...

Internet

Level 3

Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 57

For example, do you know if Level 3 is the ONLY upstream provider for Geekistan? Or are there other ways for traffic to get in and out of that network? How would you find out?

Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 58

You could look at BGP to find upstream providers for Geekistan ASN. Cogent Sprint Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 59

You could do traceroutes from random places on the Internet to IP's in Geekistan's network and see who it goes through to get there.

Cogent Sprint Internet Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 60

You could do traceroutes from random places on the Internet to IP's in Geekistan's network and see who it goes through to get there.

Cogent Sprint trace Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 61

Now with this information, what do you do next?

trace Internet

Cogent Sprint Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 62
If you are bound and determined to shape traffic into Geekistan,
you need to do 1 of 2 things...

Internet
Cogent
Sprint
Level 3
Geekistan ASN
SIGAD – US-9999
CASN – GE010100000
Page 63

1 – You'll need to find passive collection on both the Cogent and Sprint links with Geekistan. If you do, then you can have confidence in collecting exfil as you shape it into that network.

Cogent Sprint Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 64

2 – You'll need to identify the Level 3 IP address that is on the Geekistan side of the connection, and send exfil directly to that IP address...

Cogent Sprint exfil Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 65

If you do your research, and realize that Level 3 is the only connection Geekistan has to the rest of the Internet, and you are collecting that link, then you're safer to assume your exfil will get collected if you send it to anywhere in that network...

Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 66

If you do your research, and realize that Level 3 is the only connection Geekistan has to the rest of the Internet, and you are collecting that link, then you're safer to assume your exfil will get collected if you send it to anywhere in that network...

Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 67

Now to the nittier-grittier...

  • This following section could also be renamed the “I’m pulling my hair out in the fetal position while screaming ‘Why didn’t it work?!’” section.
  • The previous slides described how shaping should work at a theoretical level, following are a few reasons why it doesn’t always work in the real world.
  • The following issues are not all-encompassing of why shaping might not work, just a few examples.
Page 68

First, the multiple links problem. Below, the graph looks really pretty with 1 link per upstream provider, but that isn't always the case. Sometimes it's like this:

Cogent Sprint Internet Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 69

Notice how there are 2 links between Geekistan ASN and Level 3. It is not abnormal for ASNs to have multiple links to eachother for redundancy reasons.

Cogent Sprint Internet Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 70

Why this matters, is if we are only able to collect 1 of the 2 links, even if we try to force our exfil into/out of the Level 3-Geekistan ASN link, we can't guarantee it will traverse the link we collect.

Cogent Sprint exfil Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 71

Even if we try to shoot directly at the Level 3 interface that we can collect on, there's a chance it could go over the other Level 3 link to get there...

Cogent Sprint exfil Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 72

Next, there is the "back-up" link problem. In the ISP world, bandwidth is money, and some people charge more for letting your data traverse their networks...

Internet Cogent Sprint Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 73

Say, for example, it was twice as expensive for Geekistan to send data through Level 3 as it is for Cogent or Sprint. They would use Level 3 as little as possible (as in, only when necessary), while using the other 2 as much as they can.

Cogent Sprint Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 74

If they try to force traffic in/out Cogent and Sprint, that could spell bad news for our shaping efforts, we might see something like this:

Internet Cogent Sprint Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 75

If we try to send data directly over the Level 3 link, we might be sorely disappointed with Geekistan forcing us over a different link...

exfil Internet Cogent Sprint Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 76

Or it might look something like this if we tried to exfil out directly into Level 3's network...

Internet

Cogent Sprint Level 3

Geekistan ASN exfil

SIGAD – US-9999 CASN – GE010100000

Page 77

Lastly (for now), is the infamous collection problem...

Internet Cogent Sprint Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 78

Pretend all the stars aligned...there is only 1 Level 3 link, we can send data over it, and we do have access to it at SIGAD US-9999...

Cogent Sprint Internet Level 3 Geekistan ASN

SIGAD – US-9999 CASN – GE010100000

Page 79

In a fit of joy we start sending exfil over that link. Then we go look at US-9999, GE010100000 for our exfil, and find that it isn't there!

Internet Cogent Sprint Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 80

Without going into the gory details, US-9999 may be dropping your exfil (as in, 'able to see it, but not collecting and processing it'). If you think this is the case, please coordinate with SSO to make sure appropriate IP's/protocols are promoted at the site.

Cogent Sprint Internet Level 3 Geekistan ASN SIGAD – US-9999 CASN – GE010100000

Page 81

Contact deets... [illegible] [illegible] [illegible]

Page 82

NATIONAL SECURITY ARCHIVE

National Security Archive, Suite 701, Gelman Library, The George Washington University, 2130 H Street, NW, Washington, D.C., 20037, Phone: 202/994-7000, Fax: 202/994-7005, nsarchiv@gwu.edu

Keywords

declassifiedNational Security ArchiveCyber Vault: Delegation of Authority Oct 252017

Keep reading

More related articles from DriftSeas.