Tracking Targets Through Proxies & Anonymizers
Open PDFDocument date: -
Pages: 25
Genre: slides
Banner: TOP SECRET//COMINT//REL TO USA, FVEY
Classification: TOP SECRET SI
Dissemination: REL 5EYES
This NSA presentation details methods for tracking and de-anonymizing internet users who employ privacy tools like Tor, proxies, and VPN services such as AnchorFree's Hotspot Shield. The document explains how the agency reverse-engineers anonymization tools, creates "fingerprints" to identify their traffic in surveillance systems, and correlates encrypted connections to reveal users' real IP addresses and locations—acknowledging this work "makes for sad analysts" when targets successfully obscure their identities.
My notes
TOP SECRET//COMINT//REL TO USA, FVEY
Tracking Targets Through Proxies & Anonymizers
(and the air speed velocity of an unladen swallow)
Also known as:TTTPaAatasvoaus
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
CLASSIFICATION
- Classification of this briefing is:
TS//SI//REL 5EYES
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Agenda Items
- The issue at hand... (proxies, anonymizers & TOR, oh my!)
- What we do about, and how we approach, this issue...
- A couple examples of tracking targets through anonymizers (AnchorFree & Tor)
- Closing remarks and questions
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Up Front Caveat
Before we begin this briefing, I want to set the stage by saying that there is no silver bullet for tracking target communications through anonymizers. Any methodology set forth in this briefing requires both manual analysis and (generally) luck.
With that out of the way...
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
The issue at hand...(proxies, anonymizers & Tor, oh my!)
- Targets generally don't like to have their communications tracked by government agencies or filtered by national firewalls.
- If they are tech-savvy enough, they will use anonymizers to try to mask their real IP/ location.
- This generally makes for sad analysts. :(
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
What we do about the issue at hand...
- The only way to track communications through anonymizers is if you understand how those anonymizers work. If you don't know what the traffic looks like, how will you recognize it in SIGINT?
- Generally our process is as follows:
- Identify new proxy/anonymizer
- Research/use anonymizer, document what happens, what does traffic look like, what client traffic does it pass through(if any)?
- Create fingerprints in SIGINT to identify such proxy traffic
- Correlate proxy traffic with known target activity
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
AnchorFree owns a bunch of servers on the Internet.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
Then you have a user that downloads Hotspot Shield to proxy their traffic...
AnchorFree owns a bunch of servers on the Internet.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
Then you have a user that downloads Hotspot Shield to proxy their traffic...
AnchorFree owns a bunch of servers on the Internet.
We'll pretend they want to go to Yahoo (or whereever else)
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
When the user starts HSS, their browser sets up an encrypted connection to a randomly picked server...
AnchorFree owns a bunch of servers on the Internet.
We'll pretend they want to go to Yahoo (or whereever else)
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
When the user starts HSS, their browser sets up an encrypted connection to a randomly picked server...
AnchorFree owns a bunch of servers on the Internet.
We'll pretend they want to go to Yahoo (or whereever else)
Then they access the webpage from the AnchorFree IP address.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Anchorfree
From testing, the IP address that the user connects to, and the IP they show up as are NOT the same. But there is a direct correlation between the two.
When the user starts HSS, their browser sets up an encrypted connection to a randomly picked server...
Then they access the webpage from the AnchorFree IP address.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
AnchorFree....so what?
We can build static mappings between the inside/tunnel IP address and the IP they show up as...
So when we see a target access their account from an AnchorFree IP, we know which IP to go look for tunnel connections to in order to find their real client IP address...
We can also write XKS fingerprints to look for AnchorFree tunnels en masse from interesting locations.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
2) The user then sets up layers of SSL tunnels through them all to get to facebook. A layer is stripped off along each hop.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
2) The user then sets up layers of SSL tunnels through them all to get to facebook. A layer is stripped off along each hop.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
2) The user then sets up layers of SSL tunnels through them all to get to facebook. A layer is stripped off along each hop.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
2) The user then sets up layers of SSL tunnels through them all to get to facebook. A layer is stripped off along each hop.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Tor
Example of Tor user connecting to facebook.com
1) the user selects 3 relatively random Tor nodes to use.
2) The user then sets up layers of SSL tunnels through them all to get to facebook. A layer is stripped off along each hop.
TOR user
Facebook.com
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Interesting tidbits about Tor
- TOR uses SSL tunnels for encryption, we are able to identify what their SSL certificates look like (which allows us to identify Tor circuits in SIGINT).
- GOLDENFORTIN dataset and exit node traffic
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
And now for something completely different...
- A lot of research we do on anonymizers consists of open source research:
- The Interwebz (forums, RFC's, 2010 Circumvention Tool Usage Report, etc)
- Trial and Error / Wireshark
- Basically RTFM'ing about how stuff works and translating that to the SIGINT system.
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Contact Info
TOP SECRET//COMINT//REL TO USA, FVEY
TOP SECRET//COMINT//REL TO USA, FVEY
Questions?
NOBODY EXPECTS THE SPANISH INQUISITION!!!!
TOP SECRET//COMINT//REL TO USA, FVEY