Crypt Discovery Joint Collaboration Activity 20 January 2011

Open PDF
Creating agency: NSA National Security Agency (US)
Document date: 2011-01-20
Pages: 3
Genre: other
Banner:
Classification: TOP SECRET
Dissemination: -

A TOP SECRET NSA/GCHQ joint collaboration activity plan from January 2011 describing a research programme to understand and mitigate the threat that rising encryption adoption poses to target discovery and development — measuring encryption prevalence across SIGINT collection, assessing the impact on metadata-based targeting tradecraft, and developing new techniques for identifying targets behind encrypted traffic using a joint cloud environment fed by MUTANT BROTH, BEARDED PIGGY, KARMA POLICE, MEMORY HOLE, MARBLED GECKO, SOCIAL ANIMAL and ASDF metadata.

Long Description:
This is a TOP SECRET NSA/GCHQ joint collaboration activity plan dated 20 January 2011, with activity owners from both NSA and GCHQ. The document describes a structured research programme addressing what both agencies identified as a growing threat to their internet exploitation capability: the spread of TLS/SSL and IPSec encryption making the metadata they rely on for target discovery and development progressively harder to access. The plan is organised around four research workstreams. The first involves monitoring the prevalence of encrypted traffic across all SIGINT collection — measuring what percentage of all observed traffic is encrypted, what percentage of encrypted traffic is associated with known targets, which encryption technologies targets use, how usage breaks down across webmail, search, mapping, forums, instant messaging, social networks and banking, and what the trend lines look like over time. The second workstream assesses the specific threat to existing capabilities: which Target Detection Identifiers and presence metadata will be lost inside encrypted channels, what the impact is on target discovery and development tradecraft, and how cross-correlation and complex query capabilities are affected. The third workstream examines pairing rates — the ability to correlate encrypted sessions seen at one collection point with sessions seen at another, assessing what can be paired within and between access points, whether CA (cryptanalytic) services could handle the volumes involved, and how cloud analytics could improve pairing by looking across time. The fourth workstream researches new methods for maintaining effective target discovery tradecraft despite encryption — including developing new TDI-like identifiers that persist across sessions behind HTTPS, characterising IPSec sessions from ESP metadata analysis using duration, packet length and burstiness, and combining content with metadata for better session identification. An additional workstream addresses the mobile internet context specifically — assessing how much more challenging target discovery is behind mobile gateways and whether mitigations developed for fixed internet apply. The technical requirements section lists the data sources available in the Joint Collaboration cloud environment: MUTANT BROTH, BEARDED PIGGY, KARMA POLICE, MEMORY HOLE, MARBLED GECKO, SOCIAL ANIMAL and ASDF metadata, alongside selected content related to encryption or specific target sets, new metadata creation capability through XKEYSCORE or local processing, and a CA server internal to the JC environment. The document also references TICKETWINDOW sites as a named collection environment used for pairing correlation demonstrations. Taken together the document provides a candid internal assessment of how seriously NSA and GCHQ viewed the encryption problem in 2011 — not as a capability already defeated but as a research challenge requiring sustained joint investment, with the two agencies pooling metadata sources and cloud analytics infrastructure to develop new targeting tradecraft before rising encryption adoption made their existing methods obsolete. In plain terms this is a joint NSA/GCHQ research project plan addressing a problem both agencies were worried about: as more and more internet traffic became encrypted, the identifying information they relied on to find and track targets was disappearing inside those encrypted connections. The document lays out a programme to measure how bad the problem was, figure out what intelligence they were losing, and develop new ways of identifying suspects even when the content of their communications was hidden. It is significant as a direct statement from both agencies that encryption was genuinely impeding their work — and that they were investing heavily in finding ways around it.
Programs:
TICKETWINDOW, MUTANTBROTH, BEARDEDPIGGY, KARMAPOLICE, MEMORYHOLE, MARBLEDGECKO, SOCIALANIMAL, XKEYSCORE, ASDF

People mentioned:
-

Countries:
-

Tags:
Joint Collaboration, encryption, target discovery, target development, TLS, SSL, IPSec, HTTP, VPNs, ESP metadata, pairing rates, mobile internet, mobile gateway

Identifier:
spiegel_-_media-35548

PDF SHA-256:
ba76c51d66d616e36f2a15569ddfb4bf99dbae14d5751b09e6af4b7b68845cbe

TOP SECRET

Crypt Discovery
Joint Collaboration Activity
20 January 2011

Activity Owners
[REDACTED] NSA
[REDACTED] GCHQ

Overview

Our Internet Exploitation capability is built upon our ability to effectively conduct target development and target discovery from IP data. One target discovery strategy is to spot patterns of Internet activity (communications, web browsing, web searching, etc) from which analysts infer suspicious behaviour or intent. Developing Target Detection Identifiers (TDIs), Internet presence data and other key metadata elements are critical enablers for achieving success.

The spread of encryption, particularly Transport Layer Security (TLS), which was previously known as Security Sockets Layer (SSL), and Internet Protocol Security (IPSec), threatens our ability to do effective target discovery/development because pertinent metadata events will be locked within the encrypted channels and difficult, if not impossible, to prise out.

These activities are designed to get a sound understanding of the threat that encryption brings to our ability to do target discovery/development as well as devising mitigations that will (hopefully) allow our Internet Exploitation strategy to prevail.

Plan

Performing discovery within and beyond encryption will require new solutions. We will be investigating and creating solutions in these areas:

  1. Monitor the prevalence of encrypted usage within encryption technologies (https, vpn, etc) and across the Internet and into our SIGINT targets' domain. The questions we would like to get into a position to answer are:
    a. What percentage of encryption are we seeing in all traffic? What is the trend over time?
    b. What percentage of encrypted traffic is associated with a target? Trends?
    c. What percentage of targets are using TLS, IPSec, or other encryption technologies? Trends?
    d. How are these figures above broken out across the encrypted web services available (webmail, web searches, geo mapping, web fora, IM, social networking, online shopping, online banking, etc)? Split out the webmail stats per provider. Trends?
    e. What are the most common https sites visited (in all traffic)? Trends?
    f. What are the most common https sites visited by targets? What is the profile of usage? Trends?

Derived From: NSA/CSSM 1-52
Dated: 20070108
Declassify On: 20360101

TOP SECRET

TOP SECRET

  1. Assess the threat of https to our current exploitation capabilities and to our future intent
    a. Which of our presence and communications metadata (including TDIs) will be lost under encrypted channels?
    b. What is the impact of this loss to our Target Discovery and Target Development tradecraft?
    c. What is the impact of this loss to our ability to cross-correlate TDIs and run more complex queries against our data?
  1. Understand and improve our pairing rates within an access and between access
    points?
    a. What can be paired from what can be viewed?
    b. What could be paired between sites (demonstrate correlation between collaboration environment and TICKETWINDOW sites for example)?
    c. What are the volumes of paired encrypted links and would our CA services be able to handle these?
    d. Can we optimise our passive collection by analysing pairing rates per country and per service level?
    e. How can the cloud enable better pairing by leveraging it's ability to look over time?
  1. Research new methods for maintaining an effective TD tradecraft and mitigating the threat of encrypted http traffic.
    a. Are there other TDIs or "TDI-like” identifiers we can develop that can identify users or machines of interest, despite our targets using https? How effective are these? (e.g. are the TDIs persistent across sessions, do they uniquely identify a user or machine, do we always appear in user sessions?)
    b. Can we characterise IPSec sessions from ESP metadata analysis (e.g. using duration, packet length, burstiness)?
    c. How can we best analyze decrypted VPN traffic to develop effective TD tradecraft for VPNs.
    d. Can we use Internet presence or communications metadata with network analysis techniques to identify high-value sessions? (gah: do you think this sits better under (3) above?)
    e. Can we combine knowledge from session processing and content with metadata to enable better methods.

Acquire an understanding for how much more challenging (or easier!) this problem is in the Mobile Internet context, and whether any of the mitigations from (4) above lend themselves to discovering targets behind Mobile Gateways.

2 of 3

This information is exempt from disclosure under the Freedom of Information Act 2000 and may be subject to exemption under other UK information legislation. Refer disclosure requests to GCHQ on [REDACTED]

TOP SECRET

TOP SECRET

Technical Requirements

The ability to understand and track encryption as well as develop new methods as outlined above requires the use of existing metadata from NSA and GCHQ processing system, the ability to create new metadata to test the viability of it's use for the creation of new methods and to enhance existing data, the ability to correlate across what is known (strongly selected content) with metadata and potentially the ability to examine deeper into sessions for evaluation of exploitability. To accomplish the goals above, the following will be used in the JC environment:

  • A Joint Collaboration (JC) cloud that contains the following metadata sources:
    • MUTANT BROTH
    • BEARDED PIGGY
    • KARMA POLICE
    • MEMORY HOLE
    • MARBLED GECKO
    • SOCIAL ANIMAL
    • ASDF metadata
  • Selected content
    • Related to encryption or specific target sets
  • Ability to create new metadata or selected content as needed through XKS or other local processing sources.
  • Inclusion of enrichment data
    • TBD
  • Ability to interact with CA server internal to the JC environment

3 of 3

This information is exempt from disclosure under the Freedom of Information Act 2000 and may be subject to exemption under other UK information legislation. Refer disclosure requests to GCHQ on [REDACTED]

TOP SECRET