first commit
This commit is contained in:
commit
0d90363116
6510 changed files with 679146 additions and 0 deletions
209
data/HIMR-Oct2011/HIMR-Oct2011.md
Normal file
209
data/HIMR-Oct2011/HIMR-Oct2011.md
Normal file
|
|
@ -0,0 +1,209 @@
|
|||
TOP SECRET STRAP1
|
||||
|
||||
# Graph theory in the operational environment
|
||||
|
||||
[REDACTED] - GCHQ
|
||||
|
||||
Information & Communications Technology Research (ICTR)
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## What I will cover
|
||||
|
||||
* Finding operational closed loops:
|
||||
* The reality of target behaviour
|
||||
* The process of looking for closed loops
|
||||
* Some statistics on a large graphs components
|
||||
* A real life example
|
||||
* Cheap and disposable handsets
|
||||
* Catching targets for real
|
||||
* What Donald Rumsfeld taught me about closed loops
|
||||
* Using graphs to visualise and characterise timing relationships in a contact graph
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## What is a closed loop?
|
||||
|
||||
* A set of (for example) phones that communicate only amongst themselves as a means of communications security
|
||||
* This is generally speaking too loose a definition for practical purposes
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Tightening up the definition of a closed loop
|
||||
|
||||
* A component on three or more nodes that is neither the giant component nor a tree
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Work on this topic since 2005
|
||||
|
||||
* GCHQ initially interested in topic following use of a closed loop by the July 7, 2005 bombers
|
||||
* Analysis of anonymised meta-data for bulk UK-UK mobile call records indicated that this was a rare phenomenon and pointed to possible target discovery opportunities
|
||||
* Closed loop analysis of VOICESAIL showed promise but work was truncated
|
||||
* SANAR-08 presentation demonstrated that operational closed loops could in theory be discovered
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## The reality of target behaviour
|
||||
|
||||
* Targets from different IPT's regularly purchase groups of cheap mobile phones and use them operationally for a short period (possibly as long as three months) before getting a new set
|
||||
* Two critical features:
|
||||
* Most of the phones start life at about the same time
|
||||
* Most of the phones are cheap handsets
|
||||
* Sometimes the targets make a mistake and make a call to a phone outside the closed loop
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## The process of looking for closed loops
|
||||
|
||||
* Choose a time window (e.g. 10 days)
|
||||
* Clean up the data for that period
|
||||
* Componentize the graph
|
||||
* Extract the components that fit the closed loop definition
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## The effect of windowing
|
||||
|
||||
* We need to window the data or else slip ups by the closed loop members will render the group invisible
|
||||
* Need to choose a window that is large enough to allow the giant component to form but not so large that we never see targets who periodically goof up.
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Some statistics about components
|
||||
|
||||
* 1507405 edges and 1606330 nodes
|
||||
* 922135 nodes in the giant component (57%)
|
||||
* 222477 components, of which 218327 (98.1%) are trees
|
||||
* 4149 non-tree, non-giant components, involving 39927 nodes (2.5% of total)
|
||||
* 17 of these components had at least 70% of the nodes illuminating within a 3 week period.
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## A closed loop is born...
|
||||
|
||||
```
|
||||
Untitled 1
|
||||
File Edit Interface Selection Analysis Arrange Display Clustering Data Control Window Developer Help
|
||||
Mouse Mode
|
||||
|
||||
Base Network
|
||||
* Nodes 5
|
||||
* Links (Total) 325
|
||||
* Links (Composite) 10
|
||||
* Excluded Nodes 1
|
||||
* Excluded Links 0
|
||||
* Components 1
|
||||
* Singulars 0
|
||||
|
||||
XXXXXX1851
|
||||
XXXXXX8595
|
||||
XXXXXX7496
|
||||
XXXXXX3911
|
||||
XXXXXX9539
|
||||
```
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## When geolocation is not enough...
|
||||
|
||||
* Targets love cheap handsets, bless 'em
|
||||
* Nokia occupy the largest segment of the cheap phone market. Nokia 1* phones are nasty.
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Nokia 1616
|
||||
|
||||
"Unashamedly aimed at the bottom end of the mobile phone spectrum... under the bonnet things remain distinctly unimpressive”
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Vodafone 252
|
||||
|
||||
“A very affordable handset that comes with basic voice call and message services"
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Samsung 1150
|
||||
|
||||
"This cut-price phone offers bare-bones functionality and lacks what many mobile users would deem to be essential features”
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
|
||||
---
|
||||
|
||||
TOP SECRET STRAP1
|
||||
|
||||
## Testing the cheap phone hypothesis
|
||||
|
||||
* What type of handset did the July 7, 2005 London bombers use?
|
||||
* At least three of the four phones used on that day were a Nokia 1100
|
||||
|
||||
© Crown Copyright. All rights reserved.
|
||||
ICT RESEARCH
|
||||
80
data/HIMR-Oct2011/metadata.json
Normal file
80
data/HIMR-Oct2011/metadata.json
Normal file
|
|
@ -0,0 +1,80 @@
|
|||
{
|
||||
"title": "Graph theory in the operational environment – GCHQ",
|
||||
"type": "REPORTED",
|
||||
"description": "This document details GCHQ’s research on identifying “closed loops” of communication, where targets use a small set of phones to communicate exclusively among themselves, often with cheap, disposable handsets like the Nokia 1100. It explains the methodology for analyzing anonymized metadata, constructing contact graphs, and extracting non-giant, non-tree components to discover operationally relevant closed loops for intelligence and counterterrorism purposes.",
|
||||
"long_description": "The document presents GCHQ’s technical research into detecting “closed loops” — small groups of phones that communicate almost exclusively among themselves — using graph-theory methods applied to large anonymized mobile-call metadata sets. It explains the operational motivation (notably sparked by the July 7, 2005 London bombers who used cheap, disposable handsets), defines a practical closed‑loop criterion (components of three or more nodes that are neither the giant component nor trees), and lays out a step‑wise analytic workflow: choose an appropriate time window, clean the data, componentize the contact graph, and extract candidate components that fit the closed‑loop profile. The slides include empirical statistics from a very large dataset (millions of nodes and edges) demonstrating that closed‑loop components are rare but detectable and that a small subset show concentrated activity within short time windows — the sort of signal analysts would prioritize for further investigation.\n\nMethodologically the paper emphasizes the importance of temporal windowing and parameter choice: too large a window merges innocuous activity into the giant component, while too small a window risks missing intermittent slippage by loop members. Practical issues are addressed — handset churn (groups replacing cheap phones frequently), the dominance of low‑end models (Nokia 1100 et al.) in suspected cases, and the need to tolerate occasional “mistakes” (calls outside the loop) when tuning detection thresholds. The document also illustrates the approach with a worked example and quantitative measures (component counts, proportions of tree vs. non‑tree components), showing how graph metrics and visualization tools can turn vast metadata holdings into manageable, operational leads.\n\nThe write‑up implicitly highlights both the power and limits of metadata analytics: closed‑loop detection can surface tightly coordinated groups that evade ordinary SIGINT, offering a route to “target discovery” where content interception fails, but it is also probabilistic and error‑prone.",
|
||||
"tags": [
|
||||
"mobile phones",
|
||||
"cell phones",
|
||||
"telecommunications",
|
||||
"closed loop analysis",
|
||||
"graph theory",
|
||||
"cellphone metadata",
|
||||
"call graph",
|
||||
"operational closed loops",
|
||||
"target discovery",
|
||||
"cheap mobile phones",
|
||||
"disposable handsets",
|
||||
"Nokia",
|
||||
"Vodafone",
|
||||
"Samsung",
|
||||
"mobile networks",
|
||||
"SANAR-08",
|
||||
"IPT targets",
|
||||
"pattern-of-life",
|
||||
"node analysis",
|
||||
"edge analysis",
|
||||
"temporal windowing",
|
||||
"componentization",
|
||||
"non-tree components",
|
||||
"giant connected component",
|
||||
"network visualisation",
|
||||
"signals analysis",
|
||||
"metadata analysis",
|
||||
"contact chaining",
|
||||
"telecom research",
|
||||
"target analysis",
|
||||
"telephony",
|
||||
"network analysis",
|
||||
"metadata processing",
|
||||
"pattern recognition",
|
||||
"phone churn",
|
||||
"target identification",
|
||||
"ICT Research",
|
||||
"London bombings"
|
||||
],
|
||||
"document": {
|
||||
"creator": "GCHQ",
|
||||
"date_created": "2011-10-01",
|
||||
"classification": [
|
||||
"TOP SECRET",
|
||||
"STRAP 1"
|
||||
],
|
||||
"dissemination": [],
|
||||
"banner": "TOP SECRET STRAP1",
|
||||
"genre": "slides",
|
||||
"programs": [
|
||||
"VOICESAIL"
|
||||
],
|
||||
"countries": [],
|
||||
"people": [
|
||||
"Rumsfeld, Donald"
|
||||
]
|
||||
},
|
||||
"publication": [
|
||||
{
|
||||
"publisher": "The Intercept",
|
||||
"date_published": "2018-03-01",
|
||||
"title": "How London's 7/7 Bombings Led to \"Unprecedented\" Surveillance Tactics",
|
||||
"link": "https://theintercept.com/2018/03/01/london-7-7-bombings-gchq-nsa-surveillance/",
|
||||
"reporters": [
|
||||
"Gallagher, Ryan"
|
||||
],
|
||||
"contributors": [],
|
||||
"doc_source": "https://theintercept.com/document/2018/03/01/graph-theory-in-the-operational-environment-gchq/",
|
||||
"doc_url": "https://s3.documentcloud.org/documents/4390404/HIMR-Oct2011.pdf"
|
||||
}
|
||||
],
|
||||
"pdf_sha256": "d7b4011fc6c841dbc85a3ed8c808dc5e2e8bd8dbcba9906635e92ec31b5e0572",
|
||||
"pages": 15
|
||||
}
|
||||
Loading…
Add table
Add a link
Reference in a new issue