[Back to Keyring Analysis Page]

Key Analysis 10 Aug 2001

The following stats are being pulled from a keyring that was exported from pgp.dtype.org on August 9, 2001. Before reading this, please be sure to view the explanation of this analysis and read the FAQ before asking me any questions about it.

The strong set MSD raw analysis is available here. Please read the FAQ to explain how to read this file. This file includes all keys reachable from the strong set. Look up reports for individual keys in the raw output directory. Now you can see what keys are signed by each key (otherwise very difficult to find).

• Output directory, including individual key reports

New This Month

Thanks to Randy Harmon at NAI (pgp.com) for a full dump, and a significant increase in the keyring size. Much of the growth of the keyring came from merging this ring. As always, continued thanks to Peter Wan for his efforts in maintaining global rings.

The individual key analysis now also includes hop count histograms to each key from the strong set.

Special Report - Other Strongly Connected Sets

As mentioned in the explanation, I have chosen to focus on what is by far the largest strongly connected set, which I call the "strong set". There are, however, other smaller sets which are strongly connected. This means that groups of friends are reachable from each other, but have not exchanged signatures with anyone in the larger "strong set".

This month, I've made a scan for _all_ strongly connected sets, from the largest, at size 10,153, to the n-way tie for last place with strongly connected sets of 1.

• Raw file with lists of all strongly connected sets

As you can see, the largest strongly connected set is orders of magnitude larger than the next biggest (10,153 compared to 98), which at least somewhat validates the decision to focus on this set. Other relatively large strongly connected sets could easily merge with the larger one by simply having _any_ of the set's members exchange signatures with someone in the larger set.

Again, just because there is a path between two keys does not mean that there should be any implied trust. The reason this matters to this report is the math behind these numbers relies somewhat on strongly connected sets.

General statistics

Size of binary keyring (bytes):		1,853,471,920	(+7.52%)
Number of keys:		1,577,742	(+7.13%)
Non-revoked keys with at least one non-self sig:		147,701	(+5.82%)
Total non-self sigs on those keys:		302,662	(+5.72%)

The "strong set"

Size of largest strongly connected set:		10,153	(+6.18%)
Keys that have signed this set:		14,811	(+6.04%)
Keys that this set has signed (target of MSD calculations):		40,249	(+5.27%)

Best connected keys (shortest distance to)

Please read about the mean shortest distance (MSD) calculated here in the analysis explanation. Here are the top 50 keys. Look for your own key in this month's raw analysis (see above). Note that the only keys analyzed were those reachable from the strong set. I've included some of my own comments on people I recognize. I'm sorry if you're listed here without a comment. If you email me a quick phrase to describe what you do that would be of interest to readers, I'll put it in.

The average MSD is 6.6224, in the set of 10,153. The median value is 6.1993.

Go to this keyserver's web interface to look up these keys.

Rank	Hex ID (last 32b)	Key Name (Identifier)	Comments	MSD
1	09590CFD	Peter N. Wan	GA Tech college of computing	4.0775
2	4F570BA3	Ingmar Camphausen	PGP security maverick	4.0855
3	466B4289	Theodore Ts'o [SIGNATURE]	ext2fstools, Kerberos, LSB, other	4.1057
4	F95C2F6D	Christoph Martin	Debian maintainer & uni-mainz keyserver admin	4.1537
5	8B4608A1	Peter N. Wan PNW2048	GA Tech college of computing	4.2050
6	F081195D	Matthias Bauer		4.2149
7	09AC0A6A	L. Sassaman	cypherpunks	4.2308
8	0A2F87E5	Niels Provos (#2)	OpenBSD, OpenSSH, IPSEC	4.2601
9	C2009841	Niels Provos	OpenBSD, OpenSSH, IPSEC	4.2674
10	0679ED91	teun.nijssen@kub.nl	Manages SURFnet servers, scanned PGP source code	4.2884
11	C7A966DD	Philip R. Zimmermann	Inventor of PGP	4.3007
12	0DBF906D	Jeffrey I. Schiller	MIT Security/Network Manager	4.3037
13	5B0358A2	Werner Koch	Author of GNU Privacy Guard (GPG)	4.3064
14	1CF27FD5	Marc Horowitz	Author of pks PGP keyserver software	4.3186
15	8531327F	Martin Spill		4.3376
16	08C95A15	SURFnet-Master-Certification-Key		4.3458
17	DD934139	Patrick Feisthammel	hosts Swiss PGP keyserver, www.ch.pgp.net	4.3498
18	66A74B31	Teun Nijssen	Manages SURFnet servers, scanned PGP source code	4.3632
19	DC4ED62D	Ingmar Camphausen, DFN-PCA	PGP security maverick	4.3692
20	52D1CAB1	Nathalie Weiler	Security researcher at ETH Zurich	4.3728
21	80B07A4F	Theodore Ts'o	ext2fstools, Kerberos, LSB, other	4.3859
22	FAEBD5FC	Philip R. Zimmermann	Inventor of PGP	4.4020
23	13D9873D	Mirko Dziadzka		4.4043
24	DA0EDC81	Phil Karn		4.4076
25	A094DA25	Patrick Feisthammel CERTIFICATION ONLY, Key A		4.4171
26	28C029AF	Dave Del Torto		4.4301
27	46F3212D	LaMont Jones	Debian developer, postfix junkie	4.4304
28	2B48F6F5	Ian Goldberg	ISAAC, Crypto guru	4.4412
29	00292B81	Nathalie Weiler	Security researcher at ETH Zurich	4.4432
30	E39AF3E9	Chelo Malagon		4.4462
31	2DE30EC1	CERT Coordination Center	CERT	4.4498
32	09D3E64D	Greg Rose	USENIX, PGPMoose	4.4550
33	75BE8097	Florian Lohoff		4.4613
34	FE0B386D	Pasztor, Miklos		4.4633
35	603F2D01	Stefan Kelm		4.4648
36	C158CCED	Florian Lohoff		4.4685
37	7362BE39	Carl Ellison	Chief cryptographer, Intel	4.4697
38	C3FC4C69	Steven M. Bellovin	Security expert, AT&T Labs	4.4740
39	BB1D9F6D	ct magazine CERTIFICATE		4.4751
40	6916C873	Peter N. Wan PNWGTDH4096	GA Tech College of Computing	4.4762
41	8B05342D	Graham King		4.4802
42	1FE961A1	Harald Koenig		4.4814
43	ED9547ED	Wichert Akkerman	Debian Project Leader emeritus	4.4856
44	4413B691	Thomas Lenggenhager		4.4888
45	E12469C1	Ruediger Weis		4.4893
46	39F37F5D	Lutz Donnerhacke	privacy advocate & security expert	4.4946
47	F086CBB5	Marcel Waldvogel		4.4970
48	961F4A35	Tatu Ylonen	Inventor of SSH	4.5010
49	F82CEA91	Simon Cooper		4.5065
50	103D4013	Theodore Ts'o	ext2fstools, Kerberos, LSB, other	4.5085

In the Drew self plug, my own key moved from #272 to #80 with an MSD of 4.6027. This was probably mostly due to a good-sized keysigning at Ottawa Linux Symposium.

For next month

I would still like to see the text key IDs in some of the raw reports, to avoid having to look them up later, but this requires some rewriting in all the sections of code, to hold those text fields in RAM through from import to report. Maybe for September... I certainly wouldn't object to someone contributing this part.

Discussion about this analysis continues on the keyanalyze-discuss mailing list.

If you have any suggestions, please send them my way, especially if you have the algorithms as well. If you're so inclined, please have a look at the code as well.