1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
|
Legend:
SPEC!! - Not specified
SPEC - Spec not finalized
NICK - nick claims
ARMA - arma claims
- Not done
* Top priority
. Partially done
o Done
D Deferred
X Abandoned
For scalability:
- Slightly smarter bandwidth management: use link capacity
intelligently.
- Handle full buffers without totally borking
For 0.0.8:
NICK milestone 1:
o make all ORs serve the directory too.
o "AuthoritativeDir 1" for dirservers
o non-authorative servers with dirport publish opt dircacheport
o make clients read that and use it.
o make clients able to read a normal dirport from non-trusted OR too
o make ORs parse-and-keep-and-serve the directory they pull down
o authoritativedirservers should pull down directories from
other authdirservers, to merge descriptors.
. allow dirservers to serve running-router list separately.
o "get /running-routers" will fetch just this.
o actually make the clients use this sometimes.
o distinguish directory-is-dirty from runninglist-is-dirty
- ORs keep this too, and serve it
o Design: do we need running and non-running lists?
o tor remembers descriptor-lists across reboots.
. Packages define datadir as /var/lib/tor/. If no datadir is defined,
then choose, make, and secure ~/.tor as datadir.
o Adjust tor
o Change torrc.sample
D Change packages (not till 0.0.8 packages!)
o Look in ~/.torrc if no */etc/torrc is found.
NICK - Possible to get autoconf to easily install things into ~/.tor?
o Contact info, pgp fingerprint, comments in router desc.
o Add a ContactInfo line to torrc, which gets published in
descriptor (as opt)
o write tor version at the top of each log file
milestone 2:
. refer to things by key:
o extend cells need ip:port:identitykeyhash.
o Lookup routers and connections by key digest; accept hex
key digest in place of nicknames.
o Audit all uses of lookup-by-hostname and lookup-by-addr-port
to search by digest when appropriate.
o make sure to use addr/port in cpuworker tasks, because
OPs don't have keys.
NICK - and fix the function comments in rephist
o Rep-hist functions need to store info by keyid
- also use this in intro points and rendezvous points, and
hidserv descs. [XXXX This isn't enough.]
- figure out what to do about ip:port:differentkey
o ORs connect on demand. attach circuits to new connections, keep
create cells around somewhere, send destroy if fail.
o nickname defaults to first piece of hostname
o running-routers list refers to nickname if verified, else
hash-base64'ed.
- Mark routers as verified or unverified based on whether
running-routers list includes nickname or id hash.
o put OR uptime in descriptor
o name the secret-key directory something to discourage people
from mailing their identity key to tor-ops
milestone 3:
- users can set their bandwidth, or we auto-detect it:
- advertised bandwidth defaults to 10KB
- advertised bandwidth is the min of max seen in each direction
in the past N seconds.
NICK - calculate this
o not counting "local" connections
- round detected bandwidth up to nearest 10KB
- client software not upload descriptor until:
- you've been running for an hour
- it's sufficiently satisfied with its bandwidth
- it decides it is reachable
- start counting again if your IP ever changes.
- never regenerate identity keys, for now.
- you can set a bit for not-being-an-OR.
NICK - Reputation info needs to give better weight to recent events than
very old ones.
- Have clients and dirservers preserve reputation info over
reboots.
- clients choose nodes proportional to advertised bandwidth
- authdirserver includes descriptor and lists as running iff:
- he can connect to you
- he has successfully extended to you
- he has sufficient mean-time-between-failures
- add new "Middleman 1" config variable?
o if torrc not found, exitpolicy reject *:*
NICK - change if(options.ORPort) to what we really mean.
NICK - same with socksport.
ongoing:
. rename/rearrange functions for what file they're in
- generalize our transport: add transport.c in preparation for
http, airhook, etc transport.
For September:
NICK . Windows port
o works as client
- deal with pollhup / reached_eof on all platforms
. robust as a client
. works as server
- can be configured
- robust as a server
. Usable as NT service
- docs for building in win
- installer
- Docs
- FAQ
o overview of tor. how does it work, what's it do, pros and
cons of using it, why should I use it, etc.
- a howto tutorial with examples
o tutorial: how to set up your own tor network
- (need to not hardcode dirservers file in config.c)
. correct, update, polish spec
- document the exposed function api?
- document what we mean by socks.
NICK . packages
. rpm
- find a long-term rpm maintainer
- code
- better warn/info messages
o let tor do resolves.
o extend socks4 to do resolves?
o make script to ask tor for resolves
- tsocks
- gather patches, submit to maintainer
- intercept gethostbyname and others, do resolve via tor
- redesign and thorough code revamp, with particular eye toward:
- support half-open tcp connections
- conn key rotation
- other transports -- http, airhook
- modular introduction mechanism
- allow non-clique topology
Other details and small and hard things:
- tor should be able to have a pool of outgoing IP addresses
that it is able to rotate through. (maybe)
- tie into squid
- buffer size pool, to let a few buffers grow huge or many buffers
grow a bit
- hidserv offerers shouldn't need to define a SocksPort
- when the client fails to pick an intro point for a hidserv,
it should refetch the hidserv desc.
. should maybe make clients exit(1) when bad things happen?
e.g. clock skew.
- should retry exitpolicy end streams even if the end cell didn't
resolve the address for you
- Add '[...truncated]' or similar to truncated log entries (like the directory
in connection_dir_process_inbuf()).
. Make logs handle it better when writing to them fails.
o Dirserver shouldn't put you in running-routers list if you haven't
uploaded a descriptor recently
. Refactor: add own routerinfo to routerlist. Right now, only
router_get_by_nickname knows about 'this router', as a hack to
get circuit_launch_new to do the right thing.
. Scrubbing proxies
- Find an smtp proxy?
. Get socks4a support into Mozilla
- Extend by hostname, not by IP.
- Need a relay teardown cell, separate from one-way ends.
- Make it harder to circumvent bandwidth caps: look at number of bytes
sent across sockets, not number sent inside TLS stream.
- fix router_get_by_* functions so they can get ourselves too,
and audit everything to make sure rend and intro points are
just as likely to be us as not.
***************************Future tasks:****************************
Rendezvous and hidden services:
make it fast:
- preemptively build and start rendezvous circs.
- preemptively build n-1 hops of intro circs?
- cannibalize general circs?
make it reliable:
- standby/hotswap/redundant services.
- store stuff to disk? dirservers forget service descriptors when
they restart; nodes offering hidden services forget their chosen
intro points when they restart.
make it robust:
- auth mechanisms to let midpoint and bob selectively choose
connection requests.
make it scalable:
- right now the hidserv store/lookup system is run by the dirservers;
this won't scale.
Tor scalability:
Relax clique assumptions.
Redesign how directories are handled.
- Separate running-routers lookup from descriptor list lookup.
- Resolve directory agreement somehow.
- Cache directory on all servers.
Find and remove bottlenecks
- Address linear searches on e.g. circuit and connection lists.
Reputation/memory system, so dirservers can measure people,
and so other people can verify their measurements.
- Need to measure via relay, so it's not distinguishable.
Bandwidth-aware path selection. So people with T3's are picked
more often than people with DSL.
Reliability-aware node selection. So people who are stable are
preferred for long-term circuits such as intro and rend circs,
and general circs for irc, aim, ssh, etc.
Let dissidents get to Tor servers via Tor users. ("Backbone model")
Anonymity improvements:
Is abandoning the circuit the only option when an extend fails, or
can we do something without impacting anonymity too much?
Is exiting from the middle of the circuit always a bad idea?
Helper nodes. Decide how to use them to improve safety.
DNS resolution: need to make tor support resolve requests. Need to write
a script and an interface (including an extension to the socks
protocol) so we can ask it to do resolve requests. Need to patch
tsocks to intercept gethostbyname, else we'll continue leaking it.
Improve path selection algorithms based on routing-zones paper. Be sure
to start and end circuits in different ASs. Ideally, consider AS of
source and destination -- maybe even enter and exit via nearby AS.
Intermediate model, with some delays and mixing.
Add defensive dropping regime?
Make it more correct:
Handle half-open connections: right now we don't support all TCP
streams, at least according to the protocol. But we handle all that
we've seen in the wild.
Support IPv6.
Efficiency/speed/robustness:
Congestion control. Is our current design sufficient once we have heavy
use? Need to measure and tweak, or maybe overhaul.
Allow small cells and large cells on the same network?
Cell buffering and resending. This will allow us to handle broken
circuits as long as the endpoints don't break, plus will allow
connection (tls session key) rotation.
Implement Morphmix, so we can compare its behavior, complexity, etc.
Use cpuworker for more heavy lifting.
- Signing (and verifying) hidserv descriptors
- Signing (and verifying) intro/rend requests
- Signing (and verifying) router descriptors
- Signing (and verifying) directories
- Doing TLS handshake (this is very hard to separate out, though)
Buffer size pool: allocate a maximum size for all buffers, not
a maximum size for each buffer. So we don't have to give up as
quickly (and kill the thickpipe!) when there's congestion.
Exit node caching: tie into squid or other caching web proxy.
Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
link crypto, unless we can bully openssl into it.
P2P Tor:
Do all the scalability stuff above, first.
Incentives to relay. Not so hard.
Incentives to allow exit. Possibly quite hard.
Sybil defenses without having a human bottleneck.
How to gather random sample of nodes.
How to handle nodelist recommendations.
Consider incremental switches: a p2p tor with only 50 users has
different anonymity properties than one with 10k users, and should
be treated differently.
|