tor/doc/design-paper/roadmap-future.tex

\documentclass{article}

\usepackage{url}
\usepackage{fullpage}

\newenvironment{tightlist}{\begin{list}{$\bullet$}{
  \setlength{\itemsep}{0mm}
    \setlength{\parsep}{0mm}
    %  \setlength{\labelsep}{0mm}
    %  \setlength{\labelwidth}{0mm}
    %  \setlength{\topsep}{0mm}
    }}{\end{list}}
\newcommand{\tmp}[1]{{\bf #1} [......] \\}
\newcommand{\plan}[1]{ {\bf (#1)}}

\begin{document}

\title{Tor Development Roadmap: Wishlist for 2008 and beyond}
\author{Roger Dingledine \and Nick Mathewson}
\date{}

\maketitle
\pagestyle{plain}

\section{Introduction}

Tor (the software) and Tor (the overall software/network/support/document
suite) are now experiencing all the crises of success.  Over the next
years, we're probably going to grow even more in terms of users, developers,
and funding than before. This document attempts to lay out all the
well-understood next steps that Tor needs to take. We should periodically
reorganize it to reflect current and intended priorities.

\section{Everybody can be a relay}

We've made a lot of progress towards letting an ordinary Tor client also
serve as a Tor relay. But these issues remain.

\subsection{UPNP}

We should teach Vidalia how to speak UPNP to automatically open and
forward ports on common (e.g. Linksys) routers. There are some promising
Qt-based UPNP libs out there, and in any case there are others (e.g. in
Perl) that we can base it on.

\subsection{``ORPort auto'' to look for a reachable port}

Vidalia defaults to port 443 on Windows and port 8080 elsewhere. But if
that port is already in use, or the ISP filters incoming connections
on that port (some cablemodem providers filter 443 inbound), the user
needs to learn how to notice this, and then pick a new one and type it
into Vidalia.

We should add a new option ``auto'' that cycles through a set of preferred
ports, testing bindability and reachability for each of them, and only
complains to the user once it's given up on the common choices.

\subsection{Incentives design}

Roger has been working with researchers at Rice University to simulate
and analyze a new design where the directory authorities assign gold
stars to well-behaving relays, and then all the relays give priority
to traffic from gold-starred relays. The great feature of the design is
that not only does it provide the (explicit) incentive to run a relay,
but it also aims to grow the overall capacity of the network, so even
non-relays will benefit.

It needs more analysis, and perhaps more design work, before we try
deploying it.

\subsection{Windows libevent}

Tor relays still don't work well or reliably on Windows XP or Windows
Vista, because we don't use the Windows-native ``overlapped IO''
approach. Christian King made a good start at teaching libevent about
overlapped IO during Google Summer of Code 2007, and next steps are
to a) finish that, b) teach Tor to do openssl calls on buffers rather
than directly to the network, and c) teach Tor to use the new libevent
buffers approach.

\subsection{Network scaling}

If we attract many more relays, we will need to handle the growing pains
in terms of getting all the directory information to all the users.

The first piece of this issue is a practical question: since the
directory size scales linearly with more relays, at some point it
will no longer be practical for every client to learn about every
relay. We can try to reduce the amount of information each client needs
to fetch (e.g. based on fetching less information preemptively as in
Section~\ref{subsec:fewer-descriptor-fetches} below), but eventually
clients will need to learn about only a subset of the network, and we
will need to design good ways to divide up the network information.

The second piece is an anonymity question that arises from this
partitioning: if Tor's security comes from having all the clients
behaving in similar ways, yet we are now giving different clients
different directory information, how can we minimize the new anonymity
attacks we introduce?

\subsection{Using fewer sockets}

Since in the current network every Tor relay can reach every other Tor
relay, and we have many times more users than relays, pretty much every
possible link in the network is in use. That is, the current network
is a clique in practice.

And since each of these connections requires a TCP socket, it's going
to be hard for the network to grow much larger: many systems come with
a default of 1024 file descriptors allowed per process, and raising
that ulimit is hard for end users. Worse, many low-end gateway/firewall
routers can't handle this many connections in their routing table.

One approach is a restricted-route topology~\cite{danezis:pet2003}:
predefine which relays can reach which other relays, and communicate
these restrictions to the relays and the clients. We need to compute
which links are acceptable in a way that's decentralized yet scalable,
and in a way that achieves a small-worlds property; and we
need an efficient (compact) way to characterize the topology information
so all the users could keep up to date.

Another approach would be to switch to UDP-based transport between
relays, so we don't need to keep the TCP sockets open at all. Needs more
investigation too.

\subsection{Auto bandwidth detection and rate limiting, especially for
      asymmetric connections.}


\subsection{Better algorithms for giving priority to local traffic}

Proposal 111 made a lot of progress at separating local traffic from
relayed traffic, so Tor users can rate limit the relayed traffic at a
stricter level. But since we want to pass both traffic classes over the
same TCP connection, we can't keep them entirely separate. The current
compromise is that we treat all bytes to/from a given connectin as
local traffic if any of the bytes within the past N seconds were local
bytes.  But a) we could use some more intelligent heuristics, and b)
this leaks information to an active attacker about when local traffic
was sent/received.

\subsection{Tolerate absurdly wrong clocks, even for relays}

Many of our users are on Windows, running with a clock several days or
even several years off from reality. Some of them are even intentionally
in this state so they can run software that will only run in the past.

Before Tor 0.1.1.x, Tor clients would still function if their clock was
wildly off --- they simply got a copy of the directory and believed it.
Starting in Tor 0.1.1.x (and even moreso in Tor 0.2.0.x), the clients
only use networkstatus documents that they believe to be recent, so
clients with extremely wrong clocks no longer work. (This bug has been
an unending source of vague and confusing bug reports.)

The first step is for clients to recognize when all the directory material
they're fetching has roughly the same offset from their current time,
and then automatically correct for it.

Once that's working well, clients who opt to become bridge relays should
be able to use the same approach to serve accurate directory information
to their bridge users.

\subsection{Risks from being a relay}

Three different research
papers~\cite{back01,clog-the-queue,attack-tor-oak05} describe ways to
identify the nodes in a circuit by running traffic through candidate nodes
and looking for dips in the traffic while the circuit is active. These
clogging attacks are not that scary in the Tor context so long as relays
are never clients too. But if we're trying to encourage more clients to
turn on relay functionality too (whether as bridge relays or as normal
relays), then we need to understand this threat better and learn how to
mitigate it.

One promising research direction is to investigate the RelayBandwidthRate
feature that lets Tor rate limit relayed traffic differently from local
traffic. Since the attacker's ``clogging'' traffic is not in the same
bandwidth class as the traffic initiated by the user, it may be harder
to detect interference. Or it may not be.

\subsection{First a bridge, then a public relay?}

Once enough of the items in this section are done, I want all clients
to start out automatically detecting their reachability and opting
to be bridge relays.

Then if they realize they have enough consistency and bandwidth, they
should automatically upgrade to being non-exit relays.

What metrics should we use for deciding when we're fast enough
and stable enough to switch? Given that the list of bridge relays needs
to be kept secret, it doesn't make much sense to switch back.

\section{Tor on low resources / slow links}
\subsection{Reducing directory fetches further}
\label{subsec:fewer-descriptor-fetches}
\subsection{AvoidDiskWrites}
\subsection{Using less ram}
\subsection{Better DoS resistance for tor servers / authorities}
\section{Blocking resistance}
\subsection{Better bridge-address-distribution strategies}
\subsection{Get more volunteers running bridges}
\subsection{Handle multiple bridge authorities}
\subsection{Anonymity for bridge users: second layer of entry guards, etc?}
\subsection{More TLS normalization}
\subsection{Harder to block Tor software distribution}
\subsection{Integration with Psiphon}
\section{Packaging}
\subsection{Switch Privoxy out for Polipo}
      - Make Vidalia able to launch more programs itself
\subsection{Continue Torbutton improvements}
      especially better docs
\subsection{Vidalia and stability (especially wrt ongoing Windows problems)}
      learn how to get useful crash reports (tracebacks) from Windows users
\subsection{Polipo support on Windows}
\subsection{Auto update for Tor, Vidalia, others}
\subsection{Tor browser bundle for USB and standalone use}
\subsection{LiveCD solution}
\subsection{VM-based solution}
\subsection{Tor-on-enclave-firewall configuration}
\subsection{General tutorials on what common applications are Tor-friendly}
\subsection{Controller libraries (torctl) plus documentation}
\subsection{Localization and translation (Vidalia, Torbutton, web pages)}
\section{Interacting better with Internet sites}
\subsection{Make tordnsel (tor exitlist) better and more well-known}
\subsection{Nymble}
\subsection{Work with Wikipedia, Slashdot, Google(, IRC networks)}
\subsection{IPv6 support for exit destinations}
\section{Network health}
\subsection{torflow / soat to detect bad relays}
\subsection{make authorities more automated}
\subsection{torstatus pages and better trend tracking}
\subsection{better metrics for assessing network health / growth}
      - geoip usage-by-country reporting and aggregation
        (Once that's working, switch to Directory guards)
\section{Performance research}
\subsection{Load balance better}
\subsection{Improve our congestion control algorithms}
\subsection{Two-hops vs Three-hops}
\subsection{Transport IP packets end-to-end}
\section{Outreach and user education}
\subsection{"Who uses Tor" use cases}
\subsection{Law enforcement contacts}
      - "Was this IP address a Tor relay recently?" database
\subsection{Commercial/enterprise outreach. Help them use Tor well and
      not fear it.}
\subsection{NGO outreach and training.}
      - "How to be a safe blogger"
\subsection{More activist coordinators, more people to answer user questions}
\subsection{More people to hold hands of server operators}
\subsection{Teaching the media about Tor}
\subsection{The-dangers-of-plaintext awareness}
\subsection{check.torproject.org and other "privacy checkers"}
\subsection{Stronger legal FAQ for US}
\subsection{Legal FAQs for other countries}
\section{Anonymity research}
\subsection{estimate relay bandwidth more securely}
\subsection{website fingerprinting attacks}
\subsection{safer e2e defenses}
\subsection{Using Tor when you really need anonymity. Can you compose it
      with other steps, like more trusted guards or separate proxies?}
\subsection{Topology-aware routing; routing-zones, steven's pet2007 paper.}
\subsection{Exactly what do guard nodes provide?}

Entry guards seem to defend against all sorts of attacks. Can we work
through all the benefits they provide? Papers like Nikita's CCS 2007
paper make me think their value is not well-understood by the research
community.

\section{Organizational growth and stability}
\subsection{A contingency plan if Roger gets hit by a bus}
      - Get a new executive director
\subsection{More diversity of funding}
      - Don't rely on any one funder as much
      - Don't rely on any sector or funder category as much
\subsection{More Tor-funded people who are skilled at peripheral apps like
      Vidalia, Torbutton, Polipo, etc}
\subsection{More coordinated media handling and strategy}
\subsection{Clearer and more predictable trademark behavior}
\subsection{More outside funding for internships, etc e.g. GSoC.}
\section{Hidden services}
\subsection{Scaling: how to handle many hidden services}
\subsection{Performance: how to rendezvous with them quickly}
\subsection{Authentication/authorization: how to tolerate DoS / load}
\section{Tor as a general overlay network}
\subsection{Choose paths / exit by country}
\subsection{Easier to run your own private servers and have Tor use them
      anywhere in the path}
\subsection{Easier to run an independent Tor network}
\section{Code security/correctness}
\subsection{veracode}
\subsection{code audit}
\subsection{more fuzzing tools}
\subsection{build farm, better testing harness}
\subsection{Long-overdue code refactoring and cleanup}
\section{Protocol security}
\subsection{safer circuit handshake}
\subsection{protocol versioning for future compatibility}
\subsection{cell sizes}
\subsection{adapt to new key sizes, etc}

\bibliographystyle{plain} \bibliography{tor-design}

\end{document}




\section{Code and design infrastructure}

\subsection{Protocol revision}
To maintain backward compatibility, we've postponed major protocol
changes and redesigns for a long time.  Because of this, there are a number
of sensible revisions we've been putting off until we could deploy several of
them at once.  To do each of these, we first need to discuss design
alternatives with other cryptographers and outside collaborators to
make sure that our choices are secure.

First of all, our protocol needs better {\bf versioning support} so that we
can make backward-incompatible changes to our core protocol.  There are
difficult anonymity issues here, since many naive designs would make it easy
to tell clients apart (and then track them) based on their supported versions.

With protocol versioning support would come the ability to {\bf future-proof
  our ciphersuites}.  For example, not only our OR protocol, but also our
directory protocol, is pretty firmly tied to the SHA-1 hash function, which
though not yet known to be insecure for our purposes, has begun to show
its age.  We should
remove assumptions throughout our design based on the assumption that public
keys, secret keys, or digests will remain any particular size indefinitely.

Our OR {\bf authentication protocol}, though provably
secure\cite{tap:pet2006}, relies more on particular aspects of RSA and our
implementation thereof than we had initially believed.  To future-proof
against changes, we should replace it with a less delicate approach.

\plan{For all the above: 2 person-months to specify, spread over several
  months with time for interaction with external participants.  One
  person-month to implement.  Start specifying in early 2007.}

We might design a {\bf stream migration} feature so that streams tunneled
over Tor could be more resilient to dropped connections and changed IPs.
\plan{Not in 2007.}

A new protocol could support {\bf multiple cell sizes}.  Right now, all data
passes through the Tor network divided into 512-byte cells.  This is
efficient for high-bandwidth protocols, but inefficient for protocols
like SSH or AIM that send information in small chunks.  Of course, we need to
investigate the extent to which multiple sizes could make it easier for an
adversary to fingerprint a traffic pattern. \plan{Not in 2007.}

As a part of our design, we should investigate possible {\bf cipher modes}
other than counter mode.  For example, a mode with built-in integrity
checking, error propagation, and random access could simplify our protocol
significantly.  Sadly, many of these are patented and unavailable for us.
\plan{Not in 2007.}

\subsection{Scalability}

\subsubsection{Improved directory efficiency}

We should {\bf have routers upload their descriptors even less often}, so
that clients do not need to download replacements every 18 hours whether any
information has changed or not.  (As of Tor 0.1.2.3-alpha, clients tolerate
routers that don't upload often, but routers still upload at least every 18
hours to support older clients.) \plan{Must do, but not until 0.1.1.x is
deprecated in mid 2007. 1 week.}

\subsubsection{Non-clique topology}
Our current network design achieves a certain amount of its anonymity by
making clients act like each other through the simple expedient of making
sure that all clients know all servers, and that any server can talk to any
other server.  But as the number of servers increases to serve an
ever-greater number of clients, these assumptions become impractical.

At worst, if these scalability issues become troubling before a solution is
found, we can design and build a solution to {\bf split the network into
multiple slices} until a better solution comes along.  This is not ideal,
since rather than looking like all other users from a point of view of path
selection, users would ``only'' look like 200,000--300,000 other
users.\plan{Not unless needed.}

We are in the process of designing {\bf improved schemes for network
  scalability}.  Some approaches focus on limiting what an adversary can know
about what a user knows; others focus on reducing the extent to which an
adversary can exploit this knowledge.  These are currently in their infancy,
and will probably not be needed in 2007, but they must be designed in 2007 if
they are to be deployed in 2008.\plan{Design in 2007; unknown difficulty.
  Write a paper.}

\subsubsection{Relay incentives}
To support more users on the network, we need to get more servers.  So far,
we've relied on volunteerism to attract server operators, and so far it's
served us well.  But in the long run, we need to {\bf design incentives for
  users to run servers} and relay traffic for others.  Most obviously, we
could try to build the network so that servers offered improved service for
other servers, but we would need to do so without weakening anonymity and
making it obvious which connections originate from users running servers.  We
have some preliminary designs~\cite{incentives-txt,tor-challenges},
but need to perform
some more research to make sure they would be safe and effective.\plan{Write
  a draft paper; 2 person-months.}
(XXX we did that)

\subsection{Portability}
Our {\bf Windows implementation}, though much improved, continues to lag
behind Unix and Mac OS X, especially when running as a server.  We hope to
merge promising patches from Christian King to address this point, and bring
Windows performance on par with other platforms.\plan{Do in 2007; 1.5 months
  to integrate not counting Mike's work.}

We should have {\bf better support for portable devices}, including modes of
operation that require less RAM, and that write to disk less frequently (to
avoid wearing out flash RAM).\plan{Optional; 2 weeks.}

\subsection{Performance: resource usage}
We've been working on {\bf using less RAM}, especially on servers.  This has
paid off a lot for directory caches in the 0.1.2, which in some cases are
using 90\% less memory than they used to require.  But we can do better,
especially in the area around our buffer management algorithms, by using an
approach more like the BSD and Linux kernels use instead of our current ring
buffer approach.  (For OR connections, we can just use queues of cell-sized
chunks produced with a specialized allocator.)  This could potentially save
around 25 to 50\% of the memory currently allocated for network buffers, and
make Tor a more attractive proposition for restricted-memory environments
like old computers, mobile devices, and the like.\plan{Do in 2007; 2-3 weeks
  plus one week measurement.} (XXX We did this, but we need to do something
more/else.)

\subsection{Performance: network usage}
We know too little about how well our current path
selection algorithms actually spread traffic around the network in practice.
We should {\bf research the efficacy of our traffic allocation} and either
assure ourselves that it is close enough to optimal as to need no improvement
(unlikely) or {\bf identify ways to improve network usage}, and get more
users' traffic delivered faster.  Performing this research will require
careful thought about anonymity implications.

We should also {\bf examine the efficacy of our congestion control
  algorithm}, and see whether we can improve client performance in the
presence of a congested network through dynamic `sendme' window sizes or
other means.  This will have anonymity implications too if we aren't careful.

\plan{For both of the above: research, design and write
  a measurement tool in 2007: 1 month.  See if we can interest a graduate
  student.}

We should work on making Tor's cell-based protocol  perform better on
networks with low bandwidth
and high packet loss.\plan{Do in 2007 if we're funded to do it; 4-6 weeks.}

\subsection{Performance scenario: one Tor client, many users}
We should {\bf improve Tor's performance when a single Tor handles many
  clients}.  Many organizations want to manage a single Tor client on their
firewall for many users, rather than having each user install a separate
Tor client.  We haven't optimized for this scenario, and it is likely that
there are some code paths in the current implementation that become
inefficient when a single Tor is servicing hundreds or thousands of client
connections.  (Additionally, it is likely that such clients have interesting
anonymity requirements the we should investigate.)  We should profile Tor
under appropriate loads, identify bottlenecks, and fix them.\plan{Do in 2007
  if we're funded to do it; 4-8 weeks.}

\subsection{Tor servers on asymmetric bandwidth}

Tor should work better on servers that have asymmetric connections like cable
or DSL.  Because Tor has separate TCP connections between each
hop, if the incoming bytes are arriving just fine and the outgoing bytes are
all getting dropped on the floor, the TCP push-back mechanisms don't really
transmit this information back to the incoming streams.\plan{Do in 2007 since
  related to bandwidth limiting.  3-4 weeks.}

\subsection{Running Tor as both client and server}

Many performance tradeoffs and balances that might need more attention.
We first need to track and fix whatever bottlenecks emerge; but we also
need to invent good algorithms for prioritizing the client's traffic
without starving the server's traffic too much.\plan{No idea; try
profiling and improving things in 2007.}

\subsection{Protocol redesign for UDP}
Tor has relayed only TCP traffic since its first versions, and has used
TLS-over-TCP to do so.  This approach has proved reliable and flexible, but
in the long term we will need to allow UDP traffic on the network, and switch
some or all of the network to using a UDP transport.  {\bf Supporting UDP
  traffic} will make Tor more suitable for protocols that require UDP, such
as many VOIP protocols.  {\bf Using a UDP transport} could greatly reduce
resource limitations on servers, and make the network far less interruptible
by lossy connections.  Either of these protocol changes would require a great
deal of design work, however.  We hope to be able to enlist the aid of a few
talented graduate students to assist with the initial design and
specification, but the actual implementation will require significant testing
of different reliable transport approaches.\plan{Maybe do a design in 2007 if
we find an interested academic.  Ian or Ben L might be good partners here.}

\section{Blocking resistance}

\subsection{Design for blocking resistance}
We have written a design document explaining our general approach to blocking
resistance.  We should workshop it with other experts in the field to get
their ideas about how we can improve Tor's efficacy as an anti-censorship
tool.

\subsection{Implementation: client-side and bridges-side}

Bridges will want to be able to {\bf listen on multiple addresses and ports}
if they can, to give the adversary more ports to block.

\subsection{Research: anonymity implications from becoming a bridge}

see arma's bridge proposal; e.g. should bridge users use a second layer of
entry guards?

\subsection{Implementation: bridge authority}

we run some
directory authorities with a slightly modified protocol that doesn't leak
the entire list of bridges. Thus users can learn up-to-date information
for bridges they already know about, but they can't learn about arbitrary
new bridges.

we need a design for distributing the bridge authority over more than one
server

\subsection{Normalizing the Tor protocol on the wire}
Additionally, we should {\bf resist content-based filters}.  Though an
adversary can't see what users are saying, some aspects of our protocol are
easy to fingerprint {\em as} Tor.  We should correct this where possible.

Look like Firefox; or look like nothing?
Future research: investigate timing similarities with other protocols.

\subsection{Research: scanning-resistance}

\subsection{Research/Design/Impl: how users discover bridges}
Our design anticipates an arms race between discovery methods and censors.
We need to begin the infrastructure on our side quickly, preferably in a
flexible language like Python, so we can adapt quickly to censorship.

phase one: personal bridges
phase two: families of personal bridges
phase three: more structured social network
phase four: bag of tricks
Research: phase five...

Integration with Psiphon, etc?

\subsection{Document best practices for users}
Document best practices for various activities common among
blocked users (e.g. WordPress use).

\subsection{Research: how to know if a bridge has been blocked?}

\subsection{GeoIP maintenance, and "private" user statistics}
How to know if the whole idea is working?

\subsection{Research: hiding whether the user is reading or publishing?}

\subsection{Research: how many bridges do you need to know to maintain
reachability?}

\subsection{Resisting censorship of the Tor website, docs, and mirrors}

We should take some effort to consider {\bf initial distribution of Tor and
  related information} in countries where the Tor website and mirrors are
censored.  (Right now, most countries that block access to Tor block only the
main website and leave mirrors and the network itself untouched.)  Falling
back on word-of-mouth is always a good last resort, but we should also take
steps to make sure it's relatively easy for users to get ahold of a copy.

\section{Security}

\subsection{Security research projects}

We should investigate approaches with some promise to help Tor resist
end-to-end traffic correlation attacks.  It's an open research question
whether (and to what extent) {\bf mixed-latency} networks, {\bf low-volume
  long-distance padding}, or other approaches can resist these attacks, which
are currently some of the most effective against careful Tor users.  We
should research these questions and perform simulations to identify
opportunities for strengthening our design without dropping performance to
unacceptable levels. %Cite something
\plan{Start doing this in 2007; write a paper.  8-16 weeks.}

We've got some preliminary results suggesting that {\bf a topology-aware
  routing algorithm}~\cite{feamster:wpes2004} could reduce Tor users'
vulnerability against local or ISP-level adversaries, by ensuring that they
are never in a position to watch both ends of a connection.  We need to
examine the effects of this approach in more detail and consider side-effects
on anonymity against other kinds of adversaries.  If the approach still looks
promising, we should investigate ways for clients to implement it (or an
approximation of it) without having to download routing tables for the whole
Internet. \plan{Not in 2007 unless a graduate student wants to do it.}

%\tmp{defenses against end-to-end correlation}  We don't expect any to work
%right now, but it would be useful to learn that one did.  Alternatively,
%proving that one didn't would free up researchers in the field to go work on
%other things.
%
% See above; I think I got this.

We should research the efficacy of {\bf website fingerprinting} attacks,
wherein an adversary tries to match the distinctive traffic and timing
pattern of the resources constituting a given website to the traffic pattern
of a user's client.  These attacks work great in simulations, but in
practice we hear they don't work nearly as well.  We should get some actual
numbers to investigate the issue, and figure out what's going on.  If we
resist these attacks, or can improve our design to resist them, we should.
% add cites
\plan{Possibly part of end-to-end correlation paper.  Otherwise, not in 2007
  unless a graduate student is interested.}

\subsection{Implementation security}

We should also {\bf mark RAM that holds key material as non-swappable} so
that there is no risk of recovering key material from a hard disk
compromise.  This would require submitting patches upstream to OpenSSL, where
support for marking memory as sensitive is currently in a very preliminary
state.\plan{Nice to do, but not in immediate Tor scope.}

There are numerous tools for identifying trouble spots in code (such as
Coverity or even VS2005's code analysis tool) and we should convince somebody
to run some of them against the Tor codebase.  Ideally, we could figure out a
way to get our code checked periodically rather than just once.\plan{Almost
  no time once we talk somebody into it.}

We should try {\bf protocol fuzzing} to identify errors in our
implementation.\plan{Not in 2007 unless we find a grad student or
  undergraduate who wants to try.}

Our guard nodes help prevent an attacker from being able to become a chosen
client's entry point by having each client choose a few favorite entry points
as ``guards'' and stick to them.   We should implement a {\bf directory
  guards} feature to keep adversaries from enumerating Tor users by acting as
a directory cache.\plan{Do in 2007; 2 weeks.}

\subsection{Detect corrupt exits and other servers}
With the success of our network, we've attracted servers in many locations,
operated by many kinds of people.  Unfortunately, some of these locations
have compromised or defective networks, and some of these people are
untrustworthy or incompetent.  Our current design relies on authority
administrators to identify bad nodes and mark them as nonfunctioning.  We
should {\bf automate the process of identifying malfunctioning nodes} as
follows:

We should create a generic {\bf feedback mechanism for add-on tools} like
Mike Perry's ``Snakes on a Tor'' to report failing nodes to authorities.
\plan{Do in 2006; 1-2 weeks.}

We should write tools to {\bf detect more kinds of innocent node failure},
such as nodes whose network providers intercept SSL, nodes whose network
providers censor popular websites, and so on.  We should also try to detect
{\bf routers that snoop traffic}; we could do this by launching connections
to throwaway accounts, and seeing which accounts get used.\plan{Do in 2007;
  ask Mike Perry if he's interested.  4-6 weeks.}

We should add {\bf an efficient way for authorities to mark a set of servers
  as probably collaborating} though not necessarily otherwise dishonest.
This happens when an administrator starts multiple routers, but doesn't mark
them as belonging to the same family.\plan{Do during v2.1 directory protocol
  redesign; 1-2 weeks to implement.}

To avoid attacks where an adversary claims good performance in order to
attract traffic, we should {\bf have authorities measure node performance}
(including stability and bandwidth) themselves, and not simply believe what
they're told. We also measure stability by tracking MTBF.  Measuring
bandwidth will be tricky, since it's hard to distinguish between a server with
low capacity, and a high-capacity server with most of its capacity in
use. See also Nikita's NDSS 2008 paper.\plan{Do it if we can interest
a grad student.}

{\bf Operating a directory authority should be easier.}  We rely on authority
operators to keep the network running well, but right now their job involves
too much busywork and administrative overhead.  A better interface for them
to use could free their time to work on exception cases rather than on
adding named nodes to the network.\plan{Do in 2007; 4-5 weeks.}

\subsection{Protocol security}

In addition to other protocol changes discussed above,
% And should we move some of them down here? -NM
we should add {\bf hooks for denial-of-service resistance}; we have some
preliminary designs, but we shouldn't postpone them until we really need them.
If somebody tries a DDoS attack against the Tor network, we won't want to
wait for all the servers and clients to upgrade to a new
version.\plan{Research project; do this in 2007 if funded.}

\section{Development infrastructure}

\subsection{Build farm}
We've begun to deploy a cross-platform distributed build farm of hosts
that build and test the Tor source every time it changes in our development
repository.

We need to {\bf get more participants}, so that we can test a larger variety
of platforms.  (Previously, we've only found out when our code had broken on
obscure platforms when somebody got around to building it.)

We need also to {\bf add our dependencies} to the build farm, so that we can
ensure that libraries we need (especially libevent) do not stop working on
any important platform between one release and the next.

\plan{This is ongoing as more buildbots arrive.}

\subsection{Improved testing harness}
Currently, our {\bf unit tests} cover only about 20\% of the code base.  This
is uncomfortably low; we should write more and switch to a more flexible
testing framework.\plan{Ongoing basis, time permitting.}

We should also write flexible {\bf automated single-host deployment tests} so
we can more easily verify that the current codebase works with the
network.\plan{Worthwhile in 2007; would save lots of time.  2-4 weeks.}

We should build automated {\bf stress testing} frameworks so we can see which
realistic loads cause Tor to perform badly, and regularly profile Tor against
these loads.  This would give us {\it in vitro} performance values to
supplement our deployment experience.\plan{Worthwhile in 2007; 2-6 weeks.}

We should improve our memory profiling code.\plan{...}


\subsection{Centralized build system}
We currently rely on a separate packager to maintain the packaging system and
to build Tor on each platform for which we distribute binaries.  Separate
package maintainers is sensible, but separate package builders has meant
long turnaround times between source releases and package releases.  We
should create the necessary infrastructure for us to produce binaries for all
major packages within an hour or so of source release.\plan{We should
  brainstorm this at least in 2007.}

\subsection{Improved metrics}
We need a way to {\bf measure the network's health, capacity, and degree of
  utilization}.  Our current means for doing this are ad hoc and not
completely accurate

We need better ways to {\bf tell which countries are users are coming from,
  and how many there are}.  A good perspective of the network helps us
allocate resources and identify trouble spots, but our current approaches
will work less and less well as we make it harder for adversaries to
enumerate users.  We'll probably want to shift to a smarter, statistical
approach rather than our current ``count and extrapolate'' method.

\plan{All of this in 2007 if funded; 4-8 weeks}

% \tmp{We'd like to know how much of the network is getting used.}
% I think this is covered above -NM

\subsection{Controller library}
We've done lots of design and development on our controller interface, which
allows UI applications and other tools to interact with Tor.  We could
encourage the development of more such tools by releasing a {\bf
  general-purpose controller library}, ideally with API support for several
popular programming languages.\plan{2006 or 2007; 1-2 weeks.}

\section{User experience}

\subsection{Get blocked less, get blocked less broadly}
Right now, some services block connections from the Tor network because
they don't have a better
way to keep vandals from abusing them than blocking IP addresses associated
with vandalism.  Our approach so far has been to educate them about better
solutions that currently exist, but we should also {\bf create better
solutions for limiting vandalism by anonymous users} like credential and
blind-signature based implementations, and encourage their use. Other
promising starting points including writing a patch and explanation for
Wikipedia, and helping Freenode to document, maintain, and expand its
current Tor-friendly position.\plan{Do a writeup here in 2007; 1-2 weeks.}

Those who do block Tor users also block overbroadly, sometimes blacklisting
operators of Tor servers that do not permit exit to their services.  We could
obviate innocent reasons for doing so by designing a {\bf narrowly-targeted Tor
  RBL service} so that those who wanted to overblock Tor could no longer
plead incompetence.\plan{Possibly in 2007 if we decide it's a good idea; 3
  weeks.}

\subsection{All-in-one bundle}
We need a well-tested, well-documented bundle of Tor and supporting
applications configured to use it correctly.  We have an initial
implementation well under way, but it will need additional work in
identifying requisite Firefox extensions, identifying security threats,
improving user experience, and so on.  This will need significantly more work
before it's ready for a general public release.

\subsection{LiveCD Tor}
We need a nice bootable livecd containing a minimal OS and a few applications
configured to use it correctly.  The Anonym.OS project demonstrated that this
is quite feasible, but their project is not currently maintained.

\subsection{A Tor client in a VM}
\tmp{a.k.a JanusVM} which is quite related to the firewall-level deployment
section below. JanusVM is a Linux kernel running in VMWare. It gets an IP
address from the network, and serves as a DHCP server for its host Windows
machine. It intercepts all outgoing traffic and redirects it into Privoxy,
Tor, etc. This Linux-in-Windows approach may help us with scalability in
the short term, and it may also be a good long-term solution rather than
accepting all security risks in Windows.

%\subsection{Interface improvements}
%\tmp{Allow controllers to manipulate server status.}
% (Why is this in the User Experience section?) -RD
% I think it's better left to a generic ``make controller iface better'' item.

\subsection{Firewall-level deployment}
Another useful deployment mode for some users is using {\bf Tor in a firewall
  configuration}, and directing all their traffic through Tor.  This can be a
little tricky to set up currently, but it's an effective way to make sure no
traffic leaves the host un-anonymized.  To achieve this, we need to {\bf
  improve and port our new TransPort} feature which allows Tor to be used
without SOCKS support; to {\bf add an anonymizing DNS proxy} feature to Tor;
and to {\bf construct a recommended set of firewall configurations} to redirect
traffic to Tor.

This is an area where {\bf deployment via a livecd}, or an installation
targeted at specialized home routing hardware, could be useful.

\subsection{Assess software and configurations for anonymity risks}
Right now, users and packagers are more or less on their own when selecting
Firefox extensions.  We should {\bf assemble a recommended list of browser
  extensions} through experiment, and include this in the application bundles
we distribute.

We should also describe {\bf best practices for using Tor with each class of
  application}. For example, Ethan Zuckerman has written a detailed
tutorial on how to use Tor, Firefox, GMail, and Wordpress to blog with
improved safety. There are many other cases on the Internet where anonymity
would be helpful, and there are a lot of ways to screw up using Tor.

The Foxtor and Torbutton extensions serve similar purposes; we should pick a
favorite, and merge in the useful features of the other.

%\tmp{clean up our own bundled software:
%E.g. Merge the good features of Foxtor into Torbutton}
%
% What else did you have in mind? -NM

\subsection{Localization}
Right now, most of our user-facing code is internationalized.  We need to
internationalize the last few hold-outs (like the Tor expert installer), and get
more translations for the parts that are already internationalized.

Also, we should look into a {\bf unified translator's solution}.  Currently,
since different tools have been internationalized using the
framework-appropriate method, different tools require translators to localize
them via different interfaces.  Inasmuch as possible, we should make
translators only need to use a single tool to translate the whole Tor suite.

\section{Support}

It would be nice to set up some {\bf user support infrastructure} and
{\bf contributor support infrastructure}, especially focusing on server
operators and on coordinating volunteers.

This includes intuitive and easy ticket systems for bug reports and
feature suggestions (not just mailing lists with a half dozen people
and no clear roles for who answers what), but it also includes a more
personalized and efficient framework for interaction so we keep the
attention and interest of the contributors, and so we make them feel
helpful and wanted.

\section{Documentation}

\subsection{Unified documentation scheme}

We need to {\bf inventory our documentation.}  Our documentation so far has
been mostly produced on an {\it ad hoc} basis, in response to particular
needs and requests.  We should figure out what documentation we have, which of
it (if any) should get priority, and whether we can't put it all into a
single format.

We could {\bf unify the docs} into a single book-like thing.  This will also
help us identify what sections of the ``book'' are missing.

\subsection{Missing technical documentation}

We should {\bf revise our design paper} to reflect the new decisions and
research we've made since it was published in 2004.  This will help other
researchers evaluate and suggest improvements to Tor's current design.

Other projects sometimes implement the client side of our protocol.  We
encourage this, but we should write {\bf a document about how to avoid
excessive resource use}, so we don't need to worry that they will do so
without regard to the effect of their choices on server resources.

\subsection{Missing user documentation}

Our documentation falls into two broad categories: some is `discoursive' and
explains in detail why users should take certain actions, and other
documentation is `comprehensive' and describes all of Tor's features.  Right
now, we have no document that is both deep, readable, and thorough.  We
should correct this by identifying missing spots in our design.

\bibliographystyle{plain} \bibliography{tor-design}

\end{document}