Extending Darknets
Extending Darknets
Aaron B. Helton
August 1, 2006
Extending Darknets via Mobile Networks 2
Abstract
The future of darknets, for good or ill, depends on their ability to adapt to new
technologies designed both to defeat them and to facilitate them. One of the best ways to
strengthen and extend darknets is through mobile ad-hoc networks, which can provide
robustness, and anonymizing software that can provide a fair amount of anonymity to users of
darknets.
Extending Darknets via Mobile Networks 3
I. An Overview of Darknets
The term “darknet” was coined in 2002 by four Microsoft employees. It is typically
used to describe one of the many peer-to-peer file sharing networks in wide use today. As
Biddle, England, Peinado and Willman state (2002), a “darknet is not a separate physical
network but an application and protocol layer riding on existing networks.” Ideologically,
however, such networks are thought to exist on the fringes of the regular, “legitimate” Internet.
Despite this, evidence suggests that usage is on the rise. For example, according to Slyck.com
(2006), users of some of the most popular peer-to-peer networks have reached almost ten
million, and the trend up to this point has only been upward. Furthermore, this is only a small
percentage of darknet traffic, since some other ad-hoc darknets, such as those created by
At the time of this writing, digital piracy remains the most prevalent use of darknet
technologies. So far, the largest deterrents to such uses have been legal and technological. The
media industries, largely comprised of Motion Picture Association of America (MPAA) and
Recording Industry Association of America (RIAA) member organizations, have waged a nearly
continuous war against file sharing networks since the days of Napster. Armed with existing
copyright law, which was strengthened considerably by the Digital Millennium Copyright Act
(DMCA) in 1998, the RIAA alone “has sued just over 18,000 individuals for sharing songs
online, with 4,500 settling for about $4,000 per case” (Graham, 2006). These cases so far seem
to have had little impact, given the overall rise in popularity of darknets.
Technological deterrents for digital piracy focus mainly on preventing or delaying “the
injection of new objects into the darknet” (Biddle, et al. 2002). Any litigation targeting
Extending Darknets via Mobile Networks 4
circumvention of such technological protection measures, or TPMs, are explicitly covered under
the DMCA itself. Fred von Lohman, speaking specifically about the DMCA, surmises that
TPMs as a solution to this problem may be doomed to fail” (2004). Thus the technological
impediments seem to have as much impact on the use of darknets as legal ones. Darknets simply
refuse to die.
Despite the negative connotations darknets conjure up, they nevertheless possess
significant beneficial uses, such as protecting free speech, distributing materials as widely as
possible and reaching much larger audiences than would be possible without such technologies.
Among these uses, the most beneficial would have to be protecting free speech, especially in
countries where such freedoms are not the norm. Regardless of the actual use of the darknets,
and the legitimacy of their uses can be debated, they still must provide their users with a certain
level of protection against those who would shut them down for whatever reason.
cartels, network technology must make provisions for the greatest amount of anonymity feasible
and reasonable robustness against denial of service attacks, man-in-the-middle attacks, and any
other points of failure. This can be achieved with a combination of existing technologies, not the
Mobile ad-hoc networking is not a new concept. Its roots lie in the “packet radio”
networks developed by DARPA in the early 1970s. Originally envisioned as the future of
Extending Darknets via Mobile Networks 5
battlefield communication, mobile networking has taken a new direction. More appropriate
terminology, according to the Internet Engineering Task Force’s (IETF) charter on mobile
networking, RFC2501, is “Mobile, Multihop, Wireless Networking” (Corson and Macker, 1999).
Specifically, the applications that will be most useful, such as linking mobile networks to the
protocol. Thus, routing and data transmission will occur within a well-known framework. Note,
however, that current wireless networking technology has not reached a point of maturity such
that mobile wireless networks can feasibly exist as anything but subnets inside a larger wired
network. The largest constraint here is additional routing complexity, as mobile devices would
also have to know about routes to the wired Internet. This will most certainly change in the near
future as mobile devices become more powerful and able to handle the more complex routing
schemes.
Regardless of the protocols used, all mobile networks share certain characteristics. The
topology is dynamic, with the nodes changing location rapidly and often unexpectedly. In such
networks, available bandwidth per device is limited, often due as much to power constraints as
the signal strength constraints. Also, security in such networks is only as good as existing
spoofing, and denial-of-service attacks” (Corson and Macker, 1999) to a greater extent than the
wired counterparts.
routing itself. Because a mobile network would be useless without the ability to reconfigure
routes on-demand, a number of approaches have been suggested. All approaches are similar in
Extending Darknets via Mobile Networks 6
their basic methodologies, in that they rely on small discovery packets to roam around the nodes
on the mobile network to establish and maintain a list of current routes. The two discussed here,
however, differ somewhat in their approach to securing against the types of attacks to which
Mobile routing algorithms come in three basic flavors. Table-driven algorithms “try to
maintain routes to all other nodes at all times” (Ducatelle, 2005). This approach is the same
approach taken by the Bellman-Ford equation, and differs greatly from the approach taken by
demand-driven algorithms, which “gather routing information when a data session to a new
destination starts, or when a route which is in use fails” (Ducatelle, 2005). Both methods have
advantages and disadvantages. The third variety, a hybrid approach, combines the reactive
nature of demand-driven algorithms with the reactive nature of table-driven algorithms. Two
such routing schemes are AntHocNet and ANODR (ANonymous On-Demand Routing).
AntHocNet uses reactive measures to establish routes when such routes are requested,
and proactive measures to attempt route improvement. Its basic methodology is based on the
concept of Ant Colony Optimization (ACO), a natural phenomenon by which ants establish
direct routes to food sources. While successful versions of such protocols exist already for wired
networks. To establish routes and to maintain paths, AntHocNet utilizes small control packets
called ants, which adaptively estimate the quality of each local routing choice (Ducatelle, 2005).
The result is a distributed route discovery and path maintenance that also has the desirable
long way toward maintaining the reliability reqired for use as darknets, especially if requested
information originates from within the mobile portion of the network and is to travel to any
Extending Darknets via Mobile Networks 7
number of wired destinations on the Internet. However, this is but one facet of the overall
ANODR is mechanically very similar to AntHocNet, in that it has both a reactive path
establishment phase and a proactive path maintenance phase. It also relies on path discovery
packets similar to AntHocNet’s ants. The major difference, however, is that in addition to
extends other protocols by providing an “untraceable and intrusion tolerant routing protocol for
mobile ad hoc networks” (emphasis original) (Kong, 2003). This is achieved by adding some
complexity to the algorithm such that actual node addresses are abstracted by a pseudonym, and
the actual node-specific information is included inside a one way hash (so-called trapdoor). This
requires a bit more overhead to process the routes, but has the effect of rendering the senders and
receivers untraceable. Because of this approach, “an on-demand ANODR route is traceable only
if all forwarding nodes en route are intruded” (Kong, 2003). Given the highly distributed nature
of the networks likely to use this technology, this kind of intrusion is unlikely. While this
technology has enormous and quite obvious benefits for military use, it adapts well for use with
the purposes behind darknets by masking the sender (but not necessarily the receiver, as will be
explained shortly).
Up to this point, the only robustness that has been discussed for mobile networks has
dealt mainly with maintaining paths so that single node failures do not necessarily result in
failures for the entire network. However, tactics that are likely to be used against mobile
networks that host darknet content are the same ones used against wired networks, and include
Extending Darknets via Mobile Networks 8
denial-of-service, eavesdropping, and spoofing. Each tactic has a specific goal, and the risk of
At its heart, a denial-of-service attack is exactly what it sounds like. The goal of such
an attack is to render the source of (in this case) objectionable content inaccessible to the rest of
the network. An effective attack requires a bit of information, and can take several forms, some
unique to mobile wireless networks. The very basic requirement for such an attack to succeed is
knowing where the target node is and being able to target it directly. In wired networks, simply
picking apart packets passing through the network is sufficient to get the IP address of senders or
receivers. This can work in one of two ways: by becoming a recognized node in that network, or
by a man-in-the-middle attack. These attacks will be discussed in the following paragraphs, but
it is important to mention them here, since they offer the best ways to get the IP address of an
offending node. Once this information is known, it is trivial to initiate the denial-of-service
attack.
exhaustion), wireless networks are susceptible to radio jamming (signal saturation or network
congestion) and battery exhaustion. With radio jamming, “an attacker can
deny service to the nodes in a given area by jamming the radio frequencies they
use” (Stajano, 1999). In terms of mobile wireless networks, especially those whose edges
also touch some edge of the physical Internet, that means the threat is doubled. If the attacker is
in range of the wireless device that is the target of the attack, then simply saturating the wireless
bands available to such devices would be sufficient in reducing or eliminating that node’s
availability. The drawback of this approach is that, unless it’s done by a despotic government (or
possibly any government), it carries potential legal risk. Thus, as a tactic against, say, file-
Extending Darknets via Mobile Networks 9
sharing, this tactic is unlikely to be used. The other possible approach to deny a node’s
availability is a network flood, in which the attacker attempts to overwhelm the target with more
packets than can be processed. This approach also carries a legal risk if used by the wrong
entities, and again probably would not be employed by just anyone. Furthermore, either of these
tactics, if used by a single attacker, also have a tendency to attack the attacker such that the
The far more likely scenarios for compromising wireless hosts are eavesdropping and
spoofing. Eavesdropping allows the attacker to capture information passively to help identify the
source or destination of any packets analyzed. While this is definitely problematic if the sender
and receiver wish to maintain some confidentiality, it is far eclipsed by the threat of spoofing.
Spoofing simply means that some other node has fooled other nodes into believing it is a
legitimate source or destination. In some cases, especially where darknets are concerned, the
host itself may not be the spoofing entity, but instead have spoofed files. This is a common
tactic used by the RIAA and MPAA in their battle against file sharing networks. By creating
files that look authentic, the attacker can fool requesting nodes into believing that the files are
indeed authentic. This seriously impacts the reliability of any darknet. At the more extreme end
of the spoofing spectrum, however, is the concept of simply masquerading one host as another.
Since this tactic is likely illegal, it probably would not be employed by anything other than a
government.
Now that several points of failure have been identified, and most of these require
knowledge of the target’s network location, how can these be guarded against? The short answer
Extending Darknets via Mobile Networks 10
is that, by creating a multi-path network that uses a secure routing protocol such as ANODR, the
availability, confidentiality, and integrity of the network can be maintained. Of course, there is
more to it than that, and for reasons to be discussed, a secure routing protocol may not be
advantageous.
In general, any robust routing protocol can be used, as long as it provides the possibility
of multiple paths and can perform efficient dynamic routing. Plenty of reasons exist to keep
extensive security processing outside the scope of a wireless device’s routing algorithm, not the
least of which is performance. Insofar as such networks can be useful for hosting or extending
darknets, speed is and will continue to be a major factor in the selection of any routing protocol.
Thus, from the standpoint of security, the routing protocol should only provide robust path
software solution be used on the sending and receiving nodes such that no intermediary node can
be aware of what is being transmitted. The most mature of these software technologies is onion
routing, specifically Tor. Developed by Roger Dingledine and Nick Mathewson of Free Haven
and Paul Syverson of the Naval Research Laboratory, the Tor project “aims to defend against
traffic analysis, a form of network surveillance that threatens personal anonymity and privacy,
confidential business activities and relationships, and state security” (2006). It does this “by
distributing your transactions over several places on the Internet, so no single point can link you
to your destination” (2006). In this way, determining which machine is the actual sender is very
difficult. Tor acts as a black box with a number of defined exit nodes. Traffic moving across the
Tor network is encrypted, then routed through random Tor nodes and exits the network (thus re-
entering the public Internet) from a similarly random exit node. This has several effects on such
Extending Darknets via Mobile Networks 11
traffic. First, since no node inside the Tor network can be entirely certain of the origin of any
Tor traffic, it cannot reveal that source. Second, since the data traveling around the Tor network
is encrypted, the contents of the data stream are unknown to anyone except the requesting party.
Third, since the exit nodes are used randomly, tracking a particular host’s communications
between a Tor network and the public Internet is almost impossible. Ideally, traffic would never
need to leave such secure networks, but then their usefulness on the whole might be diminished.
The reality is that there will always be some hosts that have defined exit nodes.
Extending such anonymizing technology to mobile darknets leaves a very robust, mobile
network that provides extensively for the anonymity of its users, even if they must interact with
the public Internet. Content distribution over such a network becomes trivial, while tracking
down the source of that content injection becomes very difficult, and in some cases infeasible.
Thus, the darknets of the future will likely be a mixture of wired and wireless mobile devices,
routing their traffic through some combination of Tor (or equivalent) networks and the public
Internet. Whatever the ultimate use of such technology, this combination will be very difficult to
combat, even for governments with the resources to meaningfully combat them.
Extending Darknets via Mobile Networks 12
References
Biddle, P., England, P., Peinado, M., & Willman, B. (2002). The Darknet and the Future of
Content Distribution. 2002 ACM Workshop on Digital Rights Management.
Corson, S., & Macker, J. (1999). Request for Comments: 2501. Mobile Ad hoc Networing
(MANET): Routing Protocol Performance Issues and Evaluation Considerations. Retrieved
July 11, 2006 from http://www.ietf.org/rfc/rfc2501.txt
Ducatelle, F., Di Caro, G., & Gambardella, L. M. (2005). Using Ant Agents to Combine
Reactive and Proactive Strategies for Routing in Mobile Ad-Hoc Networks. International
Journal of Computational Intelligence and Applications, Vol. 5, No. 2. pp. 169-184.
Imperial College Press.
Graham, J. (2006). RIAA chief says illegal song-sharing 'contained'. USAToday.com. Retrieved
August 1, 2006 from http://www.usatoday.com/tech/products/services/2006-06-12-
riaa_x.htm?POE=TECISVA
Hoffman, R., & Mennecke, T. (2006). Slyck’s P2P Network Stats Page. Retrieved August 1,
2006 from http://slyck.com/stats.php
Kong, J., & Hong, X. (2003). ANODR: Anonymous On Demand Routing with Untraceable
Routes for Mobile Adhoc Networks. International Symposium on Mobile Ad Hoc
Networking & Computing, Annapolis, Maryland.
Stajano, F., & Anderson, R. (1999). The Resurrecting Duckling: Security Issues for Ad-hoc
Wireless Networks. Proceedings of the 7th International Workshop on Security Protocols.
pp. 172-194.
Von Lohman, F. (2004). Measuring the Digital Millennium Copyright Act against the Darknet:
Implications for the Regulation of Technological Protection Measures. 24 Loyola of Los
Angeles Entertainment Law Review, 635.