Ever since the advent of mainstream file-sharing networks in the late
1990s, the major music and movie studios and industry trade groups
(henceforth referred to as “the content industry”) have been
searching for a “magic bullet” solution to the problem of
online copyright infringement—one that would simultaneously
eradicate the problem of online file sharing while breathing new life
into pre-digital business models. Obviously, this magic bullet has eluded
the industry thus far, though not due to a lack of effort. In the years
since the emergence of Napster, the content industry has taken legal
action against service providers and end users, shut down centralized
file-sharing networks and flooded the web with bogus copies of commonly
traded files. Yet, as even a casual observer can attest, illicit file
sharing persists online.

The industry’s latest silver bullet comes not in the form of a
legal campaign or disruption strategy but rather in the form of a
technology that falsely promises to automatically and effectively
eradicate copyright infringement online. Copyright filtering, as it is
called, is a method whereby network appliances use a technology known as
Deep Packet Inspection (DPI) to inspect the data that travels over an
Internet Service Provider’s (ISP’s) network, identifying
content as it passes through the filter and then dealing with that
content accordingly. Unsurprisingly, the content industry has become an
outspoken advocate for the use of this technology, pushing governments
the world over to pass legislation requiring its use on ISP access
networks.

Here in the United States, the content industry was almost successful in
forcing language into the American Recovery and Reinvestment Act of
2009^[1] that would
have allowed ISPs to engage in copyright filtering under the auspices of
“reasonable network management,” an act that would have had
far-reaching consequences for citizens, businesses and the entire
Internet ecosystem.^[2] Meanwhile, based on information publicly
available,^[3] a
secretive, multilateral, international trade agreement known as the
Anti-Counterfeiting Trade Agreement (ACTA)^[4] could also pave the way for copyright filtering
in the United States, the European Union, Australia, Japan, Canada and a
number of other nations.

Before we rush to implement a technology—much less require its
use—we must objectively examine that technology and question not
only its efficacy but also its costs, consequences and drawbacks. It is
in this spirit that Public Knowledge, a non-profit public interest
advocacy group, has embarked upon an analysis of copyright filtering
technologies. The results of that analysis are presented in this
whitepaper.

In the following paper, we will take a close look at the technology
behind copyright filtering, analyze the legal ramifications of filtering
and discuss the policy implications of condoning or mandating the use of
copyright filters. We will consider the impact that filtering is likely
to have on user privacy, examine the costs associated with filtering and
contemplate the potential impact copyright filtering will have on network
security for both users and service providers.

Regardless of how they are implemented, copyright filters will alter the
fundamental behavior of the Internet and in so doing, will likely disrupt
the Internet ecosystem in ways that we cannot predict. The Internet was
designed to be an open system from end-to-end, which is to say, a system
that moves content between hosts and clients as quickly as possible on a
first-come-first-served basis—regardless of the nature of that
content. Copyright filters will inject delay into this system, make
automated judgments regarding the legality of content and will then
degrade or discard that traffic accordingly. The Internet was not
designed to support this type of activity and for this reason, the
implementation of copyright filters will assuredly result in a variety of
technical problems for all parties involved in the Internet ecosystem.
Much like traffic lights on our interstate highway system, copyright
filters on our open, high-speed networks will be a poor fit.

Technical issues aside, there are other pressing questions that must be
answered before we can determine whether copyright filters should be
installed on the networks of Internet service providers. Chief among them
is the question of efficacy: will copyright filters solve the problem
that they purport to solve? Based on our technical analysis, the answer
appears to be no. By virtue of their design, Internet filters are doomed
to be both underinclusive and overinclusive—they will fail to
identify all illegal uses of content while simultaneously blocking legal
content. The filters will be underinclusive because their technology is
not advanced enough—and will likely never be advanced
enough—to identify every instance of prohibited content on the
network. Filters will also be overinclusive; as a filter will never be
able to distinguish between fair, legal uses of content and illegal uses
of content with 100 percent accuracy. Given that even legal scholars and
courts are often unable to reach a consensus on questions of fair use,
this is not surprising—the question of whether or not a piece of
content constitutes fair use is often a difficult one for even human
beings to answer. Furthermore, as history attests, users will work to
actively circumvent the filter, thereby luring the architects of the
filter into a fruitless technological arms race.

Ultimately, we will demonstrate that copyright filtering does not
constitute “reasonable network management,” as some
proponents of the technology would have policymakers believe; rather,
copyright filtering is content management. Instead of making
determinations based on how data moves over the network, copyright
filters attempt to ascertain what that data constitutes, in
order to block, degrade or delay certain types of content. Copyright
filtering is not necessary for a network to operate properly or
efficiently and therefore, should not be considered a form of network
management at all.

The content industry would like to convince policymakers and the general
public that copyright filtering is the most effective means by which to
combat online copyright infringement and protect America’s creative
economy. This could not be further from the truth. In practice, copyright
filtering is likely to harm innovators, end users, online service
providers and Internet service providers alike. What’s more, it
will compromise the privacy of all American Internet users for the
perceived benefit of one industry. As such, copyright filtering will
discourage investment in the Internet economy—our most promising
engine for economic growth—and will harm American competitiveness
in the global market.

Finally, copyright filtering holds the potential to undermine the goals
of the National Broadband Plan (NBP), a Federal initiative to increase
the speed, adoption and affordability of broadband Internet services
nationwide. As we will see, copyright filtering, if implemented on U.S.
networks, will chip away at each of the NBP’s stated goals. The
network appliances that act as filters will slow traffic on the networks
on which they are installed, an unavoidable consequence of their traffic
analysis. The costs associated with filtering—most notably, the
purchasing and maintenance of filtering hardware and software—are
likely to be passed on to consumers, decreasing the affordability of
broadband services. And in compromising the privacy of Internet users,
copyright filters will discourage Internet use, even as the Federal
government works to promote the educational, economic and civic benefits
of broadband access.

If we are serious about combating copyright infringement online, we
should invest our time and resources in developing methods that will
effectively discourage illegal activity, without harming fair users,
innovators and the Internet economy. If we rush to require the use of a
technology that we have not fully considered, we will find ourselves
trapped by a burdensome and even dangerous mandate, one that will have
far-reaching consequences that will affect the flow of information,
knowledge, political discourse and capital. To date, the content industry
has presented the debate surrounding copyright filtering as a choice
between a specific technology and rampant online “piracy”. In
the paper that follows, we will demonstrate that this is a false choice
and that copyright filtering should not be considered as a viable
solution for U.S. ISPs.

^[1] See the
Recovery.gov website (http://www.recovery.gov/?q=content/act).

^[2] “Senator
Feinstein Trying to Sneak ISP Copyright Filtering Into Broadband Stimulus
Bill,” TechDirt, Febuary 10, 2009 (http://www.techdirt.com/articles/20090210/1050313726.shtml).

^[3] “Secret
ACTA Treaty May Include ISP Filtering,” Ars Technica, June
4, 2008 (http://arstechnica.com/tech-policy/news/2008/06/secret-acta-treaty-may-include-filtering-provisions.ars).

^[4] For more
information on ACTA, see the Electronic Frontier Foundation website
(http://www.eff.org/issues/acta).