ArticlePDF Available

Clean Feed: Australia’s Internet Filtering Proposal


Abstract and Figures

This article examines the Australian Federal government's proposal on internet filtering. Technical, policy and legal frameworks are discussed. A comprehensive comparative chart is provided looking at other countries which already have internet filtering in place.
Content may be subject to copyright.
Electronic copy available at:
Clean Feed: Australia’s Internet Filtering Proposal
Alana Maurushat* and Renée Watt**
Forthcoming Internet Law Bulletin March 2009
The Australian Proposal in the International Context
The Australian Proposal
The Australian government is seeking to make the internet a safer place, especially for
children. Cyber-safety was an electoral promise of the newly-appointed Labor
government.1 A key aspect of the Rudd government’s cyber-safety program, unofficially
known as Clean Feed, is a national internet filter.2 This means legally mandating Internet
Service Providers (‘ISPs’) such as Optus, Telstra and iiNet to implement technical means
to filter out a prescribed list of websites. Sites containing images of child abuse, in
particular child pornography, will be the initial focus of the list.
The government has not publicly stated its goals for the Clean Feed proposal; perhaps the
project’s aims remain unformulated even within the Labor Party. As there is no
Australian legislation yet on internet filtering, important details of the scheme remain
mysterious, and its configuration and scope may change. Once enacted, Clean Feed
legislation will be open to the amendment of future governments. If Parliament defeats
the eventual legislation, it is still possible that a number of ISPs will voluntarily
implement internet filtering.
The criteria for evaluation of websites to be blocked remains nebulous. As the proposal
currently stands, ACMA investigates complaints about ‘prohibited content’ based on a
classification scheme of X18, R18, RC, and MA15+ (though rare). Materials classified as
X18 or RC comprise the 1300 plus blacklist maintained by ACMA.3 Classification
decisions of such content is secret; the blacklist remains unknown. Senator Fielding, a
proponent of Clean Feed, has argued to expand the blacklist to include R18+ (eg. adult
pornography) The Clean Feed proposal has two tiers:
1) Blacklist Filtering: The first tier comprises the mandatory filtration for all
Australians (no possibility to opt-out) of sites on an ACMA-issued blacklist of
‘child pornography’ websites and ‘other prohibited’ materials, the scope of which
remains unknown. ISPs must block such sites at the URL level. Circumvention
of the blacklist will be illegal. Operating only on URLs, the filter will not block
websites with ‘child pornography’ and ‘other prohibited content’ found on:
Peer-to-peer systems (for example: bit torrent, Winny),
Electronic copy available at:
Encrypted channels,
MSN Instant Messaging, and
Mobile phones.
It remains unknown whether Clean Feed, when confronted with an offensive
URL, will block every website operating on a domain name or merely the specific
offending material. (This represents the difference for example between YouTube
– being the domain name – and a specific video on YouTube).
2) Content Filtering: The second tier, operating on an opt-out basis, will block
materials that are legal but potentially unwanted. The government has not
delineated the scope of such material but likely examples include adult
pornography and other ‘R’ rated material – content inappropriate for children but
legal for adults. Circumvention will not be illegal. The filtering techniques for
tier two filtering remain undefined. Potentially these could include URL
blacklists, deep packet inspection, peer-to-peer content inspection, and URL and
http content inspection.
National Internet Filters in Other Countries
Governmental internet filtering is the realm not only of authoritarian regimes like China
and Iran but also of such Western democracies as Canada, the United Kingdom, Sweden
and France. In Canada and the United Kingdom informal government pressure led the
countries’ major ISPs to ‘voluntarily’ institute internet filtering. The clearly articulated
goal of these systems is to block inadvertent access to websites containing child
pornography and child abuse materials. In countries such as France, Belgium and
Germany, courts have ordered ISPs to block hate speech and the illegal peer-to-peer
filesharing of copyright-protected materials.4 So Australia’s proposed system is unique
among Western democracies: we will be the first of such nations to mandate internet
filtering through formal legislation.
In Canada and the United Kingdom, using political influence instead of passing
legislation mandating internet filtering may be strategic. By removing the process from
the political arena Canada and the United Kingdom elude transparency and
accountability. Legislating internet filtering in these countries would open the provisions
to freedom of expression challenges under the Human Rights Act 1998 (United Kingdom)
and the Charter of Rights and Freedoms 1982 (Canada); it is doubtful that mandatory
internet filtering legislation would withstand such scrutiny.5 The same safeguards are not
present in Australia. There is no Bill of Rights, nor any kind of wide human rights
instrument covering freedom of expression. In theory, the constitutionally implied
freedom of political communication could prevent an overly broad legislative internet
filter, but case law indicates that in this context this right will be of limited consequence.6
The websites that Parliament blocks will thus be malleable to the political will and ethical
values of the governing party du jour.
Table 1: Comparison between Australia’s proposed internet filter and filters in
place in Canada, the United Kingdom and the People’s Republic of China.
Australia Canada United Kingdom People’s Republic of
filtering at ISP
Yes No No Yes – over 20 pieces
of legislation affect
Voluntary /
Industry filtering
at ISP level:
Perhaps – If
legislation fails,
ISPs may elect to
Yes – informal
Yes – informal
Yes – corporate self-
censorship is
Provision: No – Tier 1
Yes – Tier 2 No No No
Project Name: Clean Feed Cleanfeed Canada
Project Cleanfeed Golden Shield Project
known as ‘The Great
Firewall of China’
Filtering of
Blocked URLs:
Yes Yes Yes Yes
Purpose of
Blacklist: Unspecified Blocking
inadvertent access
to child
materials with http
inadvertent access
to child
materials with http
To block various
types of illegal
Type of
Child pornography
and other
unknown ‘illegal
pornography Child
pornography Political content,
graphic violence,
unapproved news
stories, child
pornography, and
other illegal content
Maintained By: ACMA
and Media
Authority) Internet Watch
Foundation Ministry of Industry
and Information;
Central Propaganda
Ministry of Posts
IP Address
Blocking: Not at this time Not at this time Not at this time Yes
Deep Packet
Inspection: Not at this time Yes Yes Yes
Purpose of Deep
NA Traffic Shaping Traffic Shaping Traffic Shaping,
Dataveillance &
Other Heuristic
Methods: Yes Yes Yes Yes
P2P: Not at this time Perhaps Perhaps Yes
infringement (in
negotiation with
music industry)
infringement (in
negotiation with
music industry)
Messaging: Not at this time Not at this time Not at this time Yes
Scope Creep: Inevitable Yes – suicide
sites, pro-
terrorism sites,
hate sites
Yes – suicide
sites, graphic
beheadings, pro-
terrorism sites,
hate sites
Yes – legislation
written with standard
vague and ambiguous
clauses such as the
‘state security’
Offence to
Yes – not an
offence to use a
device such as a
proxy for other
No No Yes – not an offence
to use a
circumvention device
such as a proxy for
other purposes
Safeguards: No – no Bill of
implied freedom
of political
very limited in this
context and of
little use as a
Limited – Charter
of Human Rights
does not bind
corporations such
as ISPs (no
compelling ISPs)
Convention on
Human Rights;
relevant case law
from the European
Court of Human
No – the human
rights instruments are
of little practical
significance (Eg.
Freedom of
Expression is not an
individual right)
Safeguard: No – compulsory
for all ISPs Potentially
initiative subject
to strong informal
initiative subject
to strong informal
Safeguard: No Potentially
depends where the
filtering routers
are placed (Eg.
router located on
the backbone
would affect all
depends where the
filtering routers
are placed (Eg.
router located on
the backbone
would affect all
geographical region
of access, and
bandwidth capability
affect ability to
access materials
Technical and Legal Limits of Internet Filtering
General Attributes of Internet Filtering
One of the many myths surrounding the debate on internet filtering is that a filter
operating at the ISP level will be more effective than filters that sit on the operating
system of a personal computer. Up until 31 December 2008, these were freely available
from the government under NetAlert’s National Filter Scheme7. These filters operate on
the exact same principles and will block identical websites. In short, the only difference
is that a filter on an ISP will cost more and can be made mandatory for internet users
One common problem with blacklist filters is the degradation of network speeds. The
pilot tests of filtering in Tasmania found decreases in access speeds ranging from 3% to
86%.8 The extent of degradation depends on the amount of broadband available, passive
(filter not in use) versus active (filtering of website on blacklist), and most importantly,
on the type of filtering technology utilised.9 Statistics concerning the loss of network
speed in areas with limited broadband, such as rural Australia, remain woolly. A
mandatory national internet filter lies in stark contrast with the government’s
commitment to increase both broadband access and access speeds.10
As well, there are structural, logistical obstacles to the Australian implementation of a
national internet filter. The internet in Australia developed using a free-form model with
no central nodes or gateways. Internet traffic in China and Saudi Arabia, by way of
contrast, flows through set ‘choke points’.11 This allows China, for example, to “deploy
its networks to allow censorship to occur at multiple control points, from international
gateways to the network backbone to regional network providers”.12 In the Australian
context however:
How this distributed model compares to centralized filtering in terms of
performance and effects on network speeds is not certain; filtering could be
faster if placed closer to the network edge or faster if placed closer to the
core where it could achieve efficiencies of scale and redundancy. It can also
increase the challenge of keeping block lists up to date; as there are more
network points where the lists are implemented, it becomes more
challenging to ensure each list is properly updated.13
As of December 24, 2008, most ISPs in Australia are participating in trials of the
filter. The results of these trials are yet to be finalised. It is unclear if they will be
publicly available.
What Constitutes Child Pornography?
There is no universally accepted definition of “child pornography”. The term is
ambiguous even within Australia, where State and Federal definitions vary. At a
Commonwealth level,
Child pornography material means material that depicts a person, or a
representation of a person, who is, or appears to be, under 18 years of age
and who is engaged in, or appears to be engaged in, a sexual pose or sexual
activity (whether or not in the presence of other persons); or is in the
presence of a person who is engaged in, or appears to be engaged in, a
sexual pose or sexual activity; and does this in a way that reasonable
persons would regard as being, in all the circumstances, offensive.14
Child pornographic materials include written narratives, animated cartoons (such as
manga), and fictional depictions of abuse.15 The recent decision of McEwen v Simmons16
establishes that under New South Wales17 and Commonwealth18 law depictions of sexual
acts among the children characters of the American cartoon “The Simpsons” constitute
child pornography. Early last year police removed Australian artist Bill Henson’s
photographs of naked prepubescent girls from a gallery under suspicion that the works
were ‘child pornography’.19 Both situations have raised public concern over the scope of
the definition.
URL Blacklist Filtering
URL blacklist filtering applies to the world wide web and not to the internet as a whole,
meaning that the list will only block sites employing http: protocol (that is, starting with
www) but will not target filesharing, chatrooms, password protected sites, encrypted
images, or any other digital transmission or storage of prohibited material. The blockage
of websites carries with it the very real possibility of collateral damage: an attempt to
block an offensive video on YouTube could potentially temporarily block all content on
YouTube, for example.
Similarly, in December 2008 the internet filtering system in the United Kingdom
temporarily closed Wikipedia when Internet Watch Foundation (‘IWF’), a watchdog
group equivalent to ACMA, found a Wikipedia-published image of a 1976 album cover20
to be “a potentially illegal child sexual abuse image”.21 The cover features a naked
prepubescent girl in a sexually provocative pose: a possible contravention of the
Protection of Children Act 1978 (United Kingdom). Nevertheless neither the British nor
it seems any other courts have deemed the image illegal, and earlier in 2008 Wikipedia
declined to censor the image as requested by an American complainant. The IWF did not
blacklist any other sites featuring the image, such as It did, however,
blacklist not only the image’s URL, but also that of the accompanying Wikipedia article,
which touched on the cover’s controversy but otherwise described the album’s music.
The album (including said cover) is available in British stores.
Blacklisting the site accidentally blocked 95% of the United Kingdom’s residential
internet users from contributing to Wikipedia.22 Publishing material that contravenes
Wikipedia’s own guidelines on what is inappropriate triggers an internal anti-vandalism
mechanism that prevents such participants from further contribution. The United
Kingdom’s filter requires all ISPs to reroute via a small number of proxy servers – that is,
when an internet user accesses a site, the request passes first through a “middleman”
computer that filters blacklisted content. In these circumstances Wikipedia is unable to
identify users’ individual TCP/IP (the code that defines individual computers’ internet
access), instead attributing the IP of the remote proxy to any user trying to access
Wikipedia via that proxy. By instituting a ban in the UK’s major ISPs, the IWF
instigated a blanket ban an all users employing those ISPs.
Responding to public outcry, the IWF, after a review process, confirmed the image to be
potentially an offence against the Child Protection Act, but that given the “contextual
issues … and … the length of time the image has existed and its wide availability”23
removed the URL from its list.
Content Filtering
Content filtering (as yet not part of Australia’s Clean Feed proposal) involves a number
of heuristic methods including: deep packet-inspection, keyword sniffing, traffic-
shaping, and algorithms developed to detect, absent human intervention to detect illegal
content. Inaccuracy, over-breadth and invasion of privacy render these techniques highly
Concluding Remarks
Absent a clearly stated purpose, it is difficult to assess Australia’s Clean Feed proposal,
save to say that it will require a large budget required to initialize such a system,24
potentially significantly reduce network speeds and interfere with people’s fundamental
access to information, with no definite purpose in mind. If the policy’s aim is to prevent
inadvertent access of prohibited material, the first question must surely be: how often do
Australians stumble across such content? If we are concerned about children deliberately
accessing unsavoury materials, a less expensive yet equally effective solution is to keep
the household computer in the lounge-room, and install a personal filter. If we are
concerned about deliberate adult access to child pornography, is a technologically-clumsy
filter really up to the job? Could this instead force such users to encrypt illegal materials,
making policing and prosecution increasingly difficult?
Passionate argument adorns on all sides the debate surrounding Canberra’s proposed
internet filtering system. Missing from all voices however is evidence-based policy;
equally missing is its evidence-based refutation. Hard data and sound research about
cybercrime, ‘prohibited materials’ and information security are in shortage. Research
typically comes from partisan parties such as law enforcement agencies, government,
religious groups, systems administrators and the media, and methodology is often absent.
For example, the Australian Communications and Media Authority’s report,
‘Developments in Internet Filtering Technologies and Other Measures for Promoting
Online Safety’ reaches sweeping assessments of the effectiveness of types of filters and
materials, but does not provide any information about the scope, sample size, year of the
study and who performed the study.25 Crime statistics in the online environment are
difficult due to novelty and lack of transparency. Whatever the policy, it lacks evidential
It is this author’s hope that governments around the world will start by funding non-
partisan research into the many important areas underlying cybercrime before spending
money on unpredictable schemes with no specified intended outcomes.
* Alana Maurushat is Lecturer, Deputy Director of the Cyberspace Law and Policy
Centre, and PhD Candidate – all within the Faculty of Law, the University of New South
Wales. The author is indebted to David Vaile, Derek Bambauer, and Lillian Edwards for
valuable discussion.
** Renée Watt is research intern with the Cyberspace Law and Policy Centre, Faculty of
Law, the University of New South Wales.
1 Stephen Conroy (then Shadow Minister for Communications and Information
Technology), Labor’s Plan for Cyber-Safety, Election 2007, at
2 Senator Stephen Conroy, Budget provides policing for internet safety, media release, 13
May 2008, at <>
3 The Classification Scheme of the Australian Communications and Media Authority may
be found at
4 For France, see: TGI Paris, 19 octobre 2007, SARL Zadig Production, Jean-Robert V. et
Mathieu V. c/ Sté Google Inc. et AFA and the case of LICRA et UEJF V. Yahoo! Inc. and
Yahoo France, Tribunal de Grande Instance de Paris (Superior Court of Paris 2000). For
Belgium, see: Google Inc v Copiepresse SCRL [2007] E.C.D.R. 5. For Germany, see
case numbers: 308 O 42/06 and 308 O 248/07 (both cases heard at the Hamburg Regional
5 With respect to Canada: Colangelo A & Maurushat A, “Exploring the limits of
computer code as a protected form of expression: a suggested approach to encryption,
computer viruses, and technological protection measure” McGill Law Journal, 22 March
6 Catch the Fire Ministries Inc & Ors v Islamic Council of Victoria Inc [2006] VSCA
284; Lange v Australian Broadcasting Commission (1997) 189 CLR 520; Michael
Brown& Ors v Members of the Classification Review Board of the Office of Film and
Literature [1998] FCA 319; NSW Council for Civil Liberties Inc v Classfiication Review
Board (No. 2) [2007] FCA 896; Theophanous v Herald & Weekly Times Ltd (1994) 182
CLR 104.
7 NetAlert (Australian Government), ‘Internet Content Filters’, at:
8 Australian Communications and Media Authority, ‘Closed Environment Testing of ISP-
Level Internet Content Filtering’ 41, 62-68, June 2008, at
9 Australian Communications and Media Authority, ‘Closed Environment Testing of ISP-
Level Internet Content Filtering, available at
10 Australian Labor Party ‘Australian Labor Party National Platform and Constitution
2007’ April 2007, at <>
11 OpenNet Initiative ‘China: Country Profiles’ 9 May 2007 at:
12 Bambauer D ‘Filtering in Oz: Australia's Foray into Internet Censorship’ Brooklyn Law
School, Legal Studies Paper No. 125 22 December 2008 at; Gutmann E ‘Losing The New China’ 2004, 127-32;
OpenNet Initiative ‘Internet Filtering in China in 2004-2005: A Country Study’ at
13 Bambauer D ‘Filtering in Oz: Australia's Foray into Internet Censorship’ Brooklyn Law
School, Legal Studies Paper No. 125 22 December 2008 at
14 Criminal Code Act 1995 (Cth) s 473.1
15 View ACMA’s guidelines for prohibited content here:
16 McEwen v Simmons [2008] NSWSC 1292
17 Crimes Act 1900 (NSW) s 91H(3)
18 Criminal Code Act 1995 (Cth) s 474.19(1)(a)(i)
19 Wilson A & Wilson L ‘Gallery Raid Revives Censorship Issue’ The Australian 24 May
2008 at: <,,23749181-16947,00.html>
20 German band Scorpions’ cover for the 1976 album “Virgin Killer”
21 Internet Watch Foundation, ‘IWF Statement Regarding Wikipedia Page’ December
2008 at: <>
22 Doctorow C ‘How to Make Child-Porn Blocks Safer for the Internet’ The Guardian 16
December 2008 at: <
23 Internet Watch Foundation, ‘IWF Statement Regarding Wikipedia Page’ December
2008 at: <>
24 The current government has pledged $125.8 million to “protect children online”:
Senator Conroy ‘Budget Provides Policing For Internet Safety’ 13 May 2008 at:
25 Australian Communication and Media Authority ‘Developments in Internet Filtering
Technologies and Other Measures for Promoting Online Safety’ February 2008 at
26 The OpenNet Initiative's reports in "Access Denied" do an excellent job of canvassing
filtering, including in Western democracies:; both ONI's
blog, at, and Nart Villeneuve's blog, at, are
terrific, less formal resources; for an interesting article on the negative impacts of freely
available pornography see: <
... The UK model has also been influential elsewhere, and the name "Cleanfeed" has stuck as a generic term for UK blocking systems as well as related schemes in Canada and Australia (see e.g. Watt & Maurushat 2009). ...
One of the most important trends in internet governance in recent years has been the growth of internet blocking as a policy tool, to the point where it is increasingly becoming a global norm. This is most obvious in states such as China where blocking is used to suppress political speech; however, in the last decade blocking has also become more common in democracies, usually as part of attempts to limit the availability of child abuse images. Numerous governments have therefore settled on blocking as their “primary solution” towards preventing such images from being distributed.Child abuse image blocking has, however, been extremely controversial within the academic, civil liberties and technical communities. Why have these systems been so controversial? Two lines of criticism can be identified, which might be termed the practical and the principled. The practical argument claims that blocking is ineffective, with ill-defined goals and easily evaded by widely available circumvention technologies. The principled argument, on the other hand, is that blocking systems undermine the norms associated with freedom of expression in democratic societies.This chapter introduces and evaluates these claims by examining three prominent examples of child abuse image blocking – the United Kingdom Internet Watch Foundation Child Abuse Image Content list, the European Union sponsored CIRCAMP system and United States hash value systems. It discusses the operation of each and the extent to which the critics’ concerns are borne out. It concludes by considering the lessons which might be learned for proposals to extend blocking to other types of content.
Australia's decision to implement Internet censorship using technological means creates a natural experiment: it can become the first Western democracy to mandate filtering legislatively and to retrofit it to a decentralized network. But are the proposed restrictions legitimate? The new restraints derive from the Labor Party's pro-filtering electoral campaign, though minority politicians have considerable influence over policy. The country has a well-defined statutory censorship system that may, however, be undercut by relying on foreign and third-party lists of sites to be blocked. While Australia is open about its filtering goals, the government's transparency about what content is to be blocked is poor. Initial tests show that how effective censorship is at filtering prohibited content—and only that content—will vary based on what method ISPs use. Though Australia's decision-makers are formally accountable, efforts to silence dissenters, outsourcing of blocking decisions, and filtering's inevitable transfer of power to technicians undercut accountability. This Article argues that Australia represents a shift by Western democracies towards legitimating Internet filtering and away from robust consideration of the alternatives available to combat undesirable information.
Michael Brown& Ors v Members of the Classification Review Board of the Office of Film and Literature
Michael Brown& Ors v Members of the Classification Review Board of the Office of Film and Literature [1998] FCA 319; NSW Council for Civil Liberties Inc v Classfiication Review Board (No. 2) [2007] FCA 896; Theophanous v Herald & Weekly Times Ltd (1994) 182 CLR 104.
Australian Government), 'Internet Content Filters
  • Netalert
NetAlert (Australian Government), 'Internet Content Filters', at: <>
Closed Environment Testing of ISP-Level Internet Content Filtering' 41
Australian Communications and Media Authority, 'Closed Environment Testing of ISP-Level Internet Content Filtering' 41, 62-68, June 2008, at <>
How to Make Child-Porn Blocks Safer for the Internet' The Guardian
  • C Doctorow
Doctorow C 'How to Make Child-Porn Blocks Safer for the Internet' The Guardian 16 December 2008 at: <>