The tourist complained to Te Papa. They initially tried to fob him off, but eventually he got through to someone and those sites were removed from the filter. A good outcome, right?

Not So Simple

This incident raises a number of questions:

• Why is Te Papa filtering what people see on the internet?
• What type of content is being blocked?
• Who chooses which types of content to block?
• Finally, why are they using software that flags a German political website as "Pornography (Japanese)"?

Click on the image to see it fullsize.

Why censor internet access?

We spoke to Te Papa but they couldn't tell us why they felt the need to censor their wireless. They did know that they blocked file sharing protocols to reduce internet traffic but couldn't tell us why they were blocking some websites. We'd understand if Te Papa wanted to use some censorware on internet terminals available to children, but their filter goes far beyond that.

Are they worried that people will somehow download banned material? It's not their responsibility and it's not like they're monitoring phone calls to make sure people don't have illegal conversations.

Are they worried that people will browse offensive material (pictures/video) in a public place and annoy others? An increasing number of their guests have smartphones and "bring their own internet" and someone could as easily watch a porn DVD on a portable player. In any of these cases, it would be a simple matter of asking them to stop.

We reject the idea that internet providers (for that is what Te Papa is doing by providing free wireless) are in any way responsible for what an internet user does with that connection, in the same way that they aren't responsible if someone uses Te Papa provided water or electricity.

Te Papa's Filter

Te Papa could tell us that they are using internet filtering supplied by their internet service provider, Telstra Clear, but they had very little idea about how it works.

• They don't know why they're blocking some types of content.
• They don't know what type of content is being blocked.
• They don't know who decides what to block and what criteria they use.
• They don't really want to find out, saying that they're "happy for them [Telstra Clear] to make the decisions".

Any museum and art gallery is surely aware of issues around censorship and free speech, Te Papa itself has been involved in certain controversies about what should be shown and to who. Why has Te Papa chosen to censor the internet with so little thought about why and how? As our visiting tourist put it:

Seeing this happen at Te Papa, a flagship of the capital, tells me something about democracy and the importance of free speech and human rights in NZ.

Our view

We tend to side with the visiting German tourist - it's inappropriate for a place like Te Papa to be censoring the internet.

We suggest that worries about people accessing "bad material" over public internet are overstated. Any inappropriate behaviour (e.g. viewing internet pornography in a public place) can be solved by asking them to stop.

If an organisation decides to press on with censorship anyway, it would seem at a minimum that they should:

• Be able to tell people what sort of material is blocked and why they're doing it.
• Have a process for deciding what to block.
• Provide an easy way to appeal any incorrect blocking.
• Not use software that is as badly written as that used by Te Papa and TelstraClear.

Of course, once you look at all that, doesn't it just seem easier to let people have unconstrained internet access in the first place?

The Digital Child Exploitation Filtering System is New Zealand's internet filter, run by the Department of Internal Affairs. If you're with one of the participating internet service providers you can't access the content on the blacklist. It's meant to only be used to block images of child sexual abuse (and “think of the children!” gains buy-in very effectively), but unlike other censorship decisions which must be released, the list and process is secret, so no one really knows what's blocked, and overseas experience dictates that filters rarely work as first stated.

Apart from the secret list, and the secret processes around the list, the filter is meant to be open and transparent. Here's a quote from the February 2010 Independent Reference Group minutes:

“The Group suggested that the Department publish as much information about the system as possible. This would include regular statistics and a copy of the presentation [in the presentation given to the IRG, the operation of the filtering system, the compilation of the filtering list, and the appeal process was explained].”

The presentation never made it to the DIA website, so I requested it. Sticking with the trend of being a complete mess in regards to keeping records, they have lost it.

Here's a similar Powerpoint they sent, shown at the NetSafe Conference in April 2010 (pdf).

To encourage some additional transparency, I submitted an Official Information Act request last month, along with Joshua Grainger. If you'd like to see the full responses, they're here and here (pdfs).

Scope of the filter

From the February 2010 minutes:

“The Department has no intention to expand the scope of the filter beyond child sexual abuse images and has entered into a contractual agreement with the system’s supplier that it not do so.”

From the letter to ISPs advising them of the filter's availability (pdf):

“The Department recognises that, to ensure public confidence in the DCEFS, the scope of the system must remain on child sexual abuse material and its operation must be open to scrutiny. Accordingly, the Department's contract for the use of the software that supports the DCEFS constrains its use to filtering to child sexual abuse material.”

From the Common Questions and Answers page:

What assurances are there that the filter will not in future be extended to block content other than that intended?
The Department’s contract for the use of the software that supports the DCEFS constrains its use to filtering child sexual abuse material.

I requested the section of the filter contract that discusses the limitations of the filter (the DIA refuse to provide the full contract). I received a summary of it:

“While the Department has previously refused to release the whole contract with Netclean, it has referred to clauses in that contract as one of the reasons why the scope of the filtering system can't expand. The following is a summary of the relevant conditions of the Customer Licence Agreement.

• The primary goal of the NetClean Whitebox is to block access to child pornography.
• In order to achieve the main objective, NetClean allow that even non-child pornography is filtered, as long as it is material which is illegal to possess under the country's law and that the main objective for the installation is to block access to child pornography.
• The filter must not be used to restrict freedom of expression, nor to prevent the transmission of information which in itself is legal to possess.
• Furthermore, the installation of NetClean Whitebox must not violate the articles 18 and 19 of the Universal Declaration of Human Rights.”

Does this mean just child sexual abuse material that isn't child pornography can be blocked (the DIA say that the “bad” content is wrongly called child pornography)? Or anything illegal?

To me it seems like it's wide open:

“NetClean allow that even non-child pornography is filtered”.

Appeals and anonymity

From the Code of Practice (pdf):

“5.4 The process for the submission of an appeal shall:
• be expressed and presented in clear and conspicuous manner;
• ensure the privacy of the requester is maintained by allowing an appeal to be lodged anonymously.”

I wondered how, if appeals are meant to be anonymous, the DIA can process appeals when no URL is given by the appellant.

The DIA says:

“If a user does not submit a URL when appealing, the Department does its best to identify the site that appellant was referring to. This is done by looking at the block logs to identify sites blocked shortly before and after the appeal form was accessed and ISP of the appellant. The sites identified during that period will then be reviewed.”

This makes sense because blocks probably don't happen that often. However, if the ISP of the appellant can be compared, there's still other information being collected with appeals. It doesn't seem like this actually happens though.

Here's what one of the appeal reports says:

“Checked logs for sites blocked between 13:19 and 13:21. Sites in that timeframe identified as …”

No mention of ISP comparison.

Number of sites filtered

The DIA stated in a 2009 press release that the number of sites being filtered was over 7000. This dropped down to 400-700 in 2011, which is much closer to the 500 or so URLs the Internet Watch Foundation blocks.

The DIA on the massive decrease:

“The number quoted in the press release on 16 July 2009 related to the sites on the list during the trial. As there was a 2 year gap between the trial and going live with the filter system, the majority of the sites were taken down or had ceased to operate.”

So were the URLs being blocked not being regularly reviewed during the trial, and that's why there was such a high number, or did the policy on what sites could be filtered change?

In an email from Peter Pilley at the DIA to Richard Baalham, Networks Manager at Callplus (Slingshot), he claims:

“During the trial we had 0 false positives as we [reviewed] the list each month to ensure it [was correct] and current.”

The high number of 7000 could have definitely influenced ISPs to join the filter. Here's what Allan Freeth, the CEO of TelstraClear sent to Rick Barker, the Minister of Internal Affairs regarding TelstraClear joining:

“We will add a filter to all web browsing by Clearnet and Paradise customers that stops browsers from accessing known child sex abuse sites around the world. There are more than 7,000 such sites…”

Another thing that was curious about the number of sites filtered was that no URLs were removed from the filter between April 2011 and August 2011. All URLs are supposed to be reviewed monthly, and it seems unlikely that none were due to be removed.

When asked whether the list was reviewed monthly each month in 2011, the DIA said:

“Yes, it is this strict checking that is currently keeping our list so low.”

When asked why no URLs were removed between April 2011 and August 2011:

“On review, all sites continued to contain abuse material and subsequently had not been taken down by enforcement agencies in other countries or were still under investigation?”

“Were still under investigation” seem to be the key words in this answer. It's pretty clear-cut whether a site still contains child sex abuse images, and should only take seconds to check. An investigation isn't required, and it's more likely that some sites just weren't checked at all, thus still being “under investigation”.

I raised the prospect last month that if child sex abuse sites could identify DIA access to the sites they could provide innocuous content to the DIA instead of the child sex abuse content that they're showing to everyone else. The DIA say that in January 2012 they received 449 URLs for review from the public through the ChildAlert site.

In the NetSafe Powerpoint it says that reports through ChildAlert are received and pre-processed by ECPAT to remove false reports and filter bad reports.

This means that a high proportion of URLs received by the DIA should be able to be filtered.

Only 50 URLs that were submitted by the public in January 2012 were subsequently added to the filter list.

(If you're curious, they say that they reviewed over 500 URLs in January 2012, excluding the monthly review of all URLs on the filter list. That included 21 URLs supplied through the Interpol Worst Sites Project with “a number of other sites” that came to their attention as a result of investigations also being reviewed.)

Material filtered

A Russian child model website was discussed during the IRG's March 2011 meeting:

“The Group was asked to look at a child model website in Russia. The young girl featured on the site appears in a series of 43 photo galleries that can be viewed for free. Apparently the series started when the girl was approximately 9 years old, with the latest photographs showing her at about 12 years old. The members’ part of the site contains more explicit photos and the ability to make specific requests. While the front page of the website is not objectionable, the Group agreed that the whole purpose of the site is to exploit a child and the site can be added to the filter list.”

When asked for anything held regarding this website, and whether the website was considered a case of clearly illegal, objectionable images of child sexual abuse, the DIA replied:

“The website in question is divided into a public area and a member-only area. The public area contains images of a young girl dresses in a variety of outfits, which would not be classified as objectionable. The member-only area contains more sexualised images of the same girl that are objectionable. The Independent Reference Group (IRG) agreed that, as the purpose of the site was to sexually exploit a child, it should be added to the filter list.”

The IRG misses the point that if you tell the public that you're only going to filter images of children being sexually abused, you can't turn around and filter other content too, even if it is wrong.

The DIA say that the Russian child model website was not considered to be a borderline case.

Because a whole website is being filtered when only part of it contains material we were told was going to be filtered, I asked the following:

“I understand photographs of real life children being sexually abused, CGI and drawings of children being sexually abused, and the Russian child model website are being blocked. Are any URLs being blocked that don't come under that list?”

I received this answer:

“Material being blocked by the filtering system complies with the Code of Practice, which states:

2.1 The scope of the DCEFS will be limited to preventing access to know websites that contain publications that promote or support, or tend to promote or support, the exploitation of children, or young persons, or both, for sexual purposes.

2.2 The DCEFS will focus on preventing access to known websites containing child sexual abuse images.”

Joshua asked:

“Has the filter list ever contained sites with solely written material?”

The DIA replied:

“No, there are no sites on the list that contain solely written material.”

Note that this doesn't really answer the question of whether the list ever contained those sort of URLs.

From a 21 October 2008 email from Peter Pilley at the DIA to Graham Walmsley, Wholesale General Manager at Callplus:

“We have over the last 2 years built as system for the purposes of restricting access to sites that host child sexual abuse materials such as images, movies, stories etc...”

Independent Reference Group members

Here's what Nathan Guy, Minister of Internal Affairs sent someone in regards to Steve O'Brien's membership on the IRG (he's the manager of Censorship Compliance at the DIA).

“I am advised that the InternetNZ submission on the Department of Internal Affairs' draft Code of Practice for the filtering system suggested that membership of the IRG include a wide range of interests, including law enforcement and government. As Mr O'Brien is highly experienced in the enforcement of censorship law and would be working closely with the IRG, he was considered an appropriate appointment to the Group.”

Also, if you're curious, InternetNZ and ISPANZ weren't asked to participate in the IRG.

IRG reviewing sites

The DIA have released conflicting information regarding whether the IRG will actually look at URLs on the list, you know, to make sure they contain material that should be being blocked.

From the Common Questions and Answers page:

“Will the IRG actually review/view the list of sites?
The IRG will be able to inspect the filter list and have access to the inspectors’ reports on any of the sites blocked. They will also be able to check from DIA premises any particular website on that list if they have concerns about it.”

In response to a OIA request that was disclosed with Joshua's OIA, the DIA say:

“The Department will not be subjecting the membership of the IRG to the content of the websites on the filter list. As stated in the Code of Practice, the members of the IRG will be provided with:

• the officers' reports (which identify each website by URL and describe the contents)
• details of all appeal applications and the resulting action taken,
• reports of any technical issues with the filter or connections to any ISP,
• such other information that may lawfully be provided to assist the IRG in fulfilling its function.”

If you're curious, the IRG has reviewed one URL (the Russian child model website). They have the opportunity to review URLs at their meetings, but haven't taken any of those opportunities up (if you're confused and think this conflicts with the first sentence of this paragraph, I am too).

The secret list

Here's what I asked:

“I understand requests for a full copy of the filter list have been previously declined. Could you please send me a list of just the domains from the list of URLs that are blocked, unless the whole domain is blocked (my assumption is if the whole site is blocked it's a site only for child sexual abuse material. I'm looking for the sites that have URLs blocked, but also have non-child sexual abuse content on them).

If [the above] is not possible, the domains of search engines, file sharing locker services, and social networks that have URLs blocked, and a copy of the list with the first domain name part removed, but TLD and the rest of the URL intact.”

Their reply:

“Possession of child sexual abuse material is an offence that carries a maximum penalty of 5 years imprisonment. As the release of part of the URL of the websites being filtered would facilitate a search for such material, the Department is withholding the information requested … in terms of section 6(c) of the Act (where the release of the information is likely to prejudice the maintenance of the law).”

This is a stretch, especially as a reason for refusing to disclose what search engines, file lockers, and social networks have URLs that are blocked.

Note that all other censorship decisions are released. By law, the Classification Office has to publish its decisions, which they do.

The DIA did, however, release a count of the top-level domains (TLDs) of the websites on the filter list as at 12 March 2012:

• .com - 283
• .ru - 93
• .net - 65
• .info - 23
• .biz - 6
• .in - 6
• .us - 5
• .org - 4
• .me - 2
• .tv - 1
• .ir - 1
• .su - 1
• .ws - 1

The prevalence of .com domains on the list is extremely confusing for me. The US Government loves to take them, as well as .net, .org and .tv domains down.

If they can take down a .com domain for copyright infringement (see Megaupload's homepage) why are 353 domains that the US Government should be happy to take care of on the list? Copyright infringement vs. child porn. Priorities people, priorities.

I asked whether abuse reports and take down requests are sent to hosting companies and law enforcement etc. when URLs are added to the filter.

The DIA responded with:

“The Department works with partner agencies in other jurisdictions to get international sites removed.”

The IRG's December 2011 report (pdf) states that:

“Additionally 18% of the users originated from search engines such as google images”.

I asked whether Google was informed of those images:

“We have a very good relationship with Google and they have been made aware of any objectionable links available via their services. The statement in the December 2011 report used Google Images as an example of a type of service. It was not a statement that 18% of users originated from Google Images.”

I asked how long the DIA takes to make Google aware of those objectionable links available through their services. The DIA replied:

“Google is advised of objectionable links available via its services as soon as is practicable.”

Investigator reports

I asked for a copy of all investigator reports held. I received a sample investigator's report (available in the response PDF linked in the introduction).

Here's what the DIA removed:

“Information that would identify the site, including a screen capture of the webpage has been removed in terms of section 6(c) of the Act. Information that would identify officers involved in the operation of the filter have been withheld in terms of section 9(2)(g)(ii) of the Act (to protect officers from improper pressure of harassment).”

The filter and privacy, Google Analytics

I asked for a copy of any contract the DIA has with companies that provide internet services to power the filter, including web and domain hosts for the http://dce.net.nz website (that's the website people are redirected to when the filter blocks a URL).

The DIA say they have “no contracts with providers of internet services that relate to the filtering system.”

I asked what data is collected when someone tries to visit a blacklisted site, including log data collected by the http://dce.net.nz web host:

“The filter only records the service provider name, the resource requested and date and time. No user data is stored.”

In some of the DIA's reports, statistics on device type are included. Device type isn't listed in the Code of Practice as data that's collected. I asked whether other data is collected in the course of the filtering process that isn't listed in the Code of Practice. The DIA said that no other data is collected.

This is from 6.1 of the Code of Practice (pdf):

“During the course of the filtering process the filtering system will log data related to the website requested, the identity of the ISP that the request was directed from, and the requester’s IP address.”

The Code of Practice also says that the requester's IP address is logged and it says that the system will anonymise the IP address. The DIA have previously said that the system retains the IP address for up to 30 days. The DIA clarifies(?):

“When a person requests a webpage that is blocked, the IP address of the requester will be presented to the service so that blocking page can be sent to them. IP addresses are anonymised by the system itself, no record is kept. The filtering system anonymises IP addresses using a tool developed by Netclean. By not logging the data, the system prevents anyone from reviewing source IP. All IP addresses appear as 0.0.0.0.”

The DIA additionally stated that data from the filtering system has never been used in support of any investigation or enforcement activity and that no data from the filtering system has been shared with other departments.

I asked for anything held discussing the implementation of Google Analytics on the http://dce.net.nz website. I also asked what is the data Google Analytics provides is used for and whether any privacy issues were raised regarding the use of Google Analytics:

“Google Analytics is a free service offered by Google that generates statistics about the visitors to a website, in particular the referrers used. Google Analytics is used to confirm other statistics generated from the filter and to provide better reporting to the IRG and public. The Department does not consider the the use of Google Analytics raises any privacy concerns.”

I think it's quite significant that information about New Zealanders is being sent overseas to Google.

I asked whether the DIA has a contract with Google:

“Google Analytics is free software. The terms and conditions for the use of Google Analytics are available at http://www.google.com/analytics/tos.html.”

Curiously, under the privacy section of the Terms and Conditions Google states:

“You must post a privacy policy and that policy must provide notice of your use of a cookie that collects anonymous traffic data.”

http://dce.net.nz doesn't have a privacy policy.

Google's privacy overview for Google Analytics states again that:

“All website owners using Google Analytics are required to have a privacy policy that fully discloses the use of Google Analytics.”

Google also logs whether the visitor has been to the site before on behalf of the DIA. This isn't disclosed in the Code of Practice (pdf). Google Analytics also collects IP addresses:

“Google Analytics collects the IP address of website visitors in order to provide website owners a sense of where in the world their visitors come from. This method is known as IP geolocation.”

The IP addresses are not passed to the website owner (the DIA), but it's unclear whether Google stores them after the geolocation process has taken place.

From the IRG's August 2011 minutes:

“Andrew Bowater asked whether the Censorship Compliance Unit can identify whether a person who is being prosecuted has been blocked by the filtering system. Using the hash value of the filtering system's blocking page, Inspectors of Publications now check seized computed to see if it has been blocked by the filtering system. The Department has yet to come across an offender that has been blocked by the filter.”

I asked the DIA to explain what this meant:

“Every image, photograph, document or movie found on a computer can be run through a hashing process that will generate, using a mathematical algorithm, a unique hash value for that file. A hash value is a set of numbers and letters strung together and once assigned this hash value cannot be altered. If the same image is hashed twice, the hash value will remain consistent; however, if even 1 pixel of an image is altered that new image will be assigned a new hash value.

When the Department seizes a computer or storage device as the result of exercising a search warrant, as part of the forensic examination of that device, the Department is able to look to see whether the offender has been blocked by the filter by looking for the unique hash value generated by objects on the blocking page.

While this information plays no part in the prosecution of an individual, it is useful in understanding the behaviour of persons who access child sexual abuse material and the effectiveness of the filtering system.”

This probably doesn't take into account the fact that some people have visited the http://dce.net.nz website without being redirected there because of trying to access a blocked URL (like me, and if you've clicked on the link, you too).

Correspondence with ISPs

I asked the DIA to send me any correspondence, electronic, written or otherwise, with ISPs regarding them joining or leaving the filter.

Here's what I received:

• An email (14/7/08) from the CEO of TelstraClear to the Minister of Internal Affairs and the Minister's reply (20/08/08);
• Emails between the Department and Callplus (21/10/08 and 30/10/08). Technical information regarding the operation of the filter and information regarding its location has been withheld under section 6(c) of the Act. The telephone numbers of officers have been withheld under section 9(2)(g)(ii) of the Act.
• A letter from the Department to Telecom (29/09/09).
• Emails between the Department and Telecom (1/11/10) regarding a draft press release.
• A letter from the Department to ISPs (list enclosed) explaining the filtering system and inviting them to contact the Department for more information. The telephone numbers of officers have been withheld under section 9(2)(g)(ii) of the Act.

Other correspondence with Telecom was withheld under:

• 9(2)(ba)(i) of the Act (to protect information which is subject to an obligation of confidence where the making available of the information would likely to prejudice the supply of similar information, or information from the same source, and it is in the public interest that such information continue to be supplied);
• 9(2)(j) of the Act (to enable the Department to carry on, without prejudice or disadvantage, negotiation); and
• 9(2)(h) of the Act (to maintain legal professional privilege).

Here's Allan Freeth, TelstraClear CEO to Rick Barker, Minister of Internal Affairs:

“We will add a filter to all web browsing by Clearnet and Paradise customers that stops browsers from accessing known child sex abuse sites around the world. There are more than 7,000 such sites…” “While we believe the Internet is a wonderful source of information and that people have the right to determine what they view based on personal taste, there is nothing positive about content that reflects the suffering of children.”

This is Rick Barker back:

“While participation by ISPs in the filtering programmes will remain on a voluntary basis, I expect that customer demand will mean that most ISPs will join the programme.”

He also requested that TelstraClear keep information on who has been using an IP address at a specific time for longer because ISPs are only storing information for the amount of time they require it (which is exactly what they should be doing). Note that he thinks that ISPs shouldn't place so much importance on what the Privacy Act says:

“The importance of Internet Protocol (IP) address data to DIA investigations has been recently drawn to my attention. The identification of individual computer addresses and the ability to correlate this information with the location of those computers is vital to catch offenders who distribute images of child sexual abuse. I am advised that ISPs consider that, in terms of the Privacy Act 1993, they are required to dispose of information related to IP addresses once this information is no longer necessary for the operation of their businesses. While it is up to each ISP to determine how long they keep this information, I was concerned to learn that some ISPs retain this information for only a very short time. I hope that we can continue to build on the successful partnership between government and business and that TelstraClear will continue to support my Department's investigations by retaining IP address data for a longer period of time.”

Telecom was given as a reason why other ISPs should join the filter. In a 21 October 2008 email to Callplus, the DIA claimed that “Telecom is coming online very soon”. Telecom released a press release saying they were joining the filter on 3 November 2010.

In October 2008 Telecom still had doubts about the filter, including the legality of it.

Here's a portion of a 29 September 2009 from Keith Manch, Deputy Secretary, Regulation and Compliance at the DIA to Dean Schmidt, Telecom Senior Executive Government Relations and Grant Fraser, Telecom Senior Solicitor.

“Telecom's cautious approach to date is understandable. However, as you are aware there is a compelling case that any ISP's participation in the website filtering system is lawful.

This case is based on the argument that redirecting a get request to the Whitebox and then to the Department's server is not an interception. In addition, even if a get request is a communication, and we suggest that it is not, then it is certainly not a private communication, because there can be no reasonable expectation of privacy in respect of a request that is analogous to the address on an envelope. Finally, even if a get request is in fact a private communication, there might be an argument that the ISP is a party to that communication.

Telecom should feel reassured that making out any one of these four points would be enough to ensure that the prohibition in section 216B of the Crimes Act 1961 is not breached.

If Telecom has any residual concern that redirecting a get request into the website filtering system is an interception of a private communication, then we suggest it proactively obtain the express or implied consent of its users, through the use of on-line terms and conditions of use. This would ensure that Telecom is a party to the communication, and that the offence provision in section 216B would not apply.

The Department has considered whether to utilise the provision in the Crimes Act to make an Order in Council exempting an interception device from the provisions of Part 9A. The Department does not intend to do so as we consider this unnecessary in light of the points made above. We do not see the Whitebox software as an interception device, and as a result think it would be inappropriate and confusing to seek an Order in Council premised on it being such a device.

Finally, I note your concern that regardless of the strength of our view that what is occurring is entirely legal, someone may seek to challenge it. While I accept that the potential for challenge to arise cannot be completely discounted, I suggest that this risk is minor in comparison with the benefits of joining the website filtering system. Should a challenge emerge, to the extent the Department is able to assist to overcome those proceedings, we would do so.”

And Telecom did add the following for their terms and conditions for broadband, Xtra, mobile broadband, and mobile:

Department of Internal Affairs Digital Child Exploitation Filtering System
Telecom will intercept communications for the purposes of the Department of Internal Affairs' Digital Child Exploitation Filtering System and in continuing to use Telecom's services you acknowledge and consent to this.

You can also see the final changes to Telecom's press release through emails sent between Telecom and the DIA's PR staff.

Telecom Retail CEO, Alan Gourdie's quote was changed from:

“The abuse and exploitation of children is intolerable and this filter works to block access to the worst-of-the-worst child exploitation websites.”

to

“The abuse and exploitation of children is intolerable and this filter works to block access to known child exploitation websites.”

and

“The system will be applied in coming weeks.”

was added to the bottom of the release.

IRG minutes refer to detailed traffic reports and information about patterns which is given to ISPs.

In the March 2011 IRG meeting minutes:

“Officials noted that more detailed reports on traffic through the filtering system is being distributed to each ISP. ISPs use this data to assist in the management of their systems, including the operation of their internal filtering systems that they offer customers.”

And October 2010:

“Officials noted that the data obtained from the filter can demonstrate patterns of requests for blocked websites that may be of interest to ISPs. This information includes the 50 most blocked sites and the time of day that the filter is most active but cannot identify particular ISPs. The Group agreed that the DIA should draw any such patterns to the attention of ISPs.”

I asked for this information, but received this:

“The information has been withheld under section 9(2)(b)(ii) of the Act (would be likely unreasonably to prejudice the commercial position who is the subject of the information).”

Integrity of the list

I asked whether a URL could be added to the filter list without the approval of three inspectors and without the knowledge of the IRG, and what the limitations are that would prevent that from happening:

“No. The addition of a URL to the filter list requires three inspectors of publications to agree that the website comes within the scope of the filter system. Once a change to the filter list is agreed, only one officer has the ability to edit the filter list. As the task of reviewing the filter list is shared between members of the Censorship Compliance Unit it is unlikely that the same three inspectors will be involved in the review of a website.”

What does the filter achieve?

I asked whether the DIA has any statistics or figures to back up what they say on the Common Questions and Answers page:

“In the long term, if it is made more difficult for persons with a sexual interest in children to access this material, the market will decline and fewer children will be exploited.”

“The Department firmly believes that if the market for child sexual abuse material is reduced, then fewer children will be abused to support that market. The problem is a global one, to which the Department's website filtering system can only make a small contribution. The Department therefore has no statistics or figures to confirm that the filtering system has lead to fewer children being exploited.”

Chief Censor

I asked whether the Chief Censor has been consulted over decisions relating to the filter.

I received this reply:

“Many of the publications blocked by the filter have been the subject of classification and are therefore on the online database of classified material that is accessible on the Office of Film and Literature Classification website.”

ISPs that were asked to participate/sent a letter about the Digital Child Exploitation Filtering System

If you want to know who was asked (pdf), but didn't cave.

• Actrix
• ASC Data
• Airstream Metworks [Networks?]
• Airnet NZ
• BorderNET
• BorgWiFi
• Compass Communications
• Plain Communications
• Cybermedia New Zealand
• Enternet Online
• Evolution Wireless Consultants
• Teldave Communications
• Farmside
• Freenet
• GetRheel
• Go2 Internet
• AGRE Enterprises
• Helix Wireless Ltd
• Internet Hawke's Bay
• ICONZ
• Inspire Net
• KC Internet
• Kinect
• Kiwi Online
• KTSA Internet
• NATCOM
• Netsmart
• Netspeed Data
• NZNET Internet Services
• NZWireless
• Orcon Internet
• PlaNet Internet
• PrimoWireless
• Slingshot
• Snap Internet
• TelstraClear
• thepacifiicnet [thepacificnet?]
• The Packing Shed
• thinair Communications
• Uber Networks
• Vodafone New Zealand
• Web World
• WirelessWeb
• WIZwireless
• Woosh
• WorldNet Services
• Xnet
• Xtreme Networks

• Very few people (only 9%) knew whether their ISP used the government filter. The ISPs using the filter represent more than 90% of the NZ internet market.
• Less than a quarter (23%) wanted the government choosing whether to filter their internet connection.
• Two-thirds want the filter to include other, non-specified, content.

Tech Liberty's Comment

We've always been opposed to the government's internet censorship system but support the right of people to choose filtering for themselves or their families. We're pleased to see that the people of New Zealand agree with us, rejecting the idea of letting the government impose centralised censorship.

Unfortunately we already have such a system. While it is voluntary at the ISP level, their users get no say in the matter and this survey shows that most are unaware that they are covered by it. We also note that with Telecom, Vodafone and 2 Degrees all having implemented the filter there are no major providers of censorship free mobile data in New Zealand, further undermining any voluntary aspect to the current filter.

At the same time it also seems obvious that the internet has a lot of disturbing content that you might want to block other than just child pornography. Therefore it makes sense that someone wanting "cleaner internet" at their home would be looking for a more general purpose filter than the government's one. A number of ISPs do offer such a service (either free or as an add-on) and it seems that they should be promoting this further.

In conclusion, it seems that the survey shows that the current government internet filter is implemented the wrong way for the wrong purpose and by the wrong people.

What inspired you to offer the service?

People will always choose the easiest way to get the latest movies and TV shows and downloading off the internet is it. Unfortunately options are now limited to VPN tunnels not only for p2p but also for using an American IP address to get access to TV sites like www.hulu.com.

Are you getting many customers and what do they want it for?

Yes I am. VPN tunnels have a number of uses that include getting an American IP address to watch tv sites such as hulu.com, encrypt internet traffic when they are using a public WiFi point and of course hiding their real IP address.

How do you feel about the fact that some of your customers will probably be using your service to break NZ law?

No Comment.

What sort of information do you keep about your customers?

We only keep the email address.

What sort of information do you keep about your customers connections? (Such as when they connect, how long they connect for, anything they do through the service.)

We do not log what the user does or transfers over our network but we do log the time of connection and disconnection. We use this data to strategically deploy network resources. We also log the country the user is logging in from, this helps us to detect hijacked accounts and abuse. We do not log IP addresses.

Do you think your business has an obligation under the Telecommunications (Interception Capability) Act to allow the NZ police or other enforcement agencies to monitor traffic?

No, I am not a network operator. A network operator means a person who owns, controls, or operates a public telecommunications network or a person who supplies (whether by wholesale or retail) another person with the capability to provide a telecommunications service. CyberDodge does not provide anyone else with the capability to provide a service and CyberDodge is not a public telecommunications network. Public telecommunications network means a public switched telephone network and a public data network. CyberDodge is not a public switched telephone network nor a public data network. A public data network means a data network used, or intended for use, in whole or in part, by the public and includes, without limitation, the following facilities: Internet access and email access. CyberDodge requires that you have internet and email access already. This law applies to ISPs, which CyberDodge is not.

Do you think your business has an obligation under the Copyright (Infringing File Sharing) Act to store customer IP addresses so that you can pass on notices?

No, I am not a IPAP. IPAP, or Internet protocol address provider, means a person that operates a business that, other than as an incidental feature of its main business activities, offers the transmission, routing, and providing of connections for digital online communications, between or among points specified by a user, of material of the user's choosing and allocates IP addresses to its account holders and charges its account holders for its services and is not primarily operated to cater for transient users. CyberDodge does not offer the transmission nor providing of connections for digital online communications. CyberDodge only routes digital online communications. This law applies to ISPs and CyberDodge is not a ISP.

A thread over on gpforums.co.nz has discussed problems Telecom users have had accessing content delivered by various CDNs (content delivery networks - used by many sites to handle video streaming).

Network traces showed a large amount of packet loss and the path taken by the data looked a bit unusual.

This appears to be the first sign of a site being either adversely affected or actually blocked by the DIA filter. We've also had confirmation of other ISPs (Internet service providers) believed to be using the filter having access blocked.

What we believe is happening

The filter works by creating alternative routes to particular network IP addresses and passing them onto the participating ISPs. Traffic to those IP addresses is then passed to the DIA and checked by the filter to see whether it is going to the blocked site or another site on the same IP address. If it is going to a blocked site, the user is redirected to www.dce.net.nz, or else it allowed through the DIA's ISP and out onto the Internet. (Read more in our Filtering Frequently Asked Questions article.)

Inspection of the traces shows that the traffic is going through an ISP with a relationship with the Department. The address 124.150.165.62 in the traces is from that ISP. The traffic is then going out through a link that the ISP has to Australia.

This ISP's link to the Internet appears to be either under considerable pressure or is simply broken. The level of traffic being dropped by it (as reported by users and our own investigation) is likely to be degrading access significantly to any site hosted - but not actually blocked - by any IP address the DIA is wanting to inspect.

What does this mean?

The site in question hosts anime (animated video from Japan and other countries). While we believe that some anime work has been found objectionable in New Zealand, we cannot find any reference to this site being banned by the Chief Censor.

Even if one video at the site has been blocked by the DIA, this blocking appears to be generally degrading performance to other material on that site or any other site hosted by the same content delivery network.

The Department has repeatedly denied access to the filter list in the expectation that hiding the list will prevent people from accessing it. As this story illustrates, it's not difficult to uncover the filter given the effects it has on an IP address being filtered/intercepted.

We're very interested in hearing from anyone else having difficulties accessing a site where 124.150.165.62 appears in a traceroute to the site. We're particularly interested in legal content being degraded by passing through the DIA's filter.

Please do your job.

Yours, Tech Liberty

We believe that secret censorship is a threat to our democracy. We need to be careful when giving our government the ability to limit what we can see and hear - which is why we require the Chief Censor to publish their decisions. This openness, the ability for anyone to review and challenge, helps prevent abuse of the censorship scheme.

One of our objections to the government's Internet censorship filter was that the Department of Internal Affairs has refused to release the list of censored sites. They say that they'll only censor certain types of material, but how can we know that they're sticking to this without being able to see the list?

The DIA did respond to these concerns by establishing the Independent Reference Group to provide at least some semi-independent oversight of the filter, although they had to be persuaded to let the IRG have access to the list of blocked sites. Then, from the minutes of the IRG's meeting on 15th October 2010:

Members of the Group were invited to identify any website that they wish to review. They declined to do so at this stage.

Now, we quite understand that members of the IRG don't want to look at those sites. But that's not the point - they have a responsibility to ensure that the filter "...is operated with integrity and adheres to the principles set down in the Code of Practice."

This oversight isn't going to work if the IRG don't exercise it. The filter list grew from 153 entries in June to 538 in November - surely it would have made sense to have a look at the list and select some of the additions for a brief review?

Recommendation

We recommend that at each meeting the IRG should randomly select a sample of newly added sites and review the content to ensure that the filter is not being abused. Anything less is neglecting their duty.

It seems that a lot has happened since we did our last update.

Increase in the number of ISPs

The ISPs using the system are now:

• Maxnet
• Watchdog
• TelstraClear
• Airnet
• Xtreme

Telecom are obviously next and Vodafone are also apparently well on the way to implementing it. According to the DIA, "Discussions are continuing with Ihug/Vodafone, Woosh, Orcon and 2degrees. Design changes are being investigated to adapt the system for performance on mobile devices." However public statements from Orcon have said they have no plans to implement the filter.

Even so, this means that most users of the Internet in New Zealand will be using a filtered connection.

The filter list

The number of entries has risen from 153 (as at 29th June 2010) to 538 (as at 2nd November 2010), representing 463 unique domain names.

What is being filtered

From the DIA's Independent Reference Group minutes: "Aware that the inclusion of drawings or computer generated images of child sexual abuse may be considered controversial, officials advised that there are 30 such websites on the filtering list. Nic McCully advised that officials had submitted computer generated images for classification and she considered that only objectionable images were being filtered. It was noted that images of popular television cartoon characters engaged in sexual acts, which are quite common on the internet, would not be added to the filter list."

Further reading

• Tech Liberty's Internet Filtering FAQ
• Tech Liberty's Internet Filtering Technical FAQ
• Which ISPs Will Filter?
• Why We Oppose Internet Filtering
• Stop the Filter campaign website

References

• http://www.dia.govt.nz/diawebsite.nsf/wpg_URL/Services-Censorship-Compliance-Independent-Reference-Group-Meeting-Minutes-15-October-2010?OpenDocument
• IRG 2nd Quarter Briefing Document from the DIA (not online)
• Letters from the DIA (not online)
• http://www.nbr.co.nz/article/telecom-installs-govt-website-blocking-filter-132625

• They completed trials of the system nearly two years ago.
• They have signed up just two small ISPs, Watchdog and Maxnet, although we believe that Maxnet are not using it yet.
• The list of banned webpages has only 153 entries on it (well down from the 7000 they were claiming earlier).

While they've been doing this:

• Internet NZ has announced their opposition to the filter on technical and practical grounds.
• Six ISPs have said they definitely won't use the filter, another couple have said they have no plans to implement it, and only three have stated an intention to sign up alongside the current two.
• Political pressure has forced the Australian Labor government to delay implementing their filter, and the opposing Coalition has said they'll scrap the system if elected.
• The US government has opposed internet censorship and stated their commitment to developing tools that allow people to circumvent it.

And all the normal reasons against the DIA's proposed censorship scheme remain:

• The filtering system only works against unencrypted websites and doesn't stop the main ways used to distribute objectionable images - torrenting, email, chat. Only 8.5% of the traffic going through the filter can be checked.
• It's easy for motivated people to work around the filter.
• Secret censorship (the list of banned sites is kept secret) is offensive in an open and democratic society.
• Better filtering solutions that cover a wider range of objectionable material are available for those who want them for their family or business.

Time to stop?

The filtering system doesn't seem to be getting anywhere and isn't going to work if it ever does - surely it's time to just close the project down. The money saved could be much better spent funding the DIA's efforts at infiltrating the groups that trade in objectionable images and shutting them down.

Which ISPs?

We've updated our list of which internet service providers will or won't be implementing the DIA's filter.

New additions to the not filtering camp:

• Unleash - "we will not be implementing this filter and we have no plan to do so in the future."
• Iconz - "ICONZ are not currently signed up to the filter, nor do we have any intention of doing so in the future."

And an ISP that will be filtering:

• Xtreme Networks - "It's a no-brainer for us, it's free."

That gives us three ISPs that will be filtering, six that won't, and two that are undecided. We're still waiting for hear from Woosh, FX, Compass, Inspire, and Actrix.

More material from the DIA

The Department of Internal Affairs has posted a couple of new documents on their website:

1. They have done their own FAQ in opposition to Tech Liberty's FAQ.
2. There is also an Information Pack (pdf).

Filtering Delayed in Australia

The fight against the Australian filter system continues. Significantly worse than the New Zealand system, the Australian filter will be compulsory, covers a wide range of material and there is no oversight of the list of filtered sites.

In a sign that protests are working, it was recently announced that passing of the filtering law will be delayed until after the election. Further links and comment from Electronic Freedom Australia.

See OpenInternet.com.au for more information about the Australian anti-filtering campaign.

The obvious question has to be, why was Tech Liberty announcing something that the Department of Internal Affairs had done? Where was their announcement that the filter had gone live on the 1st of February? Don't civil servants have a duty to communicate to the people that they serve?

Sadly this reticence with information has been typical of the Department of Internal Affairs in relation to the implementation of the Internet filter.

Deleting Public Records

Last year we used the Official Information Act to ask for copies of the reports that the inspectors has used to justify banning the websites on the list. The DIA refused. After we appealed this refusal to the Ombudsman, the DIA then said that those records had been deleted and therefore it was impossible for them to give them to us anyway. The Department has an obligation under the Public Records Act to keep such information.

We complained to the Chief Archivist, who investigated and confirmed that the DIA had deleted public records without permission. He told us that the DIA has promised to do better in the future, but naturally this didn't help us access the missing records.

The Secret Go-Live Date

Why has the DIA been so secretive about the filter going into operation? Here's two examples where we believe that they have failed to be open and honest about what they are doing, even in response to direct questions.

We wrote to the DIA and asked them, again, when the filter was going to go live. They wrote back on January the 20th and said that as they were about to make an announcement, the Official Information Act gave them grounds to refuse our request. This was 11 days before Watchdog was the first ISP to start using the filter. It's now the 16th of March, nearly two months later, and there's still no announcement from the DIA.

Secondly, on February the 15th we rang Keith Manch, Deputy Secretary of Internal Affairs, and directly asked him when the filter was going live. Keith is responsible for Regulation and Compliance and has been heavily involved in the implementation of the filter. Did he admit that the filter had gone into operation two weeks earlier? No, he carefully took note of our questions and then wrote in a follow-up email that as we had already asked those questions by letter he wouldn't answer. We finally got our answer on March the 8th, admitting that the system had gone live on February the 1st.

Open and Democratic Government

Tech Liberty is at the intersection of technology and civil liberties. We are strong supporters of the right to self-rule as expressed through democratic government. An important element of democratic government is the principle that government must be open and accountable, as without this governments tend to become corrupt and self-serving.

New Zealand recognises this and the Official Information Act and Public Records Acts are some of the ways we use to ensure that our government remains open and accountable. However, the law isn't enough on its own, it also requires a commitment from government departments to honour the spirit of the law and not try to use or misuse the letter of it to conceal information.

We don't believe that the Department of Internal Affairs has been living up to this standard when it comes to the issue of internet filtering.