Internet Watch Foundation

 

Author – Tony Fagelman

Input: Howard Lamb, Peter Robbins, FC and Board

 

At meetings of the Board and Funding Council in July 2004, a document was presented that outlined the issues and opportunities surrounding the interest in the Child Abuse Images database, with a subsequent addendum that provided more information. Since details of BT’s Cleanfeed project entered the public domain there has been intense national and international interest in our database

 

This document draws on the original documents and brings together various views of contributors into some specific policy recommendations for adoption.

 

In the context of this paper the IWFs primary aim is to minimise the availability of potentially illegal child abuse images on the Internet hosted anywhere in the world.

 

 

Due to inconsistencies that exist around the world in having child abuse content ‘taken down’ the IWF database of child abuse URLs may help to reduce the possibility of internet consumers gaining access to potentially illegal child abuse images whether inadvertently or through actively seeking out the content, by providing data for filtering options to as many organisations as possible.

 

At their meeting on the 18th July 2002 and following legal advice, IWF Board decided to release the child abuse URL section of the IWF database to subscribing companies to use the data to block access to potentially illegal child abuse content.

Termed the Child Abuse Images URL service or CAI URL service, this is a dynamic system, up-dated daily as new URL’s are added. For inclusion on the list, the URL must contain images of child abuse that are deemed potentially illegal according to UK law, including the URLs of websites which advertise indecent images of children. If the potentially illegal images are displayed on the home page then the entire site or top level URL is included, but if they are only found in specific locations within that site, then only those URL’s containing such content are recorded and added.

 

The CAI database holds every confirmed child abuse URL that has been reported to the IWF since inception. This data has always been passed to law enforcement agencies and to European and International Hotlines where they exist and where potentially illegal child abuse images are apparently hosted.

 

The IWF employs analysts, working full-time viewing, classifying and tracing images. The analysts undergo a comprehensive in-house training programme as well as attending training days with police experts.

 

The IWF uses the image classifying system as stipulated by the Sentencing Advisory Panel. These guidelines list five levels of abuse from Level 1, i.e. children under the age of 18 depicted in erotic poses through to Level 5, i.e. children engaged in sadism and bestiality. At least two analysts must confirm that the image is Level 1 before it is added to the database. All reports include the name of both the analysts involved in such a classification. Quality control mechanisms are in place to ensure consistency of decision making.

 

Once a request to receive information stored in the database is received, the IWF set up a secure unique access point through its web-site. Each client is allocated their own specific URL login page with unique password. The system authenticates the client then opens a secure link to generate a request to the CAI database. This in turn passes the data to the web-site and then to the client. At no time does a client have direct access to the database.

 

Data can be transferred to the client either manually, i.e. the client requests a full six month download, or an up-date, since their last request. The system acts like an anti-virus service, the client can request regular updates or they can refresh their entire list. If the client has an automatic update arrangement the system knows what to deliver when a request is made.

 

The data is downloaded using industry standard 128bit encryption using SSL and HTPPS, similar to the method personal online banking services are managed.

 

The IWF adds approximately eighty new URL’s to the list weekly. The system also has a built in reminder system, so that random URL’s are tested after three months inclusion to check their status.

 

The IWF also will also maintain a list of “removed” URL’s. These are URL’s that would be considered as containing potentially illegal data, but might have become “legal”. It should be noted, that there are no URL’s on this list.

 

Since November 2003, the IWF has provided access to the database to member companies on an individual request basis. Each member signs a detailed legally drafted agreement that covers the terms of use, the IP, the liabilities and other standard contractual clauses in preserving the integrity of the data.

 

The IWF currently delivers the data via a secure encrypted method. However, once the data has arrived with the member it is in a raw format and is readable by anyone who has access to it. The agreement requires the member to limit access to that data and if asked, provide the IWF with a list of all personnel who have access to that raw data, and that those personnel must be kept to a minimum. Additionally, the agreement recommends that an automated service be setup that requests the data and passes it directly into the filtering service the member is using.

 

The IWF has considered a number of methods for delivery of the data and subsequent capture by the member, but because of the diverse filtering options that are available, it was decided to provide the data in its simplest form, i.e. .csv file format. The IWF could adopt a policy whereby a more secure method is agreed and that all licensees had to adhere to that method.

 

Any person or organisation that feels that a URL has been included on the CAI list incorrectly, can appeal. If, on receipt of a complaint, the relevant URL is subsequently found not to contain potentially illegal images, then the URL will be removed from the database. If, however, in the expert opinion of the management at the IWF, the content is still potentially illegal under UK law, then the URL will remain on the list. The complainant can then make further representations in which case the matter will be referred to the NCS POLIT (National Crime Squads Paedophile Online Investigation Team) who will further review the URL concerned and make their judgement. The police judgement is final.

 

three issues have emerged:-

 

1)                 use of the CAI data by non-member organisations

2)                 use of the CAI data by members who wish to commercialise it

3)                 members with an international presence who wish to offer the list to their clients/users

 

 

 

 

 

 

 

 

For international consumption it is possible to provide segregated lists that exclude “level 1” images. Such segregation will then render the list compatible with many non UK territories, however, differences in law in different territories means that the IWF cannot be responsible for images that are illegal under UK law, not being deemed illegal under a specific countries law, where the list may be being used as a filter. The member must consider the use of the list both under the countries laws and under its acceptable use policies.

 

The IWF is currently considering two potential options to improve the security of the CAI list.

 

1)                 The use of cryptography to maintain the data in a form that is unreadable from when it is requested from the IWF through to when it is placed in the filtering solution.

2)                 The use of Digital Rights Management as a method of ensuring that only the specific people or machines can access the data and then limit the actual use of that data. i.e. only allow the file to be opened once.

 

Both of these require more investigation, but at the moment, the crypto solutions is looking increasingly unlikely, without enforcing the need for a specific single solution that all licensees must adopt, which may be prohibitive to some due to increased costs.

 

A further option could be for the IWF to provide a separate Server that contains the unique software for each specific service that enables the CAI list to be encrypted in those services. The licensee would then query the separate server as required to obtain their own unique code. Such a service will require further investigation.

 

Tony Fagelman

General Manager