A European parliament study evaluating `Feasibility of censoring and jamming pronography and racism in informatics': "SUBSCRIBER PROTECTION: In this scenario [...] it would be necessary for access providers to use firewall and PICS technology (in some instances implemented with in proxy applications) to implement a management regime consistent with provisions for protection of society as in other media, ie media parity. Implementation on this basis would be necessary to ensure subscribers could not tamper with management settings. The impact of implementing the necessary technology on the operating economics of access providers appears to be small. That technologies of these types have already been employed by access provider organisations [...] could potentially be considered to confirm this." From: kris@koehntopp.de (=?ISO-8859-1?Q?Kristian_K=F6hntopp?=) Newsgroups: de.soc.zensur Europarl: STOA Homepage http://www.europarl.eu.int/dg4/stoa/en/default.htm Europarl: STOA Publications http://www.europarl.eu.int/dg4/stoa/en/publi/publi.htm Europarl: STOA Workplan 1996 http://www.europarl.eu.int/dg4/stoa/en/wp96.htm A. Information and telecommunications policy options A.LP.2 INTERNET AND MULTIMEDIA IN THE INFORMATION SOCIETY (In cooperation with the specific research programme of DGIV: REX/Internal Market/New technology/STOA) Sub-projects: (iv) Technical feasibility of jamming or censoring pornography and racism in informatics - Civil liberties ------------------------------------------------------------------------------ STOA European Parliament * Batiment Schuman Bureau 4/82 L-2929 Luxembourg Luxembourg Phone: +352 4300-2511 Fax: +352 4300-2418 * Batiment Eastman Bureau 113 B-1047 Brussels Belgium Phone: +322 284-3747 Fax: +322 284-9059 Email the STOA Team in Luxembourg at _STOA@europarl.eu.int_ Email the STOA Team in Brussels at _ksercu@europarl.eu.int_ ------------------------------------------------------------------------------ This document is a working Document for the 'STOA Panel'. It is not an official publication of the STOA. This document does not necessarily represent the views of the European Parliament. Feasibility of censoring and jamming pronography and racism in informatics ***DRAFT*** Final Report Luxembourg, May 1997, PE number: 166 658 <del del del> 7 Conclusions 7.1 Technical Feasibility 7.1.1 The central question posed to this study is whether it is technically feasible to use the content managment technology to censor and jam pornography and racism on the Internet. 7.1.2 It is concluded that available technology does no and is not likely to provide any satisfactory means for automatically classifying such material. Such judgements will therefore have to be made by humans. 7.1.3 Once such judgements are made, however, technology provides for these judgements to be distributed and used by content managment systems to make the following content managment actions feasible (in principle) to a high degree: - subscriber choice, in that individuals can choose the extent to which they permit or exclude such material on their own behalf or on behalf of others for whom they have responsibility; - subscriber protection, in that precautions against unintentional or unsolicited exposure to such material can be implemented as defined by statute or other form of code or regulation; - prevention of Internet abuse, in that it is possible to detect the exchange of such material within closed groups of offenders to a very substantial extent; - detection of offenders, in that offenders can be identified when offences have been detected or notified involving sites and services where the user has explicitly authenticated identity; - flexible and multi-cultural implementation of content management, in that content management policies can be defined: - to meet the requirements of very small subscriber groups, eg. selected subscribers or a selected group of terminals; - differentially for different types of Internet service, eg Web access and e-mail; - to meet the needs of one culture with respect to the material originating from another without the cooperation of the originator; 7.1.4 The principal technical limitations in these capabilities are: - automated content classification systems can not, because of the complexity of legal definitions, identify Internet content to be racist, pornographic, illegal or harmful - hence this classification must be made by human judgement; - the potential flexibility and coverage offered by the technology can not be achieved in full, at present, because - currently implementations of the key technology (PICS) in Internet applications is incomplete and hence the ratings information can not yet be used to manage all types of Internet content explicitly; - ratings databases are currently incomplete in terms of the types of service they consider and the full range of Internet services available; - it can not be absolutely guaranteed that Internet anonymity services can not be used to frustrate detection of offenders in certain circumstances. 7.1.5 The principal practical issues involved in implementing these technologies are: - defining rating systems which are sufficiently explicit regarding content managment requirements to allow automatic content management whilst still being: - practical to apply to large volumes of information; - meaningful outside the culture in which they are originally applied; - funding the large effort required to categorise the huge and growing volumes of Internet content. 7.2 Implications of policy scenarios 7.2.1 The study investigated the following policy scenarios: Subscriber choice 7.2.2 In this scenario, essentially the status quo, subscribers would either coose to exercise explicit controls over their Internet experience, eg through the use of parental control software, or not. Potentially, subscribers could elect to implement management regimes consistent with privision for protection of society as in other media, ie to provide media parity, but this could not be enforced. 7.2.3 Currently, the technical capability of these products is increasing as vendors invest in the development of new facilities. In the longer term, however, it is not clear that the current subscriber uptake of such tools (on a purely discretionary basis) will sustain their development or the maintenance of databases upon which they depend on a long term basis. 7.2.4 The principal social objection to this policy scenario is that it does not in any way act to reduce abuse either on the Internet or in the 'real' world. The principal legal issue which arises is that it does not provide for media parity. Subscriber protection 7.2.5 In this scenario, subscribers would continue to have the option to exercise management controls accordings to their preferences - potentially including screening unsolicited communications from 'strangers' or anonymous sites. In addition, however, it would be necessary for access providers to use firewall and PICS technology (in some instances implemented with in proxy applications) to implement a management regime consistent with provisions for protection of society as in other media, ie media parity. Implementation on this basis would be necessary to ensure subscribers could not tamper with management settings. 7.2.6 The impact of implementing the necessary technology on the operating economics of access providers appears to be small. That technologies of these types have already been employed by access provider organisations (without discernible impact on end-user charges) could potentially be considered to confirm this. 7.2.7 The principal legal issues which arise in this scenario are: - the liabilities of those who apply labels to content provided by others; - the liabilities of those who are charged with implementing content management on the basis of labels applied by others; - the potential for content labelling providing a legal obstacle to free trade. 7.2.8 These legal issues potentially have substantial social impact in that: - if liabilities are potentially large, those who accept them, either through compiling ratings databases or providing Internet access service, are likely to be very conservative with respect to the material with which they will deal; - international e-commerce in certain services may be compromised by nations exploiting differences in ratings procedures as a pretext for blocking material originating from other countries. Preventing Internet abuse 7.2.9 It is concluded that only implementation of the most intrusive surveillance technology could be exspected to impact heavily upon abuse of the Internet on committed offenders (because of their ready access to strong encryption and covert communication technologies). To enable this policy scenario, it would be necessary for access providers to constrain the equipment used by subscribers to access the Internet to be such that: - no content could be removed; - all content and processing could be subject to automatic and manual surveillance. 7.2.10 The technical and economic feasibility of the necessary technology is evidenced by the emergence, with purely commercial motivations, of technical systems which privde the basic facilities required to implement the necessary measures. It is notes that this coincidental emergence of such systems may raise siginificant privacy issues outside of the scope of this study. The economics of manual surveillance, however, can not be assessed. 7.2.11 This technical approach is, however, completely non-specific in the terms of the subject of censorship which could be implemented and in the scope of (potentially private) material to which surveillance access would be required. It therefore seems highly unlikely that, when the commercial implications, legal obstacles and social objections are considered, application of this technology in this way can be considered a proportionate general response. 7.2.12 Finally, it seems highly probable that the principal of impact of implementation of these measures, if they were to be implemented generally and were found to be effective, would be to displace abuse to other communications means. Detecting offenders 7.2.13 If access providers and service providers maintain proper subscriber authentication systems and system management environments, offending actions at the site to which the subscriber authenticates can be traced to the offender. Detection of offenders committing offences 'at a distance' (ie a subscriber in a country committing an offence in another), however, would require: - access and service providers to take steps to ensure subscriber identity information is carried in their transmissions; - monitoring of services provided on the Internet to prevent subscribers misusing anonymity services to remove such information. 7.2.14 Whilst a variety of means exist to attach identity information, detection of anonymity services is not straightforward. However, monitoring such services would be exspected to limit the number available for overt, eg to transmit offensive material to innocent parties, offensive use. 7.2.15 Legally, handling offences detected through voluntary authentication would be relatively straightforward as the offending subscriber would be resident in the country where the offence was committed. However, offences can only be detected through forced identification, ie those committed 'at a distance', would currently be subject to extradition arrangements and it is not clear that new opportunities for prosecution would arise in practice. Meanwhile, it may be that the social implications of forced and general identification of Internet subscribers would leave subscribers open to a variety of forms of intrusion on privacy and discrimination. 7.3. Options for action at Member State level 7.3.1 It is concluded that, in technical terms, Member States could implement combinations of the technical measures identified in this report according to their own perceptions of harm and proportionality of response in accordance with principles of media parity. This could, technically, be achieved without the cooperation of any external party and provide for different levels of content management according the requirements/vulnerabilities of individual subscriber groups. 7.3.2 The principal potential implications for a member state adopting measures at variance with international norms would be: - obtacles to free trade, including possible infringement of Single Market requirements; - reduced availability of Internet content; - having to bear the costs of maintaining classifications of Internet content alone. 7.4 Options for action at European level 7.4.1 It is concluded that cooperation at the European level could mitigate the implications of uni-lateral action by Member States if action were taken to establish a common content management system. This would comprise: - technical standards in the definition of rating systems to minimise costs of researching Internet content and spread them over the largest possible population; - regulations for harmonisation in the application of agreed standards to ensure that their use did not lead to artifical obstacles in trade; - technical standards in the implementation of content selection mechanisms in Internet applications to ensure that ratings information can be used in a consistent and appropriate way.