Safer Internet study: significant shortcomings in today’s content filters and parental control technologies

The test lab of the German Kommission für Jugendmedienschutz (KJM) has evaluated content filtering products used for protecting children in the Internet. The final statement is very clear and to the point: The filtering performance, i.e. the accuracy and efficiency, is too low in general and in particular for the blocking of objectionable content depicting violence, racism or other topics not suited for minors. The filtering systems tested by the KJM were also prone to an inadequate degree of overblocking, i.e. the blocking of “innocent” and harmless content.

Internet content rating systems

The negative review of the KJM should be especially hard for the filtering system of the Internet Content Rating Association (ICRA), which has been praised in the past as the best possible content filtering system because it relies on voluntary rating of Internet content (think: special digital labels included in a web page) by the creators and authors themselves. Though the idea of putting the control of the description of a web page into the hands of the owner instead of giving it to third parties (in particular, automated filtering algorithms, Internet security service providers, or governmental agencies) is very interesting and aims at preventing censorship concerns, it has several design flaws and shortcomings in practice, the most obvious one being the question “Why should criminals include digital labels into their (illegal) Internet content?” For example, a hot topic on last year’s Safer Internet Forum was the fight against child pornography in the Internet. Now, how should a system such as ICRA help against that? As if a child molestor would care for creating such digital content rating labels in web pages depicting pictures or videos of abused children…in order to protect children and other users from viewing said pages.

One of my previous studies has shown that the usage of content rating systems such as ICRA is marginal in practice, and that parents would be advised not to rely on rating-dependent content filters to protect their children, at least not without another tier of security. In our random sample of more than 150,000 web pages, only 0.3% were correctly and fully labeled (the percentage “increases” to 0.7% if we just look at any rating label at all, even erroneous ones).

New ways and alternatives

Again, this is not to say that the approach and general idea of ICRA and other rating systems focusing on the content creators and authors are per se inadequate – they are in fact a good first step. But they are at the moment far away from being a viable solution in practice.

In my paper Design and Anatomy of a Social Web Filtering Service, I propose a new, alternative approach to web filtering – you might want to call it “Web (Filtering) 2.0″. In this paper, I describe the design and anatomy of an open architecture for creating a social filtering service and show how it can be implemented. The system is designed to overcome the deficiencies of today’s content rating and filtering systems by empowering and involving the recipients and not the creators of Internet content, i.e. the end users, and to actively support user collaboration and cooperation by using techniques and methodologies with a low cost of participation combined with ease of use. Of course, there is still some work to be done, but it’s already a promising start.

Related links

• SIP-BENCH: part of the European Commission’s Safer Internet Programme; evaluates and studies the usefulness of tools that can be installed at home or at school and that filter bad content