FILTERING GAMBAR DAN VIDEO PORNO PADA JARINGAN

Utama, Airlangga Cahya and Pramadihanto, Dadet FILTERING GAMBAR DAN VIDEO PORNO PADA JARINGAN. EEPIS Final Project.

Preview

PDF - Published Version Download (886Kb) | Preview

Abstract

As a source of information, the internet contributes greatly to modern society. Unfortunately, not all information it contains is suitable for everyone. Take for instance, materials containing information about war, violence, drugs, and pornography, themes which are not very suitable for children. Various filtering techniques are used to separate these unsuitable information. Among these techniques, are text based filtering and domain based filtering. Text based filtering works by searching for inappropriate words and calculating its weight. Domain based filtering uses an extensive list of domains it considers as unfavorable. Both of these approaches have several disadvantages. Text based filtering is considered least effective because most harmful contents are image or video. Domain based filtering requires huge resources to check every possible domain name for harmful content. Furthermore, domain names and its contents are frequently updated, domains that were once considered neutral may become harmful anytime in the future. This research uses pornography image detection technique. With this approach, filtering is applied on various data that pass through the network. A computer that stands between the user and internet is used to capture and process this data. Using this technique, we are able to detect about 60% of pornography image positively. It's false positive rate reaches to about 10%. The detection process used in this paper requires a high amount of computation resources. An image that consist of 200.000 pixels takes more than 3 seconds to process. An extensive research at performance area will be able to overcome this limitation.

Actions (login required)