ABSTRACT

Pay TV has evolved from a walled garden, set-top box, model to include online services. Although there are numerous operator and consumer benefits that result from this shift, it also opens up tremendous piracy threats that are a nightmare to control.

The sheer volume of information being shared in a more open environment, means that manpower alone is insufficient to process and detect threats effectively. As Artificial Intelligence (AI) technology develops in the media space, its application in security must focus on more than closing gaps and locking down assets.

Security threats must be spotted and managed faster and more efficiently, before a security instance even occurs.  In this paper, we explain how to leverage AI  to fight piracy by using content monitoring solutions that search and identify pirated content on the internet.

At the core of this technology is an AIpowered computer vision system that identifies the original source of distributed content based on the visual information present in the image (e.g. a broadcaster logo). We cover practical issues around building such a system, including its workflow, training and performance. 

INTRODUCTION

AI is becoming entrenched in our day-to-day lives. Due to rapid innovation over the last few years, AI is now capable of simulating a range of human brain functions, including pattern recognition. AI is disrupting a variety of industries, including the pay media industry, where it is being considered for implementation of complex security and anti-piracy schemes.

In this paper, we present how AI can assist in fighting piracy. It is widely believed that the most significant threat to the content production and distribution businesses today is the illegal redistribution of content. Pirates recompress content that has been decrypted using a legitimate subscription and stream the content either online or via various streaming IPTV devices.

The business model for these types of piracy includes advertising, hardware sales and subscriptions. A typical example of such illegal activity is near real-time redistribution of live sports events. Such events have high value when they are live. Pirates publish links on social media and aggregation sites to illegal content streams. They re-broadcast the original content with limited modifications. Redistribution is done either via web streaming, with or without using a Content Delivery Network (CDN), or via Peer-to-Peer applications such as Sopcast or Acestream.

This form of piracy has become widespread and it is hard to analyse which streams of content are decoys and which are actual pirated streams. One of the problems with the metadata that is provided along with the pirated content (if any), is that it is unstructured and inconsistent.

This means that it is impossible to determine the actual content in a re-broadcast video stream without analysing the stream. This is further complicated when multiple television channels are transmitting the same sports event. Even if the event has been identified as being a specific football match, the logo of the broadcaster is the only indication of the original source of the pirated content.

Download the full technical paper below