Flash media server is a media server which streams flash video. Basic difference between streaming and progressive (youtube) is that incase of streaming, the video is never stored on client's disk. As soon as it is displayed, video data is discarded. This makes it difficult for a user to store the video on it's disk to view it offline. Then how are tools like 'Replay Media Catcher' etc able to do this so easily?
To start with, RTMP (Real Time Media Protocol) has been completely reverse engineered. Being a text based protocol, doing this wasn't very difficult for experienced hackers. A lot of open source implementations of RTMP are available, which allow a user to write C++ or Java applications to connect to FMS and subscribe to streams. Once you know the name of the FMS server and the stream name, you can use these applications to subscribe to that stream and dump audio/video data into a FLV.
Is it really that simple? The answer is yes, only if the server side application developer forgot to add any kind of authentication mechanism. This is 2nd reason why these so called 'media catchers' are able to work without a glich. Wihtout any kind of authentication, 'Replay Media Catcher' is easily able to open a new NetConnection with FMS, which is gracefully accepted by the server since the application programmer forgot to authenticate the client. That would be a silly thing to do especially if one is planning to commercialy host video content.
Now to the ultimate question. How does 'Replay' figure out the name of the application to connect to and the stream name to subscribe when a .swf file(in a separate process) is trying to subscribe to a stream. My initial guess is that it is trying to hack into flash player's memory to get this information. It would be very difficult to extract this information by looking at the packets being sent over the network and virtualy impossible if RTMPE is used by the .swf. There is no way it can decrypt data being exchanged between client and the server incase of RTMPE.
So what should one do to protect their content? Simple, authenticate the clients which try to connect to your application. Don't just blindly accept connections. The auth mechanism need not be complex. Very simple mechanisms like exchanging random numbers during the connection process can do the trick. Also use RTMPE to avoid any kind of packet sniffing. RTMPE is a lot faster than SSL and doesnt require certificates.
The following link provides a nice introduction of various sceurity features offered by FMS:
Finally i'd like to say is that content is precious. Don't let these tools hack into your hard work.