one of the ways is to use LSB (least significant bit) which is the simplest method on the frames, you can read the stream frame by frame and include the information in these frames. you can create a random generator to distribute the data randomly on the pixels in each frame..
There are no top algorithms. Straight-forward LSB is certainly not secure. You have to make sure that whatever distortion you introduce is introduced consistently in every frame throughout a scene, as a minimum, to bar attacks based on inter-frame correlation.
The key to steganographic security is low pay-load. Andrew Ker's square root law for steganography tells you that. You won't ever get asymptotically non-zero information rate in steganography. If the pay-load is small enough, you can do just about anything; it isn't statistically significant so the attacker cannot make any inference.
What I would recommend is to use a hashing of the steganogram as the decoder, and let the encoder simply choose a cover with the correct hash as the steganogram. That should work OK for 16-32 bit messages or so, and I reckon it is theoretically secure too.
I do not think enough work is done on Audio or video steganography . but you may go through transformation techniques and framing and the some technique like LSB or or any other embedding technique.
It is certainly true that little work has been done on audio and video steganography, but some of the results obtained in image steganography are generic, and so are some from video watermarking. Steganography can only be kept a secret with small scale use.
But then, the answer to the original question probably depends on what the purpose of the study is.
How do you define, operationalise and measure «high security»? The question is serious; security isn't well-defined, and a definition which makes sense for an algorithm out of context, is hardly relevant in a practical application.
If security is primary, capacity secondary, and practicality is of no particular concern, then you should look into the theoretical and abstract work which has been done (oblivious to host medium) in cryptology (e.g. Backes and Cachin).
If you really want to make the requirement on /any steganalysis algorithm/, then you are into provable security, and you have to search in cryptology and information theory. Backes and Cachin is a good place to start.