OK, I need to clarify :

The point is the submitter can add information in the publication, let's say by steganography, this information allowing to open a communication channel. Then  the reviewer would use this channel to give proof of being a reviewer to the submitter, without disclosing identity of course.

This is possible I think, by using for instance a hashblock of a part of the text as a proof of ownership. Transmitting the hashblock together with a BTC address through the communication channel.

This left the question of adding a communication channel to the publication that would be costly to detect by automatic means. This do not seems impossible to me for reasons i can expose.. 

I agree with Peter:

You have got the wrong model for how peer reviewing of scientific papers currently works.

1. The paper is not assigned to a reviewer by chance, but by

a. who the editor knows who is a best fit to review the paper,

b. who actually has the unpaid time to review the paper

2. The reviewer is never in any channel of communication with the writer. All communication is via the editor to ensure anonymity.

3. Because of 2 above, the writer cannot contact the reviewer to offer any sort of unethical rewards.

4. About half the editors adopt a double-blind system, which means that the reviewer does not know who the writer is, stopping all possible unethical behaviour.

5. There are always more papers than there are willing volunteer referees, so the idea of “bidding” for papers is flawed.

6. The acceptance of a paper depends only on the merit of the paper, and there is never any unethical passing of money to make the paper more “popular”.

There are however three holes in the system, which may be harmful:

A. If it is a small field, then every expert knows everybody else.

B. If the reviewer is not an expert in the field, then the reviewer may make a wrong recommendation and science will not be advanced.

C. Reviewers are unpaid, they only get secret fame and a secret preview of work in the field as rewards.

If an unethical referee exists, then this crook could conceivably identify the writer through some scholarly and academic searches, E-mail the writer, and demand a “ransom” paid via a cheque or an electron stream to let the paper be favourably reviewed. However, this does not guarantee publication, as the other (anonymous) referees would have to also recommend publication of the paper. Furthermore, since almost all researchers are ethical, whistle-blowing is highly likely.

You might do better by modelling the reader reviewing of Amazon books, where there has been a high-profile fraud case in the last month.

A channel of communication between reviewer and writer is unnecessary and potentially dangerous. All that is required is a signal “Accept / Revise / Reject” to be transmitted to the editor, the editor to collate similar signals, and to convey the majority decision / management decision to the writer.

It seems to me that the assumption of impossibility of hidden communication between reviewer and submitter is the main point of the question, actually a full communication channel is not necessary and dangerous as Ian underline.

The process can be simplified to one information exchanges. The one from the reviewer to the submitter, giving proof of owning the paper and an anonymous bitcoin address. The proof and the bitcoin address are supposed to give no information on the emitter. 

To be possible, the information exchange needs the submitter to hide in his paper a way to receive this information. He obviously do not want the journal to detect this hidden information. This may not be possible with certitude, but it may be possible to make the automatic fraud detection very costly.

To make it costly can be done by requiring actual understanding of the text to decipher the posting address. What about : "The address is the result of a given hashcode method applied to the sentence in part 2 that contradict with a sentence in introduction"? 

The deciphering function should change often enough I guess.

I did not meant to offend anyone, this question was quite genuine. @Ian I believe most researcher have indeed enough integrity or self respect to not even give a look at a bribery attempt. And I am not trying to find partner in crimes, it would be a terrible place for that anyway.

Working in academic for big money do not sound like a smart idea either, I acknowledge. I guess there are lot more way in politics.

Although the question seemed interesting to me, and I was not able to find something related to structural vulnerabilities in blind peer review systems. 

The problem of such an attack is summed up :

"If you do want a way to organise a hidden channel, fine, but (a) it's way not as hard as that, and (b) the punishment for being caught is way too high to risk, so forget it."

So fair enough, I owe you a bottle of whisky I guess! 

Not to be consider as a attempt of corruption, the whisky will have to be terrible though. :)