Reviewers are humans and might be affected by cognitive biases when information overload comes into play. In fact, no amount of scientific training will completely mask the human impulses to partisanship. And the consequence is that authors may receive incorrect editorial decisions in their submissions to peer-reviewed journals. For instance, the journal editor issues a substantial revision when in fact a moderate one would suffice. This would be over-revision in peer review. In this situation, there exists a fraud cost if the journal editor tries to request the author to make a substantial revision when in fact a moderate one would be sufficient. Thus, in this article, we identify a set of conditions under which the peer review process involves equilibrium fraud and over-revision. An equilibrium in peer review is efficient if the first peer-reviewed journal to which the author submits their research paper makes a truthful editorial decision, which the author accepts. When the fraud cost is sufficiently high, there exists an efficient equilibrium. Otherwise, when the fraud cost cannot sustain an efficient equilibrium, it may arise a specialization equilibrium in which the author first submits the manuscript to a top journal which makes a truthful editorial decision. This specialization equilibrium may explain why academic journals with higher quality standards more often attract authors who write articles of higher quality. Finally, when the fraud cost is not too large, we show that a new type of equilibrium emerges in our model, equilibria involving costly fraud, in which the first peer-reviewed journal to which the research paper is submitted always requests substantial revisions. If the review time and the probability of very serious concerns from reviewers were large, the author would prefer to send the research paper to one single peer-reviewed journal even if that would involve over-revision. In the fraud equilibrium, the author's revision cost is high and independent of the true quality of the manuscript.