When a question comes up about possible duplicate records and fraud, I think it's a good idea to step back and ask if you understand the record groups you are looking at.
We have a similar question here about a different record type: Two ships, two manifests, same passengers? where someone asked about fraud on a passenger list, but there's a much simpler explanation if you know how to interpret the records.
You say "there is someone with almost identical details" but have you considered how many people across the entire record set might share those same details? The more common a name is, the more identifiers you need to distinguish similar-named individuals from each other.
You've only talked about census records here, and you don't say what other record types you may have looked at. One of the elements of the Genealogical Proof Standard is to do a reasonably exhaustive search. Have you looked at other records? If you can build a separate timeline, from cradle-to-grave/probate, for each of the people in the two census records you are talking about, then you are likely to be looking at two different people.
Another principle of the Genealogical Proof Standards is to evaluate the records belonging to a person or family as a group.
As a case study, I will list two memorials from FindAGrave:
Memorial ID 66775782 for Frederick Maximilian Paul “Max” Bretschneider and his wife Memorial ID 66776488 for Ida Emily Poehler Bretschneider (whose maiden name is not Poehler, but Ida Georgi)
The owner of these memorials has cited many source records in his biographies for these two individuals. He lists their parents and six children. However, if you examine the records closely and look at other record groups in the localities, it quickly becomes apparent that he has conflated the records belonging to two different couples named "Max" and "Ida". The parents and half of the children listed in these biographies belong to a couple who lived elsewhere in the state.
I suspect the author of these bios was led astray by an Ancestry hint which suggests the 1930 Census record of this Newton Massachusetts Max Bretschneider to fill in a gap for the missing 1930 Census record for the Max who didn't live in Boston. If you look at all the records available online, such as city directories, historical newspapers, and court records, you see that the records fall neatly into two separate groups. Not only do the records separate themselves neatly by geography, the children's records make it especially apparent, because they fall into two groups of three siblings. If you have a set of six people but their records fall into two groups of three, and both group A and group B only reference each other's records, then you need to step back and reconsider what's going on. Sure, family feuds happen, but it's much more likely that Group A's records don't mention Group B and vice-versa because they don't belong to the same family.
Use the FAN principle (Friends, Associates, Neighbors) as well as extended family to learn more about the people in your two census records. QuickLesson 11: Identity Problems & the FAN Principle is a worked case study to identify a woman named Mary Smith using records of the people around her.
Because of mix-ups like the one shown on Find A Grave, my general rule when I see almost-matching records with not-quite-enough information is to assume they are for two different people. I try to prove that they aren't the same person (by building out the timeline for each one). If I'm wrong, and the records do belong to the same person, the information in the records will point back in that direction.
Records can provide evidence of fraud, and as part of the Genealogical Proof Standard, we do need to provide an explanation for contradictory information. But it's hard for me to see what an individual would have to gain by deliberately being counted on a census twice. A census is a very different type of record than something like a military enlistment where someone might lie about their age because they were too young but wanted to sign up anyway. Be careful that you aren't basing your conclusion too much on a coincidence of names and ages in the absence of other identifiers.