Most integrity problems at fair level are not fraud. They are shortcuts taken under deadline pressure by students who did not realise the line was there.
Attribution. Any idea, method, figure, or sentence that is not yours needs a citation. Adapting someone's protocol is normal and expected — presenting it as your own design is not.
Mentor contribution. If a mentor ran an instrument, prepared a reagent, or wrote analysis code, say so explicitly in your write-up and your interview. Judges respect a clearly delineated contribution far more than a vague "I did everything".
Data handling. Never delete an inconvenient data point without a rule you set in advance and stated in the write-up. Outlier exclusion is legitimate; silent exclusion is not.
Reporting negatives. A result that does not support your hypothesis goes in the poster. Omitting it is the most common serious integrity failure in student work.
AI assistance. Follow your fair's rules, and be able to say exactly what you used it for. Text you did not write and cannot explain is a problem in an interview regardless of the rules.
The working test: would you be comfortable explaining exactly how this was produced, to a judge, out loud? If not, fix it now rather than hoping.