If I understand correctly, the paper has only been submitted for peer review, and this is thus not yet the camera-ready final version that will be archived and widely disseminated. As such, you have an opportunity to fix the error before it enters the literature.
The key question, then, is whether your error will mislead the peer reviewers into having a different judgement of the paper than they would otherwise (whether better or worse). The first question I would thus ask is this: how different are the wrong plots from the right plots?
- If the results are essentially identical (just unsupported because of the error), then I would let it go and just make sure you fix it in subsequent revision.
- If there is a major difference, then you need to update the paper, because otherwise you will be misleading the reviewers.
Updating the paper may be easy or may be difficult depending on the conference and its submission and review timeline. Since it's quite soon after the submission deadline, there's a fairly high chance that no reviewers will yet have actually reviewed your paper, and so updating isn't likely to cause problems.
As for how to actually do it: occasionally conferences will allow you to continue updating a submission (whether intentionally or by misconfiguring their review system); for most, though, you will need to email the program chairs who are running the review process. If they are feeling generous, they will allow you to update; if they are not and the error is a major problem, then you may need to withdraw your paper and instead submit more carefully to the next good opportunity.
There are a lot of ways to salvage this situation, though, so hopefully you'll have no difficulty in correcting your paper, whether now or at the camera-ready stage.
