Question: How to stop and infinite regression
The regression you mention can be very useful in questioning one's assumptions, but it can also be fallacious. Each "how do you know" question is essentially questioning the truthfulness of the previous statement. At some point, the burden of proof shifts to the person asking "how do you know" to demonstrate that what you have claimed is false. For example:
Person A: People who regularly eat donuts for breakfast are almost all obese.
Person B: How do you know? (reasonable - burden is on person A to provide evidence for the claim)
Person A: I research this area, and conducted a meta-analysis comprising 12 studies that have been done in the last decade. The results were clear.
Person B: How do you know you didn't make a serious mistake? (reasonable, but less so - burden of proof can be argued either way)
Person A: I have been doing this for years and I am good at it. Besides, my work was peer reviewed and no serious mistakes were found.
Person B: How do you know that this time you didn't make a mistake, and that those who reviewed your work didn't all make errors as well? (unreasonable - we leave rational skepticism and enter denialism)
Person A: What makes you think that I might have and that all those who reviewed my work made errors as well?
Person B: Err... because, I..., just seems that way...
Person A: How do you know?
A deeper philosophical issue here is epistemology and foundational knowledge. There are many schools of thought here, and not worth getting into in casual argumentation. The technique above should prove very useful.