How do you redact secrets in logs?
PICTURE THIS: A SENTENCE BECOMES TOKENS
The model does not read letters like humans. It reads these pieces, then predicts the next one.
Simple meaning
I never log tokens, passwords, or full connection strings.
WHY — Security instead of guessing?
Why interviewers care about Security:
question about Security.
trade-offs, and what you would actually do on a .NET project - not buzzwords.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens with tokens?
Before the model can read a sentence, it goes through these steps:
- 1I never log tokens,
passwords, or full connection strings.
- 2Structured logging enrichers can
scrub known keys.
- 3Embeddings
Exception messages also get reviewed.
- 4Context mix
Attention looks at nearby tokens together.
- 5Next token
The model scores what should come next.
- 6Decode
IDs turn back into readable text.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
I never log tokens, passwords, or full connection strings. Structured logging enrichers can scrub known keys.