He may be referring to opsec more generally (onramps, offramps), or he may be specificially referring to poisoned output attacks: https://www.youtube.com/watch?

He may be referring to opsec more generally (onramps, offramps), or he may be specificially referring to poisoned output attacks: https://www.youtube.com/watch?v=iABIcsDJKyM https://www.youtube.com/watch?v=iABIcsDJKyM TL;DR Ring signatures, like all sender-obfuscation methods, have a limited anonymity set: it limits you to a pool of possible senders. If Alice frequently sends funds to Bob, who frequently sends funds to Carl, who frequently sends to Alice, she can see that Alice->Bob->Carl->Alice is one possible outcome. She does this because she can trace the coin she associates with Bob to a coin she associates with Carl. There is a ton of plausible deniability at first, but the relationship between Bob and Carl becomes more obvious the more Alice->Bob->Carl->Alice continues to happen. Alice can be multiple exchanges collaborating using KYC. How to resist poisoning: limit your risk, churn, bigger ring-size/anonymity set, do atomic swaps (this severs chain of ownership, but is not generally sybil-proof), do multi-output transactions if you are sending to multiple people at once who can co-ordinate (this reduces the number of coins they can co-ordinate).