Page 70

It explained some Sybils can employ sophisticated algorithmic strategies that are indistinguishable from real people.

“They fund accounts from many distinct exchange addresses, never interact with each other, use randomized amounts, and use software to randomize daily patterns of human behavior, and even perform activities unique to the project (for example, using zkSync paymasters),” it explained.

“The majority of such bots are completely undetectable, even with the most advanced anti-Sybil methodology.”

ZkSync claimed being too aggressive with filtering could eliminate some Sybils but also might falsely flag many organic users — so it chose to reward organic users with “high likelihood” by using a combination of value scaling and multipliers.

Essentially, this assigned fewer eligibility points to wallets with low funding (a sign of Sybil behavior) but would give them a multiplying boost if there was onchain behavior that signaled human behavior.

ZkSync explained that Sybils typically creates many accounts but uses small amounts of crypto to fund each account to be capital efficient.