Aardvark uses vector commitments in a new way to build an authenticated dictionary suited for a blockchain. Updating vector commitment proofs costs a lot of CPU time, so Aardvark versions the dictionary to avoid these updates.
(Published at USENIX Security, 2022.)
Vault describes how Merkle tree sharding, delayed block certification, and a bit of careful design can help a blockchain scale in spite of growing storage and bootstrapping costs. (Bootstrapping here refers to the process of helping a new user join the system.) Vault is built on the Algorand cryptocurrency.
(Published at NDSS, 2019.)
Stadium is a metadata-private chat system which hides the participants of a conversation over the network. It improves upon the Vuvuzela chat system by enabling it to scale: Stadium efficiently serves more users with each additional server added.
(Published at SOSP, 2017.)
I evaluated the usability of some smart door locks. Smart door locks are difficult to get right because they’re part of a distributed system (for instance, the lock owner can remotely set access to the lock), so they must choose between failing open and failing shut when they can’t get a network connection.
(Published at AsiaCCS, 2016.)
I helped run a study on whether people notice that a hacked webcam is recording them. Most people don’t notice if they’re distracted by another activity (e.g., reading, writing, or playing a game).
(Published at SIGCHI, 2015.)