Ittai Abraham Profile picture
I do research in algorithms and distributed computing
Sep 14, 2024 6 tweets 2 min read
"Even if MPC works you still need TEEs"

Pioneer blockchain researcher @socrates1024 in a great podcast about TEEs 🔥

Some of my notes ⤵️ 1. Must watch explanation about delegation and encumbrance - this is clearly going to be a major future use case. Wonderful clarity around ZK enabling read permissions vs TEE also enabling smart contract based write permissions

ZK, TEEs and MPC will all play a major role here
Jan 23, 2024 4 tweets 1 min read
1/4 Motorway: a new super high throughput and low latency BFT system that is **not a DAG** by lead authors Neil Giridharan and Florian Suri-Payer !!

arxiv.org/abs/2401.10369 2/4 BFT systems in partial synchrony loose liveness in asynchrony and regain it once synchrony comes back.

But when operating on constant load most systems suffer a **hangover** - they have to spend considerable time catching up on all the transactions they could not process.
Jan 10, 2024 8 tweets 2 min read
Decentralized Thoughts: we are nearing 5 years and 100+ posts!

A few updates for this year 🧵 First, check out the awesome post by Ling Ren on a Simpler Security proof for Nakamoto Consensus

It's amazing that Bitcoin is 15 years old and it took so long to get clear security proofs for it

decentralizedthoughts.github.io/2023-10-30-Ana…
Apr 24, 2023 10 tweets 2 min read
1/From the seminal work of BGW88 and CCD88 we know that perfectly secure (unbounded adversary - just private channels) optimally-resilient (t<n/3) secure Multi-Party Computation (MPC) is possible. But is it efficient? 2/ For a circuit with depth d, there are, in general, two families of protocols:
1. Efficient but slow: O(n) words per gate, but O(n+d) rounds
2. Fast but not efficient: O(d) expected rounds, but O(n^4) words per gate
Can we get the best of worlds?
Jan 28, 2023 5 tweets 2 min read
1/5 Chained Byzantine Fault Tolerant systems leverage pipelining and leader rotation for efficiency and fairness. Pipelining significantly increases throughput. Leader rotation means that each chosen leader speaks only once, significantly reducing risk of bribery and attacks 2/5 Existing protocols require a sequence of three **consecutive** honest leaders to commit. Even simple leader crashes weaken liveness in theory and practice. Obtaining chained BFT that commits even if the sequence of honest leaders is non-consecutive, remains an open question
Jan 25, 2023 5 tweets 2 min read
1/5 View synchronisation is an important component of many modern Byzantine Fault Tolerant State Machine Replication (SMR) systems in the partial synchrony model. The efficiency of view synchronisation has emerged as a bottleneck in the efficiency of SMR systems as a whole 2/5 A key question remained open: Do there exist view synchronisation protocols with asymptotically optimal quadratic worst-case word complexity that also obtain linear message complexity and responsiveness when moving between consecutive correct leaders?
Jan 20, 2023 5 tweets 1 min read
1/5 My mental model: a "synchronous BFT protocol" is a shorthand for saying a protocol has a proof of safety and a proof of liveness under the synchronous network model. As you know, the best fault tolerance you can get in this model is f<n/2. 2/5 Similarly, a "partially synchronous BFT protocol" is shorthand for saying a protocol has a proof of safety under the asynchronous network model and proof of liveness under the synchronous network model. As you know, the best fault tolerance you can get in this model is f<n/3.
Sep 5, 2022 15 tweets 7 min read
What is your definition of a blockchain?

decentralizedthoughts.github.io/2022-09-05-wha… 2/ the post has many links, here are some as a thread:

@bhorowitz “a computer with trust”