6.876J (MIT) and CS294175 (Berkeley)
Advanced Topics in Cryptography Spring 2020
Announcements
 Lecture videos for some lectures are recorded and posted; access is limited to class participants.
Course Information
INSTRUCTORS

Shafi Goldwasser
Yael Kalai
Vinod Vaikuntanathan

LOCATION 
Calvin Hall 116 (at Berkeley) and 9151 (at MIT)

TIME 
F 10am  12

FORMAT 
Each course participant will deliver one or more lectures on recent research results in cryptography, in consultation with the lecturers.

GRADING 
Each student gives one or more lectures or scribes one or more lectures.

Schedule (subject to change)
Abstracts:
Learning vs Verifying
This talk will address the following question: In what cases does learning a good hypothesis require more resources than verifying a hypothesis proposed by someone else?** If verification is significantly cheaper than learning, that could have important practical implications for delegation of machine learning tasks in commercial settings, and might also have consequences for verification of scientific publications, and for AI safety. Two results will be discussed: (1) There exists a learning problem where verification requires quadratically less random samples than are required for learning. (2) The broad class of Fouriersparse functions (which includes decision trees, for example) can be efficiently verified using random samples only, even though it is widely believed to be impossible to efficiently learn this class from random samples alone.
Jonathan is a PhD student at UC Berkeley. This is joint work with Shafi Goldwasser (UC Berkeley), Guy Rothblum (WIS), and Amir Yehudayoff (Technion).
Publicly Verifiable NonInteractive Delegation
We construct a delegation scheme for all polynomial time computations. Our scheme is
publicly verifiable and completely noninteractive in the common reference string (CRS) model.
Our scheme is based on an efficiently falsifiable decisional assumption on groups with bilinear maps. Prior to this work, publicly verifiable noninteractive delegation schemes were only
known under knowledge assumptions (or in the Random Oracle model) or under nonstandard
assumptions related to obfuscation or multilinear maps.
We obtain our result in two steps. First, we construct a scheme with a long CRS (polynomial
in the running time of the computation) by following the blueprint of Paneth and Rothblum
(TCC 2017). Then we bootstrap this scheme to obtain a short CRS. Our bootstrapping theorem
exploits the fact that our scheme can securely delegate certain nondeterministic computations.
Lisa is a PhD student at MIT. This is joint work with Yael Kalai (Microsoft Research and MIT) and Omer Paneth (TelAviv University).
CONGEST Lower Bounds via Communication Complexity
One of the wellstudied models of distributed graph algorithms is the CONGEST model. In this model, there is a network of n nodes that can communicate with each other via synchronous communication rounds. In each round, each node can send a lognbit message to each of its neighbors. The goal of the nodes is to compute some function of the network (e.g., the value of the maximum independent set) while minimizing the number of communication rounds.
In the first part of this talk, I will discuss a very fruitful technique for proving lower bounds for the CONGEST model. I will present recent progress in hardness of approximation of maximum independent set in the this model.
In the second part of this talk, I will discuss the classical lower bound for 2party setdisjointness (fooling sets method).
TwoRound MultiParty Computation without Lattices
Recent developments in the field of secure multiparty computation have resulted in protocols that allow parties to jointly compute arbitrary functionalities over their private inputs, with minimal interaction and based on 20th century tools and assumptions. In this talk, we will see the obfuscationbased “roundcompressing compiler” of [GGHR14] and how it was later instantiated using garbled circuits and oblivious transfer [GS18, BL18] to yield a tworound multiparty computation protocol from minimal assumptions. Depending on time, we’ll also see how to use the DDH assumption to reduce interaction even further by making the first round messages reusable across an arbitrary number of second round executions, each of which may compute a different function over the parties’ inputs [BGMM20].
SafetyPin: Encrypted Backups with HumanMemorable Secrets
We present the design and implementation of SafetyPin, a system for encrypted mobilephone backups. Like existing mobilebackup systems, including those of Apple and Google, SafetyPin requires users to remember only a short PIN and prevents bruteforce PINguessing attacks using hardware security protections. Unlike today’s systems, SafetyPin provides strong protection against hardware compromise: the system protects the confidentiality of backedup user data even against an attacker that can adaptively compromise many of the system’s constituent hardware security modules. To achieve these security properties while respecting the resource limits of today’s hardware security models, we develop a collection of new cryptographic tools.