Grants Funded by the Center

Current/Past Grants

Current Research Grants

Improved Algorithms for Generic Attacks on Cryptographic Hash Functions

PI: Keller N.

Cryptographic hash functions are one of the basic primitives in cryptography.

They are also used for numerous applications in cyber security, for example for signing messages, applications, and software updates. One of the main methods of comparison between strategies for designing hash functions is generic attacks. While usually non-practical, attacks of this type point at a structural weakness in a design strategy, and may make us prefer a more conservative (or just a different) design. Numerous generic attacks were presented in recent years. This research project aims at improving the existing algorithms for generic attacks on hash functions, and (more ambitiously) at devising new such algorithms.

Personal Information Management and Sharing

PI: Amsterdamer Y.

The distribution of personal data such as emails, photos, and social interactions over platforms of commercial vendors such as social networks, is of great current concern.

By giving up the management of personal data the users do not realize the full potential of the data that they generate: e.g., they cannot enforce certain policies and access control schemes on data they share with peers across multiple vendors. The goal of this research project is to develop a declarative, generic and robust framework that will allow users to jointly access and manage data they store in different platforms, and to easily share this data with others without compromising access control rights and data integrity.

Algebraic Lightweight Cryptography

PI: Tsaban B.

Present-day secure computation is mostly based on abelian mathematical structures and problems: Present-day secure computation is mostly based on abelian mathematical structures and problems: the discrete logarithm problem in finite cyclic groups (Diffie–Hellman), integer factorization (RSA), and more recently, lattice-based problems.

To reduce the dependence of cyber-security on a small number of problems, it is desirable to also have candidate problems of substantially different types. On the practical side, this would make it easier to tailor optimal implementations in constrained environments, such as RFID tags, that are more suitable for Internet of Things (IoT) applications.   Recently, some candidates for lightweight cryptographic primitives, based on nonabelian structures, begin to stand out as potentially secure. We will consider practical aspects of their security and efficiency.

Computing with Crypto-currency

PI: Moran T.

Crypto currencies, such as Bitcoin and Ethereum, are more than simply “decentralized digital cash”. One of their most interesting features is the ability to create “smart contracts”, that are enforced automatically by the decentralized system, as opposed to by a government or a court.  That is, crypto currencies allow us to combine computation with money, in new and exciting ways.


This research project aims to deepen our understanding of this interaction. On one hand, we will study the applications opened up by computing with existing crypto currencies (such as auctions with guaranteed fair execution and decentralized online games), looking at their complexity under various metrics, as well as searching for new applications.  On the other hand we will explore the underlying crypto currency protocols themselves, proposing extensions and alternative constructions that could improve the existing state of the art.

The vulnerability of deep networks to adversarial examples

PI: Keshet J

Deep neural network models are used in all segments of the modern industry from self-driving cars to automated dialog agents. It becomes critical to evaluate their vulnerability to security threats, such as adversarial examples.

An adversarial example is a synthetic pattern carefully crafted by adding perturbations to an existing pattern. The resulted pattern is indistinguishable from the original pattern by a human, yet they have demonstrated a strong ability to cause catastrophic failure of state of the art systems.

The goal of this proposal is to explore the effect of the surrogate loss functions, the model, and the training data on the robustness of neural networks to adversarial examples.  We would like to understand the phenomenon and to propose more robust training algorithms.


Students: Yossi Adi, Shir Aviv, Felix Kreuk