Michal Zajac

Konrad Zolna, Michal Zajac, Negar Rostamzadeh et al.

Neural networks are prone to adversarial attacks. In general, such attacks deteriorate the quality of the input by either slightly modifying most of its pixels, or by occluding it with a patch. In this paper, we propose a method that keeps the image unchanged and only adds an adversarial framing on the border of the image. We show empirically that our method is able to successfully attack state-of-the-art methods on both image and video classification problems. Notably, the proposed method results in a universal attack which is very fast at test time. Source code can be found at https://github.com/zajaczajac/adv_framing .

Antoine Rondelet, Michal Zajac

Transaction privacy is a hard problem on an account-based blockchain such as Ethereum. While Ben-Sasson et al. presented the Zerocash protocol [BCG+14] as a decentralized anonymous payment (DAP) scheme standing on top of Bitcoin, no study about the integration of such DAP on top of a ledger defined in the account model was provided. In this paper we aim to fill this gap and propose ZETH, an adaptation of Zerocash that can be deployed on top of Ethereum without making any change to the base layer. Our study shows that not only ZETH could be used to transfer Ether, the base currency of Ethereum, but it could also be used to transfer other types of smart contract-based digital assets. We propose an analysis of ZETH's privacy promises and argue that information leakages intrinsic to the use of this protocol are controlled and well-defined, which makes it a viable solution to support private transactions in the context of public and permissioned chains.