Searcharxiv⌕ Search

arXiv subjects

Marcin Andrychowicz

Publications and source records attributed to Marcin Andrychowicz.

22 records · Page 2Linked to original sources

Neural Random-Access Machines

In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.

cs.LG↗

Modeling Bitcoin Contracts by Timed Automata

Bitcoin is a peer-to-peer cryptographic currency system. Since its introduction in 2008, Bitcoin has gained noticeable popularity, mostly due to its following properties: (1) the transaction fees are very low, and (2) it is not controlled by any central authority, which in particular means that nobody can "print" the money to generate inflation. Moreover, the transaction syntax allows to create the so-called contracts, where a number of mutually-distrusting parties engage in a protocol to jointly perform some financial task, and the fairness of this process is guaranteed by the properties of Bitcoin. Although the Bitcoin contracts have several potential applications in the digital economy, so far they have not been widely used in real life. This is partly due to the fact that they are cumbersome to create and analyze, and hence risky to use. In this paper we propose to remedy this problem by using the methods originally developed for the computer-aided analysis for hardware and software systems, in particular those based on the timed automata. More concretely, we propose a framework for modeling the Bitcoin contracts using the timed automata in the UPPAAL model checker. Our method is general and can be used to model several contracts. As a proof-of-concept we use this framework to model some of the Bitcoin contracts from our recent previous work. We then automatically verify their security in UPPAAL, finding (and correcting) some subtle errors that were difficult to spot by the manual analysis. We hope that our work can draw the attention of the researchers working on formal modeling to the problem of the Bitcoin contract verification, and spark off more research on this topic.

cs.CR↗

How to deal with malleability of BitCoin transactions

BitCoin transactions are malleable in a sense that given a transaction an adversary can easily construct an equivalent transaction which has a different hash. This can pose a serious problem in some BitCoin distributed contracts in which changing a transaction's hash may result in the protocol disruption and a financial loss. The problem mostly concerns protocols, which use a "refund" transaction to withdraw a deposit in a case of the protocol interruption. In this short note, we show a general technique for creating malleability-resilient "refund" transactions, which does not require any modification of the BitCoin protocol. Applying our technique to our previous paper "Fair Two-Party Computations via the BitCoin Deposits" (Cryptology ePrint Archive, 2013) allows to achieve fairness in any Two-Party Computation using the BitCoin protocol in its current version.

cs.CR↗

Efficient Refreshing Protocol for Leakage-Resilient Storage Based on the Inner-Product Extractor

A recent trend in cryptography is to protect data and computation against various side-channel attacks. Dziembowski and Faust (TCC 2012) have proposed a general way to protect arbitrary circuits against any continual leakage assuming that: (i) the memory is divided into the parts, which leaks independently (ii) the leakage in each observation is bounded (iii) the circuit has an access to a leak-free component, which samples random orthogonal vectors. The pivotal element of their construction is a protocol for refreshing the so-called Leakage-Resilient Storage (LRS). In this note, we present a more efficient and simpler protocol for refreshing LRS under the same assumptions. Our solution needs O(n) operations to fully refresh the secret (in comparison to Ω(n^2) for a protocol of Dziembowski and Faust), where n is a security parameter that describes the maximal amount of leakage in each invocation of the refreshing procedure

cs.CR↗