Audio fingerprinting techniques were developed to index and retrieve audio samples by comparing a content-based compact signature of the audio instead of the entire audio sample, thereby reducing memory and computational expense. Different techniques have been applied to create audio fingerprints; however, with the introduction of deep learning, new data-driven unsupervised approaches are available. This article presents Sequence-to-Sequence Autoencoder Model for Audio Fingerprinting (SAMAF), which improved hash generation through a novel loss function composed of terms: Mean Square Error, minimizing the reconstruction error; Hash Loss, minimizing the distance between similar hashes and encouraging clustering; and Bitwise Entropy Loss, minimizing the variation inside the clusters. The performance of the model was assessed with a subset of VoxCeleb1 dataset, a“speech in-the-wild” dataset. Furthermore, the model was compared against three baselines: Dejavu, a Shazam-like algorithm; Robust Audio Fingerprinting System (RAFS), a Bit Error Rate (BER) methodology robust to time-frequency distortions and coding/decoding transformations; and Panako, a constellation-based algorithm adding time-frequency distortion resilience. Extensive empirical evidence showed that our approach outperformed all the baselines in the audio identification task and other classification tasks related to the attributes of the audio signal with an economical hash size of either 128 or 256 bits for one second of audio.

Blockchain, or distributed ledger, provides a way to build various decentralized systems without relying on any single trusted party. This is especially attractive for smart contracts, that different parties do not need to trust each other to have a contract, and the distributed ledger can guarantee correct execution of the contract. Most existing distributed ledger based smart contract systems process smart contracts in a serial manner, i.e., all users have to run a contract before its result can be accepted by the system. Although this approach is easy to implement and manage, it is not scalable and greatly limits the system's capability of handling a large number of smart contracts. In order to address this problem, we propose a scalable smart contract execution scheme that can run multiple smart contract in parallel to improve throughput of the system. Our scheme relies on two key techniques: a fair contract partition algorithm leveraging integer linear programming to partition a set of smart contracts into multiple subsets, and a random assignment protocol assigning subsets randomly to a subgroup of users. We prove that, our scheme is secure as long as more than 50% of the computational power is possessed by honest nodes. We then conduct experiments with data from existing smart contract system to evaluate the efficiency of our scheme. The results demonstrate that our approach is scalable and much more efficient than the existing smart contract platform.

Transaction system build on top of blockchain, especially smart contract, is becoming an important part of world economy. However, there is a lack of formal study on the behavior of users in these systems, which leaves the correctness and security of such system without a solid foundation. Unlike mining, in which the reward for mining a block is fixed, different execution results of a smart contract may lead to significantly different payoffs of users, which gives more incentives for some user to follow a branch that contains a wrong result, even if the branch is shorter. It is thus important to understand the exact probability that a branch is being selected by the system. We formulate this problem as the (+-)-Biased Ballot Problem as follows: there are n voters one by one voting for either of the two candidates A and B. The probability of a user voting for A or B depends on whether the difference between the current votes of A and B is positive or negative. Our model takes into account the behavior of three different kinds of users when a branch occurs in the system -- users having preference over a certain branch based on the history of their transactions, and users being indifferent and simply follow the longest chain. We study two important probabilities that are closely related with a blockchain based system - the probability that A wins at last, and the probability that A receives d votes first. We show how to recursively calculate the two probabilities for any fixed n and d, and also discuss their asymptotic values when n and d are sufficiently large.