Glossary

AI and web3 terminology

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

A

51% Attack

A type of assault on a blockchain that occurs when a miner or a group of miners control more than half of the network's hashing rate.

A/B Testing

An experimental comparison of two system variants, A and B, conducted in a real-world setting.

Activation Function

In Artificial Neural Networks, it's a function that produces an output value for the next layer by taking the weighted sum of inputs from the previous layer.

Active Learning (Active Learning Strategy)

A subcategory of Semi-Supervised Machine Learning, where a learning agent actively queries an oracle (usually, a human annotator) to gain labels for new data points.

Airdrop

A free distribution of cryptocurrency tokens intended to promote a new project or boost community adoption.

Algorithm

A clear set of instructions defining how to solve a class of problems, capable of carrying out computations, data processing, and automated reasoning.

Altcoin

An abbreviation for "alternative coin," which refers to any cryptocurrency other than Bitcoin.

Annotation

A piece of metadata attached to a data unit, typically supplied by a human annotator.

Area Under the Curve (AUC)

A technique in Machine Learning used to identify the highest-performing model among several.

Arithmetization

Arithmetization is the method of converting computational challenges into mathematical formulas. Essentially, it transforms computer code into "mathematical expressions," allowing for examination through cryptographic algebraic methods.

Artificial Intelligence

Systems that are designed to perform tasks with intelligence comparable to that of a human, such as learning, reasoning, problem-solving, and decision making.

Artificial Neural Networks

An architecture comprised of interconnected layers of basic units known as artificial neurons, interspersed with non-linear activation functions. This structure is loosely based on animal brain neurons.

Association Rule Learning

A rule-oriented Machine Learning method that uncovers interesting relationships between variables in large data sets.

Autoencoder

A type of Artificial Neural Network used to create efficient data representations in a non-linear and unsupervised way, often used for dimensionality reduction.

Automated Speech Recognition

A field within Computational Linguistics that focuses on techniques allowing computers to recognize and translate spoken language into text.

B

Backpropagation (Backpropagation through time)

A method employed in training Artificial Neural Networks to calculate the gradient necessary for adjusting the network's weights.

Batch

A collection of examples used in one gradient update during model training.

Bayes' Theorem

A well-known theorem used by statisticians to determine the likelihood of an event based on prior knowledge of related conditions.

Bear Market

A prolonged period characterized by declining prices in the financial market.

Bias (Inductive Bias, Confirmation Bias)

Inductive Bias: The assumptions used by the learning model when predicting outputs for unseen inputs.

Confirmation Bias: A bias where one favors and interprets information in a way that confirms their own beliefs or hypotheses, neglecting contradictory information.

Bias-Variance Tradeoff

A conflict encountered when trying to simultaneously minimize bias and variance, hindering supervised algorithms from generalizing beyond their training data.

Bitcoin

The first and most capitalized cryptocurrency globally, created in 2009 by an unidentified creator known as "Satoshi Nakamoto."

Blockchain

A distinct type of distributed ledger technology used to chronicle data, characterized by its unchangeable nature.

Blockchain trilemma

The blockchain trilemma presents a dilemma where one can only prioritize two out of three factors: decentralization, security, and scalability. The Ethereum community has opted for decentralization and security at the expense of scalability. This trade-off leads to higher gas prices, potentially limiting Ethereum's widespread adoption by making it cost-prohibitive for some users.

Boosting

A Machine Learning meta-algorithm that's primarily used to reduce bias and variance in supervised learning. It's a set of algorithms that transform weak learners into strong ones.

Bounding Box

A rectangular region that encloses an object of interest in computer vision tasks.

Bull Market

An extended period marked by increasing prices in the financial market.

C

Central Bank Digital Currency (CBDC)

A digital rendition of a country's official fiat currency, administered by the central bank.

Chatbot

A computer program or an AI specifically designed to converse with human users.

Circuit

A circuit is a representation of computation through a series of logic gates. It illustrates the data flow and the operations executed on the data. In zero-knowledge proofs, a circuit is formulated by encoding the claims to be verified, along with the inputs, outputs, and transaction steps. The prover then produces a proof, often by mimicking the circuit's execution, but only discloses if the execution was valid. The verifier, leveraging the circuit's representation, ensures the prover's computations are accurate and the results are valid.

Classification

The task of determining a function that maps input variables to discrete output variables. In the broader context, it refers to a class of Machine Learning algorithms used to categorize specific instances.

Clustering

In Machine Learning, it's an unsupervised task of grouping objects in such a way that objects in the same group (or cluster) are more similar to each other than those in other groups.

Cold-Start

A potential problem that arises when a system can't make inferences for users or items due to insufficient information.

Collaborative Filtering

A technique used in recommender systems to predict a user's interests by collecting preferences from a larger user group.

Computer Vision

A subset of Machine Learning that focuses on deriving high-level understanding from images or videos.

Confidence Interval

A type of interval estimate that is likely to contain the true value of an unknown population parameter. The interval is associated with a confidence level that the parameter is within the interval.

Constraint

A constraint is a stipulated condition applied to mathematical processes or computations, aiming to guarantee the accuracy, efficiency, or safety of a cryptographic technique.

Contributor

A human worker who provides annotations on the Appen data annotation platform.

Cryptography

The art and science of securing communication in the presence of adversaries.

D

DApps

Abbreviation for "decentralized applications." These are autonomous apps that run on distributed networks using smart contracts, which automatically execute actions like transactions when specific conditions are met.

Data (Structured Data, Unstructured Data, Data Augmentation)

The cornerstone of all Machine Learning and Artificial Intelligence projects.

Unstructured Data

Raw and unformatted data. Text-based data is a prime example as it's not organized into specific features.

Structured Data

Data organized in a way that makes it suitable for ingestion by a Machine Learning algorithm. In the context of Supervised Machine Learning, it refers to labeled data or data post-processing on the Appen data annotation platform.

Data Augmentation

The practice of enriching a data set with additional information derived from internal and external sources, usually via annotation.

Decision Tree

A type of Supervised Machine Learning algorithm that iteratively divides the data based on a specific parameter or criteria.

Deep Blue

An IBM-developed chess-playing computer famous for being the first of its kind to win a game and match against a reigning world champion under standard time controls.

Deep Learning (Deep Reinforcement Learning)

A subset of Machine Learning techniques emphasizing learning data representations rather than algorithm-specific tasks.

DeFi

Stands for "decentralized finance," an umbrella term for all finance-based DApps built on public blockchains.

Digital Currency

Refers to any form of money that is entirely digital and transferred electronically over the internet.

Dimensionality (Dimensionality Reduction, Curse of Dimensionality)

Dimensionality Reduction: The act of decreasing the number of random variables by obtaining a set of principal variables.

Curse of Dimensionality: The complications encountered when analyzing and organizing data in high-dimensional spaces.

Dogecoin

The proprietary cryptocurrency of the Dogecoin blockchain, a popular blockchain project created around a viral internet meme featuring a Shiba Inu dog.

E

Elliptical Curve Cryptography

Elliptic Curve Cryptography (ECC) is a branch of public-key cryptography based on the algebraic structure of elliptic curves over finite fields. Instead of traditional mathematical methods, ECC utilizes the properties of elliptic curves, offering the same level of security with smaller key sizes, which can result in faster computations and reduced resource usage. The strength of ECC derives from the difficulty of the Elliptic Curve Discrete Logarithm Problem (ECDLP), making it computationally infeasible to determine the private key from a known public key. Due to its efficiency and strong security characteristics, ECC has become increasingly popular in modern cryptographic applications, from securing websites, enhancing mobile device encryption, and blockchain applications.

Embedding (Word Embedding)

A mathematical structure, such as a group, that exists within another similar structure.

Ensemble Methods

In Statistics and Machine Learning, ensemble methods combine multiple learning algorithms to achieve better predictive performance.

Entropy

The average amount of information produced by a data source with randomness.

Epoch

In the context of training Deep Learning models, an epoch signifies a complete pass through the entire training data set.

Ethereum

The blockchain supporting the second-largest cryptocurrency by market capitalization, Ethereum's currency is called ether. The Ethereum network incorporates smart contracts directly and is the basis for numerous digital currencies and projects, including decentralized apps (DApps), non-fungible tokens (NFTs), and decentralized finance (DeFi).

F

False Negative

An error in which the null hypothesis is incorrectly accepted when it should have been rejected.

False Positive

An error in which the null hypothesis is incorrectly rejected.

Feed-Forward (Neural) Networks

A type of Artificial Neural Network where the connections between neurons do not loop back or create a cycle.

Feature

An individual measurable property or characteristic serving as an input to a model.

Feature Learning

A range of methods designed to automatically uncover the required representations for feature detection or classification from raw data.

Fiat

A term used for any official currency issued by a government, such as the U.S. dollar or the Chinese yuan.

FOMO

An acronym for "fear of missing out," often used to express worry about potentially missing a promising opportunity.

F-Score

A metric that gauges a model's accuracy by considering both precision and recall to calculate the score.

FUD

An abbreviation for "fear, uncertainty, doubt," frequently used in the crypto world to categorize the dissemination of negative news or information.

G

Gas

Ether required to be paid as a fee to perform a transaction or contract on the Ethereum network.

Garbage In, Garbage Out

A concept asserting that flawed input data will invariably lead to erroneous or nonsensical output.

General Data Protection Regulation (GDPR)

An EU law that provides data protection and privacy for all individuals within the European Union.

Genetic Algorithm

A search heuristic influenced by the Theory of Evolution, mirroring the process of natural selection.

Generative Adversarial Networks (GANs)

A category of Artificial Intelligence algorithms used in Unsupervised Machine Learning, constructed as a pair of competing Neural Networks in a zero-sum game framework.

Graphic Processing Unit (GPU)

A dedicated electronic circuit built to quickly process and modify memory to speed up image rendering.

Ground Truth

Information acquired through direct observation as opposed to inference.

H

HALO2

A proving system grounded in Plonkish arithmetization and incorporates features influenced by UltraPLONK, which is an enhanced version of the PLONK proving system that supports custom gates and lookup tables. Beyond its application in ZCash, Halo2 has been adopted by entities like Protocol Labs, the Ethereum Foundation's Privacy and Scaling Explorations (PSE) team, Scroll, and Taiko. This has cemented its position as one of the leading zkSNARK architectures in the present day.

Halving

A 50% cut in mining block rewards for a specific cryptocurrency. For Bitcoin, these are programmed to occur every four years.

HODL

A deliberate misspelling of "hold" and common slang in the crypto world used to endorse or describe the concept of never selling crypto. HODL is retrospectively considered an acronym for "hold on for dear life."

Homomorphic Encryption

Fully homomorphic encryption, often just called homomorphic encryption, is a category of encryption techniques first conceptualized by Rivest, Adleman, and Dertouzos in 1978 and initially realized by Craig Gentry in 2009. What sets homomorphic encryption apart from conventional encryption approaches is its capability to conduct operations directly on encrypted data, all without needing the secret key. The outcome of this operation stays encrypted, and only the secret key's possessor can decrypt and uncover it later.

Human-in-the-Loop

Human-in-the-Loop (HITL) is a field of artificial intelligence that combines human and machine intelligence to develop machine learning models. In a conventional HITL approach, humans are involved in a continuous cycle of training, tuning, and testing a specific algorithm.

Hyperparameter (Hyperparameter Tuning)

The manual process of determining the optimal configuration for training a specific model.

I

Image Recognition

The task in Computer Vision that involves identifying if an image contains a specific object, feature, or activity.

Inference

The act of making predictions by applying a trained model to new, unlabeled instances.

Information Retrieval

A domain within Computer Science that explores the process of searching for information in a document, searching for the documents themselves, and also for metadata that describes data, as well as searching for databases of text, images, or sounds.

Inflation

An economic concept describing when the prices of goods and services increase, resulting in a decrease in a fiat currency's purchasing power.

Interactive Proof

A Zero-Knowledge Proof in which the prover and verifier participate in several attempts to confirm a proof's authenticity.

L

Layer (Hidden Layer)

A collection of neurons in an Artificial Neural Network that process an input feature set or, alternatively, the output of those neurons. Hidden Layer: a neuron layer where the outputs are connected to the inputs of other neurons, hence not directly observable as a network output.

Layer 1

A Layer 1 (L1) blockchain serves as the fundamental and primary framework upon which additional layers, like Layer 2 (L2) solutions, can be built. Essentially, it's the main blockchain infrastructure that ensures network security, transaction validation, and consensus. Ethereum, as an example, is an L1 blockchain, encompassing node operators responsible for network security, block producers who record transactions, a historical archive of those transactions, and the specific consensus method by which the blockchain operates.

Layer 2

A Layer 2 (L2) blockchain is an auxiliary structure designed to enhance transaction speeds and capacity, drawing its security either entirely or in part from Ethereum. It mandates that L2 solutions post their transaction data to Ethereum to ensure data availability. By consolidating multiple transactions into a single submission on Ethereum, L2 solutions alleviate congestion on the primary Layer 1, allowing for improved scalability. As a result, L2s benefit from Ethereum's robust data reliability, security, and decentralized nature.

Learning-to-Learn

An emerging domain in Machine Learning that examines how algorithms can self-modify their generalization process by scrutinizing and improving upon their own learning.

Learning-to-Rank

The usage of Machine Learning to build ranking models for Information Retrieval systems.

Learning Rate

A scalar value employed by the gradient descent algorithm at each iteration of an Artificial Neural Network's training phase to multiply with the gradient.

Liquidity

A term indicating how quickly an asset can be bought or sold at any given price, volume, and time.

Logit Function

The reverse of the sigmoid "logistic" function utilized in mathematics, predominantly in statistics.

Long Short-Term Memory Networks

A form of Recurrent Neural Network proposed as a resolution to the vanishing gradient issue.

M

Machine Learning

A subset of AI, machine learning refers to machines and systems that learn from data and perform tasks without direct programming, instead using algorithms to analyze data, make predictions and recognize patterns.

Machine Translation

A sector within computational linguistics that explores the use of software to translate text or speech from one language to another.

Mainnet

A mainnet in blockchain technology is the primary network for a specific cryptocurrency, formed by interconnected nodes in a peer-to-peer (P2P) fashion. Key elements of a mainnet include its nodes, the associated cryptocurrency offering economic incentives to support and sustain the network, a consensus protocol facilitating transaction validation, and linked storage blocks creating the characteristic "chain" structure. It's the operational platform where real transactions take place on the distributed ledger.

Market Cap

A term used to calculate the total value of a project, company, or other entity. In crypto, the market cap is determined by multiplying the current price of a project's token by the total circulating supply.

Mining

The practice of using computer equipment to validate and add new data to a blockchain ledger, as well as introducing new coins into circulation.

Model

An abstract representation of what a Machine Learning system has learned from the training data during the training phase.

Monte Carlo

A technique that employs repeated random sampling to generate synthetic simulated data.

Multi-Modal Learning

A segment of Machine Learning focused on interpreting multimodal signals collectively and constructing models capable of processing and relating information from various types of data.

Multi-party Computation (MPC)

Multi-party computation is a cryptographic technique allowing several participants to collaboratively compute a function using their private inputs. Throughout this process, each participant's input remains confidential, ensuring that while the collective outcome is achieved, individual data is not disclosed to the other participants.

Multi-Task Learning

A segment of Machine Learning that leverages similarities and differences across tasks in order to address multiple tasks simultaneously.

N

Naive Bayes

A series of straightforward probabilistic classifiers based on applying Bayes' Theorem with rigorous independence assumptions among the features.

Named Entity Recognition

A subset of Information Extraction that aims to identify and classify named entities in text into pre-set categories such as names, locations, parts-of-speech, etc.

Natural Language Processing (NLP)

A field within Artificial Intelligence that explores interactions between computers and human languages, specifically how to process and analyze large volumes of natural language data.

Neural Networks

see Artificial Neural Networks

Neuron

A component in an Artificial Neural Network that processes multiple input values to yield a single output value.

NFTs

Stands for "non-fungible tokens." NFTs are digital tokens that validate ownership of unique physical and non-physical items.

Node

see Neuron

Non-interactive Proof

A Non-Interactive Zero-Knowledge Proof is a cryptographic method where the prover can produce a single zero-knowledge proof without interaction between the prover and verifier.

O

Optical Character Recognition

The transformation of images of printed, handwritten or typed text into a machine-readable textual format.

Optimization

The choice of the optimal element (based on some criteria) from a range of available alternatives.

Oracle

This is a computational unit or service that delivers data or responses from external sources (off-chain) to the blockchain (on-chain) for utilization by smart contracts.

Overfitting

A situation where a model inadvertently discerns patterns in the noise and interprets these as representing the underlying structure; the creation of a model that fits too precisely to a specific data set, hence failing to generalize well to unseen observations.

P

Pattern Recognition

A segment of Machine Learning dedicated to the identification of patterns in data, carried out in either a supervised or unsupervised manner.

Polynomial commitment

A polynomial commitment scheme is a cryptographic mechanism that enables a party to publicly pledge to a value or information without disclosing its actual content. Such a scheme permits the validation that a polynomial meets specific criteria without ever unveiling the polynomial's identity, proving especially beneficial for zk-rollups due to the compact nature of the commitment relative to the polynomial. In this scheme, a prover creates a commitment to a polynomial, which can subsequently be opened at any designated point to demonstrate that the polynomial's value at that spot aligns with a specified value. Once the prover shares this commitment value, represented by an elliptic curve point, with a verifier, they are bound to that polynomial. They can only generate valid proofs for that specific polynomial, and any deceptive attempts will lead to the prover's inability to present a proof or the verifier's rejection of the proof.

Pooling (Max Pooling)

The technique of downsizing a matrix, produced by a convolutional layer, to a smaller one.

Personally Identifiable Information

Any information that can independently or in conjunction with other information identify a specific individual.

Precision

The ratio of correctly predicted positive outcomes to the total number of positive outcomes predicted by a classifier.

Prediction

The output produced by a trained model when provided with an input instance.

Preprocessing

The process of converting raw data into a more digestible format.

Pre-trained Model

A model, or a part of it, which has been pre-trained, generally utilizing a different dataset. Also, see Transfer Learning.

Principal Component Analysis

A process that uses orthogonal transformation to convert a set of possibly correlated variables into a set of linearly uncorrelated variables, known as principal components.

Prior

The probability distribution that encapsulates the preexisting beliefs about a specific variable before incorporating new evidence.

Private Key

A unique alphanumeric code that validates a person's claim over a particular cryptocurrency wallet, enabling them to access and manage the assets within.

Public Key

A string of characters utilized to generate an address allowing a cryptocurrency wallet to accept transactions.

Public Ledger

A globally accessible, transparent distributed digital record of transactions that anyone can download.

R

Random Forest

An ensemble learning method that constructs numerous decision trees during the training phase and outputs a combined version (like the mean or mode) of the individual tree results.

Recall

The proportion of all pertinent samples that are accurately classified as positive.

Rectified Linear Unit

A unit that uses the rectifier function as its activation function.

Recurrent Neural Networks

A type of Artificial Neural Network where neuron connections form a directed graph along a sequence. This configuration allows it to demonstrate dynamic temporal behavior for a time sequence and process sequential signals using their internal state (memory).

Regression (Linear Regression, Logistic Regression)

A collection of statistical methods for determining the relationships among variables.

Linear Regression: A straightforward type of regression that uses a linear combination of features as input and outputs a continuous value.

Logistic Regression: A type of regression that generates a probability for each possible discrete label value in a classification problem by applying a sigmoid function to a linear prediction.

Regressor

A feature, or an explanatory variable, used as input to a model.

Regularization

The process of adding extra information to avoid overfitting.

Reinforcement Learning

A branch of Machine Learning, inspired by human behavior, which studies how an agent should act in a given environment to maximize some concept of cumulative reward.

Reproducibility (Crisis of)

A methodological issue in science where researchers find it challenging or impossible to replicate or reproduce the results of many scientific studies in subsequent investigations, either by independent researchers or by the original researchers themselves.

Restricted Boltzmann Machines

A Restricted Boltzmann Machine (RBM) is a generative stochastic artificial neural network that can learn a probability distribution over its set of inputs.

S

Semi-Supervised Learning

A variant of supervised learning methods that also utilize unlabeled data for training. This typically involves a small amount of labeled instances combined with a larger volume of unlabeled data. Also, see Supervised Learning and Unsupervised Learning.

Sentiment Analysis

The application of natural language processing, text analysis, computational linguistics, and biometrics to systematically recognize, extract, measure, and examine affective states and subjective information.

Sharding

A term that refers to the division of a blockchain into multiple identical chains, known as "shards." These shards simultaneously execute transactions and smart contracts, enhancing a network's efficiency and scalability.

Smart Contract

A unique kind of computer program designed to automatically execute a transaction when a set condition is met.

Speech Recognition

see Automated Speech Recognition.

Statistical Distribution

In statistics, an empirical distribution function corresponds to the distribution function associated with the empirical measure of a sample. This step function jumps up by 1/n at each of the n data points. Its value at any specified value of the measured variable is the fraction of observations of the measured variable that are less than or equal to the specified value.

Stablecoin

A category of digital assets designed to maintain a constant value by pegging it to another asset, like gold or the U.S. dollar.

Support Vector Machines (SVM)

A type of discriminative classifiers formally defined by a separating hyperplane. Given labeled training data, the algorithm outputs an optimal hyperplane that categorizes new examples.

Supervised Learning

The task in Machine Learning of formulating a function that maps an input to an output based on example input-output pairs.

Sybil attack

A Sybil attack is a type of security threat in peer-to-peer (P2P) networks where a single adversary controls multiple nodes in the network, primarily to subvert its functionality or achieve a malicious outcome. By creating a large number of pseudonymous identities, the attacker can influence the network or gain a disproportionately large influence within it.

For instance, in networks that rely on redundancy and trust, having many nodes controlled by one entity can undermine mechanisms that are based on redundancy, reputation, or voting.

One of the most common contexts where Sybil attacks are discussed is in blockchain and cryptocurrency networks. For example, if an attacker can control a majority of nodes, they might be able to double-spend coins or prevent certain transactions from being verified. To defend against Sybil attacks, many networks implement mechanisms like requiring some form of validation for participation, such as solving cryptographic puzzles (Proof of Work) or proving ownership of certain assets (Proof of Stake).

Synthetic Data

Data artificially generated when real data cannot be gathered in sufficient quantities or when the original data does not satisfy certain requirements.

T

TensorFlow

A widely-used open-source library within the Machine Learning community for data flow programming across numerous tasks. It serves as a symbolic math library and is also utilized for machine learning applications such as neural networks.

Testnet

A testnet in the context of blockchain is a parallel network created explicitly for testing and developmental purposes. It operates similarly to the main blockchain but doesn't carry real value, ensuring that there's no financial risk involved in the trials. Testnet tokens, distinct from mainnet tokens, are valueless and can often be acquired for free from specific sources known as faucets. This environment facilitates developers in experimenting, refining, and receiving feedback on new features, applications, or upgrades without compromising the stability or security of the primary blockchain network.

Testing Data

A subset of the total data that a data scientist selected for the testing phase of model development.

Time Series Data

A series of data points listed at specific times and indexed in accordance with their occurrence order.

Token

A type of cryptocurrency. Tokens represent units of value issued by platforms that are built on existing blockchains.

Topic Modeling

A type of Unsupervised Machine Learning algorithm that employs clustering to uncover hidden structures in text data, interpreting them as topics.

Training Data

In the context of Supervised Machine Learning, it involves the development of algorithms that can learn from and make predictions on data.

Transfer Learning

A branch of Machine Learning focusing on the application of knowledge acquired while solving one problem to a different, yet related problem.

Turing Test

A test designed by Alan Turing to assess a machine's capability to exhibit intelligent behavior equivalent to or indistinguishable from human intelligence. The test involves the machine chatting with a human. If a human evaluator can't reliably distinguish between the machine and the human, the machine is considered to have passed the Turing test.

Type I Error

see False Positive

Type II Error

see False Negative

U

Underfitting

A situation where a Machine Learning algorithm is unable to accurately capture the underlying data structure, typically because the model is either not complex enough or unsuitable for the task. It's the opposite of Overfitting.

Unsupervised Learning

An area of Machine Learning that involves inferring a function that describes the structure of unlabeled data.

V

Validation

The procedure of using hold-out data to evaluate a trained model's performance. Unlike the testing phase used for the final assessment of the model’s performance, the validation phase is used to determine if any modifications need to be made to the model iteratively.

Vanishing/Exploding Gradients

A challenging issue encountered when training Artificial Neural Networks with gradient-based learning methods and backpropagation. It is due to the neural network’s weights receiving an update proportional to the partial derivative.

Variance

A form of error that arises from the model's sensitivity to small fluctuations in the training set. It is calculated as the expected value of the squared deviation of a random variable from its average.

Volatility

A term used to illustrate the degree of fluctuation in a crypto token's price relative to its average over a certain period.

W

Wallet

Either hardware or software that securely stores a user’s private and public key pair and facilitates interaction with a blockchain network.

Warrant Canary

A warrant canary is a method used by organizations to inform users indirectly about the receipt of government or law enforcement requests for data, which may come with non-disclosure requirements. It's essentially a regularly updated statement affirming that no such secretive requests have been received. If the statement disappears or isn't updated, it can act as a silent alarm, suggesting that the organization might have been served with a request. The idea is that while legal orders can compel silence about a request, they cannot force an entity to lie or continue publishing a false statement.

Whale

A term used to characterize investors who possess substantial amounts of specific cryptocurrencies.

Z

Zero Knowledge Proof

Or zk-proof, is a cryptographic technique enabling an individual to affirm the truth of a statement without revealing any supplementary details. For such a proof to be effective, it must satisfy two primary conditions: completeness and soundness. Completeness implies that the prover can convincingly show their knowledge of the pertinent data, while soundness ensures the verifier can dependably ascertain whether the prover genuinely possesses the said information. Significantly, true zero-knowledge is attained when both these criteria are met without any direct exchange of the core data between the prover and verifier. Beyond privacy implications, zk-proofs are valuable in scaling blockchain networks via rollups, as they minimize the data passed to foundational layers. Essentially, it's the evidence that a robust computational system offers for trustworthy calculations.

zk-SNARK

(Zero-Knowledge Succinct Non-Interactive Argument of Knowledge) Zero-knowledge proofs enable a prover to demonstrate to a verifier that a statement is valid, without divulging any information other than the statement's truthfulness. For instance, using the hash of a random number, the prover can assure the verifier of the existence of such a number without disclosing it. In the extended zero-knowledge "Proof of Knowledge," the prover goes a step further, convincing the verifier not only of the number's existence but also their knowledge of it, still without revealing the number. "Succinct" zero-knowledge proofs are incredibly efficient, verifiable in milliseconds and only a few hundred bytes in length, even for extensive program-related statements. Originally, zero-knowledge protocols required multi-round communications between the prover and verifier. "non-interactive" versions consist of a single message from the prover. Before the advent of Halo, the most efficient method to produce such non-interactive, succinct proofs entailed an initial setup phase to create a common reference string for both parties, known as the system's public parameters. A vulnerability exists where anyone with access to the secret randomness used in generating these parameters could fabricate seemingly valid but false proofs. These non-interactive zero-knowledge proofs, termed zk-SNARKs, allow the verification of vast transaction volumes in consistent time, presenting significant scaling benefits.