papers/zoo-poai-consensus/sections/02-background.tex

\section{Background}


\subsection{Proof of Useful Work}

Prior attempts at useful PoW include:

\begin{itemize}
    \item \textbf{Primecoin}~\cite{king2013primecoin}: Finding prime number chains
    \item \textbf{Gridcoin}~\cite{gridcoin2014}: BOINC scientific computing
    \item \textbf{FoldingCoin}: Protein folding simulations
\end{itemize}

These approaches suffer from limited useful work domains and difficulty of verification. Our work extends to general AI computation.

\subsection{Trusted Execution Environments}

\TEE{}s provide hardware-enforced isolation:

\begin{definition}[Trusted Execution Environment]
A \TEE{} is a hardware-protected region where:
\begin{enumerate}
    \item Code and data are encrypted in memory
    \item Execution cannot be observed or modified by the host OS
    \item Remote attestation proves what code ran and its output
\end{enumerate}
\end{definition}

Major \TEE{} implementations:
\begin{itemize}
    \item Intel SGX / TDX
    \item AMD SEV-SNP
    \item ARM TrustZone / CCA
    \item NVIDIA Confidential Computing
\end{itemize}

\subsection{BLS Signatures}

BLS signatures enable efficient aggregation:

\begin{definition}[BLS Signature Aggregation]
Given signatures $\{\sigma_i\}_{i=1}^n$ from signers $\{pk_i\}$ on message $m$:
\begin{equation}
\sigma_{\text{agg}} = \prod_{i=1}^n \sigma_i
\end{equation}
Verification: $e(\sigma_{\text{agg}}, g) = e(H(m), \sum_i pk_i)$
\end{definition}

This enables $O(1)$ verification of $n$ signatures.