> ## Documentation Index
> Fetch the complete documentation index at: https://docs.notional.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Risks

## Overview

This page summarizes the protocol’s main risk areas across four buckets:

* Compromised operator or guardian control
* Infrastructure-level attacks on Azure, AMD, or network connectivity
* Stale or manipulated oracle prices
* Operational liveness failures

Responsibilities are intentionally split between the operator, guardians, and exchanges to
limit blast radius when something goes wrong.

## Compromised Operator or Guardian Set

A compromised operator may attempt to produce incorrect state or submit unauthorized trades. A
compromised guardian majority could approve malicious withdrawals or authorize a rogue operator.

### Mitigations

* The operator cannot move funds without guardian approval.
* Guardians can revoke the operator’s trading wallet at any time.
* Canonical state is derived from the Notional ledger and can be independently replayed.
* Guardian compromise requires coordinated collusion rather than a single failure.
* The guardian set is intended to be distributed to reduce concentration risk.

## Infrastructure-Level Attacks

Cloud, hardware, or network failures may enable malicious operator behavior, degrade availability,
or expose secrets.

### Mitigations

* The operator runs inside AMD SEV-SNP on Azure Confidential VMs.
* Azure Secure Key Release ties key access to valid attestation.
* Guardian-controlled withdrawals prevent infrastructure compromise from directly moving funds.
* Canonical state remains auditable through the Notional ledger.
* Long term, Notional plans to diversify across TEE vendors to reduce single-vendor risk.

### Trust Assumptions

Users still rely on a few core assumptions:

* AMD hardware does not contain a backdoor.
* Azure’s attestation service does not forge reports.
* The system is not broken by advanced side-channel attacks.

## Oracle Price Manipulation

Stale or manipulated prices may cause incorrect margin calculations, incorrect liquidation triggers,
or opportunistic withdrawals during volatile markets.

### Mitigations

* The protocol uses Hyperliquid mark prices for perps and spot.
* Notional operator and guardian nodes are expected to run Hyperliquid nodes directly.
* Public Hyperliquid nodes are a fallback if local infrastructure fails.
* The goal is to keep price feeds current even during partial infrastructure failures.

## Operational Liveness Risks

Cloud failures, network partitions, TEE errors, or exchange outages may temporarily halt block
production, trading, or withdrawals.

### Mitigations

* Critical keys are guardian-controlled.
* State is fully replayable from canonical data.
* Backup operators can assume control if needed.
* The network can be safely restarted from canonical ledger state after a failure.

## Responsible Disclosure

If you discover a security vulnerability, please report it to:

**[security@notional.xyz](mailto:security@notional.xyz)**

We follow a coordinated disclosure process and offer bug bounties for qualifying vulnerabilities.