Security

Space implements multiple layers of risk controls to protect users and maintain market integrity.

​

Oracle Security

Multi-Source Validation:

• Median of Pyth + Switchboard prices
• Reject if oracles disagree by >2%
• 15-minute TWAP averages out manipulation

Liveness Checks:

• Require minimum updates in TWAP window
• Extend window if oracle temporarily down
• Mark invalid if insufficient data

Price Clamping:

• Maximum 1% price movement per minute
• Outlier detection: reject prices >3 standard deviations
• Multi-feed median prevents single-source manipulation

​

Circuit Breakers

Markets automatically pause if price moves:

• 10% in <60 seconds
• 25% in <5 minutes

On Trigger: Prevents flash crashes and manipulation attempts.

​

Liquidation Controls

Position Value < Maintenance Margin → Liquidation Liquidation Process: Insurance Fund:

• Funded by liquidation penalties (10% of liquidated value)
• Plus 5% of all platform fees
• Covers negative balances from latency/congestion
• Balance publicly visible on-chain
• Target reserve: 2% of platform TVL The insurance fund protects the platform from user accounts going negative due to rapid price movements or network congestion preventing timely liquidation.

​

Pre-Event Freeze

Markets can implement freeze mechanisms before resolution: Option 1: Fee Escalation Time to resolution < 10 minutes: $\text{Taker Fee} \times 3$ Time to resolution < 5 minutes: $\text{Taker Fee} \times 5$ Option 2: Position Freeze Time to resolution < 5 minutes:

• New positions: DISABLED
• Exit positions: ENABLED

Prevents last-second manipulation with insider information or unfair advantage from faster information access.

​

Slippage Protection

Users set maximum acceptable slippage on market orders (default 5%). If $\frac{|\text{Actual Fill Price} - \text{Expected Price}|}{ \text{Expected Price}} > \text{Max Slippage}$: → Revert transaction → No trade executed → No fees charged Protects users from unexpected execution prices in volatile or thin markets.

​

Platform Upgrades

Timelock Requirements:

• Parameter changes: 48-hour timelock
• Critical code changes: 72-hour timelock

Multisig Control:

• 3-of-5 multisig for admin functions
• 2-of-3 multisig for resolution submissions
• 2-of-5 multisig for emergency pause (faster response)

All governance actions are transparent and logged on-chain.

​

Future: zkML & zkTLS Research

Space is researching zero-knowledge proofs for an even more cutting-edge, trustless subjective resolution yet to be utilized within prediction markets: zkML (Zero-Knowledge Machine Learning):

• Prove AI model output without revealing model weights
• On-chain verification of resolution logic
• No trust required in resolution operators

zkTLS (Zero-Knowledge Transport Layer Security):

• Prove data was fetched from specific URL
• Cryptographic proof of source authenticity
• Combine with zkML for fully automated resolution

Combined Architecture: Fetch news article via zkTLS → Process with zkML → Submit proof on-chain → Automatic resolution with cryptographic guarantees.