Papertrade: Unguarded BBO Midpoint Oracle Yields $1.28M Manipulation Attack on Launch Day

A Hyperliquid veteran exploited Papertrade's use of Hyperliquid's live best-bid-offer midpoint as its sole, unprotected price oracle on the protocol's first day, simultaneously moving the underlying spot price to profit on Papertrade positions across 296 of 305 trades for $1.28M.

Root Cause

Papertrade, a synthetic perpetuals venue on HyperEVM (Hyperliquid L1), settled all positions against Hyperliquid's live BBO (best-bid-offer) midpoint — a single-block spot-market quote — with no manipulation-resistance mechanisms:

No time-weighted average price (TWAP) or median lookback
No maximum single-block deviation circuit-breaker
No open-interest cap relative to oracle liquidity depth
Oracle and trading venue share the same underlying market (reflexive risk)

// Simplified vulnerable oracle pattern
function getSettlementPrice() external view returns (uint256) {
    // Returns Hyperliquid BBO midpoint at this exact block —
    // no TWAP, no staleness check, no deviation guard
    return hyperliquidOracle.getSpotMidpoint(ASSET);
}

function closePosition(uint256 positionId) external {
    uint256 exitPrice = getSettlementPrice();  // live midpoint — manipulatable
    _settle(msg.sender, positionId, exitPrice);
}

Because an actor with sufficient Hyperliquid capital can widen the BBO or move the inside quote for a single block, and because Papertrade entry and exit both read this raw midpoint, the attacker could arbitrarily set their entry and exit prices on Papertrade by moving the Hyperliquid market at the right moment.

Attack Steps

Step	Hyperliquid Action	Papertrade Action	Direction
1	Buy heavily, move mid-price up	Open large long just before buy	Long
2	Mid-price rises	Close long at inflated midpoint	Profit
3	Sell heavily, move mid-price down	Open large short just before sell	Short
4	Mid-price falls	Close short at deflated midpoint	Profit
5–N	Repeat ~305 times over ~8 hours	296 wins, 9 losses	$1.28M net

The attacker's final 30 shorts were liquidated (~$450K loss), capping total profit at ~$1.28M. Some reports speculate the team manually intervened to block the final batch of closes.

Impact

Attacker Net Gain: ~$1.28M
House Pool Loss: Protocol's liquidity pool absorbed the losses
Date: October 10, 2026 (within first ~8 hours of mainnet launch)
Chain: HyperEVM (Hyperliquid L1)
Attacker Profile: Pre-existing Hyperliquid trader with $14B all-time volume, $13M+ PnL — capital sufficient to move the oracle
Response: Papertrade tightened open-interest caps post-incident; no contract upgrade announced

Lessons for Auditors

Single-block spot midpoints are manipulable by capitalized actors. Any oracle derived from a live order-book midpoint without TWAP or lookback is exploitable by anyone who can move the underlying market temporarily. Use a TWAP of at least several minutes, or a multi-source aggregate.
Bound OI relative to oracle manipulation cost. Maximum position size on a synthetic venue should be set so the expected profit from moving the oracle by X% is less than the cost of that manipulation (i.e., slippage + fees on the underlying market). Audit whether this relationship holds at launch-day liquidity levels.
Reflexive oracles amplify risk. When the settlement oracle is the same venue where the attacker trades to manipulate prices, the attacker extracts profit from the synthetic protocol while the manipulation cost is partly offset by their own Hyperliquid P&L. Prefer oracles on markets the attacker cannot profitably move in isolation.
Launch-day OI limits must account for thin liquidity. Protocols should constrain OI much more aggressively at launch, scaling up only as oracle depth (liquidity on the source market) grows. A staggered OI ceiling tied to VWAP depth metrics is one mitigation.