Could attackers combine pulse-level circuit attacks with crosstalk sabotage to amplify damage across multiple victim circuits simultaneously?

# Can Combined Pulse-Level and Crosstalk Attacks Amplify Damage?

Yes—the evidence suggests these attack vectors could compound each other's effects, though the field is still mapping exactly how.

The Two Attack Surfaces

Pulse-level attacks exploit a critical gap in quantum software: high-level gate instructions don't always match their low-level pulse implementations [AG-2024.06-1682]. An attacker who controls the pulse specification can inject seven distinct attack types—qubit plunder, timing mismatches, frequency mismatches, and others—without the gate-level description betraying the sabotage. This is powerful because most current SDKs are vulnerable to this inconsistency.

Crosstalk attacks exploit unwanted interference between neighboring qubits in shared cloud quantum computers [AG-2024.02-1368]. When an adversary designs a circuit to deliberately induce crosstalk noise, they can disrupt computations on nearby qubits that belong to other users. The effect is measurable: crosstalk is a significant source of errors on real IBM hardware, and simple scheduling changes can improve accuracy by 3× [AG-2024.02-1554].

The Amplification Risk

A combined attack becomes plausible because:

1. Spatial coupling: Pulse-level manipulations (e.g., frequency mismatch [AG-2024.06-1682]) could be designed to *deliberately maximize* crosstalk to adjacent victim qubits, turning a single-qubit vulnerability into a multi-qubit cascade.

2. Detection evasion: Crosstalk naturally produces correlated errors on physically nearby qubits [AG-2024.05-2051], so an attacker exploiting both vectors simultaneously could hide pulse-level tampering within the "noise signature" of legitimate crosstalk.

3. Shared environment: Cloud quantum computers allow multiple users on the same hardware [AG-2024.02-1368], so an attacker controlling one circuit can craft pulse-level instructions *and* spatially target victim qubits simultaneously.

The preprints provide evidence that both attack types work independently; combining them for amplification is a logical next step, but the archive doesn't directly analyze a hybrid attack.