Quantum Trojan Insertion: Controlled Activation for Covert Circuit Manipulation
Jayden John, Lakshman Golla, Qian Wang
TL;DR
The paper tackles the threat of untrusted quantum compilers by introducing a controllable quantum Trojan that activates only under predefined input conditions. It details a DAG-based two-phase insertion method that places Trojan gates in empty circuit slots without increasing depth, preserving functionality when deactivated. Experimental results on RevLib benchmarks show minimal depth overhead and substantial alterations to output distributions (TVD ≈ 0.9) in many cases, highlighting covert manipulation risks. The work emphasizes the need for new defense strategies tailored to dynamic, conditionally activated Trojans in quantum circuits and compiler pipelines.
Abstract
Quantum computing has demonstrated superior efficiency compared to classical computing. Quantum circuits are essential for implementing functions and achieving correct computational outcomes. Quantum circuit compilers, which translate high-level quantum operations into hardware-specific gates while optimizing performance, serve as the interface between the quantum software stack and physical quantum machines. However, untrusted compilers can introduce malicious hardware Trojans into quantum circuits, altering their functionality and leading to incorrect results. In the world of classical computing, effective hardware Trojans are a critical threat to integrated circuits. This process often involves stealthily inserting conditional logic gates that activate under specific input conditions. In this paper, we propose a novel advanced quantum Trojan that is controllable, allowing it to be activated or deactivated under different circumstances. These Trojans remain dormant until triggered by predefined input conditions, making detection challenging. Through a series of benchmark experiments, we demonstrate the feasibility of this method by evaluating the effectiveness of embedding controlled trojans in quantum circuits and measuring their impact on circuit performance and security.
