Hardware-Agnostic Modeling of Quantum Side-Channel Leakage via Conditional Dynamics and Learning from Full Correlation Data

summary

Video file (mp4)

The gist

The gist The study investigates a sequential coherent side-channel model where an adversarial probe qubit interacts with a target qubit during hidden gate sequences, yielding an empirical

In short

The study models a sequential side-channel where an adversarial probe qubit interacts with a target qubit during unknown gate sequences. It found that retaining the full joint histogram of probe bitstrings is crucial for accurate gate recovery, significantly outperforming marginal statistics. This suggests that analyzing the complete temporal correlation structure provides a much stronger signal for post-processing than looking at individual time steps alone.

Key concepts

Full-Correlation Record
This is the joint histogram of all possible probe bitstrings observed over many shots during a sequence of gates. It captures the entire history and correlation between successive interactions with the probe qubit, serving as a rich statistic for decoding.
Sequential Probe Framework
This model describes Alice applying an unknown gate sequence to a target qubit, followed by repeated interactions with an adversarial probe qubit that is measured mid-circuit. The record is history-dependent because the statistics at any point depend on the gates applied previously.
Marginal Statistics vs. Full Correlation
Marginal statistics only track information from single time steps, which degrades quickly with depth and noise. In contrast, retaining the full joint histogram allows for much better strict sequence recovery because it preserves the necessary correlations between all gate operations.
Amortized Decoding via TCN
A deep learning approach using a Temporal Convolutional Network (TCN) is used to decode the full correlation data. This network processes the entire history as a sequence, allowing it to predict gate labels at each step efficiently without needing to retrain for every specific physical setting.

Terminology used across episodes

This episode discusses

The paper

Hardware-Agnostic Modeling of Quantum Side-Channel Leakage via Conditional Dynamics and Learning from Full Correlation Data · Read on arXiv

Brennan Bell, Andreas Trugler, Konstantin Beyer, Paul Erker

University of Graz · Stevens Institute of Technology · Technische Universitat Wien

DOI: 10.1109/qCCL70331.2026.11668566

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: Today's paper: "Hardware-Agnostic Modeling of Quantum Side-Channel Leakage via Conditional Dynamics and Learning from Full Correlation Data".

Mira: The gist The study investigates a sequential coherent side-channel model where an adversarial probe qubit interacts with a target qubit during hidden gate sequences,

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So to recap where we are, this paper is about modeling a sequential coherent side-channel where an adversarial probe qubit interacts with a target qubit during some hidden gate sequence Brennan Bell et al. claim that repeating the same hidden sequence for N shots yields an empirical full-correlation record, which they call Pbg(b) over probe bit-strings b in zero one k <ref:2602.15966#pg1,repeating the same hidden sequence for N shots yields an empirical full>.

Mira: They argue this full-correlation record is a sufficient statistic for classical post-processing under identically and independently distributed shots but that it grows exponentially with circuit depth. They first describe the sequential probe framework in a coupling- and measurement-agnostic form, emphasizing the scaling of the observation space and why exact analytic distinguishability becomes intractable as you add more circuit depth.

Lev: So they set up this scenario where Eve couples a probe through a gate V theta after each hidden gate gi, and measures it mid-circuit to get that probe bit-string over depth k. Repeating N times gives the histogram Pbg(b).

Kai: They then show that for this specific case—controlled-rotation coupling with fixed projective readout and a commuting Rx gate alphabet—they can derive a depth-dependent leakage envelope and find its maximum, theta star k in Eq. (twenty-three) for the coupling strength that maximizes distinguishability assuming Eve only uses marginal statistics <ref:2602.15966#pg2,23) for the coupling strength that maximizes distinguishability>.

Mira: But they also introduce an operational decoder, which is a neural model, that predicts gates from Eve’s observed statistics as a single parameter conditioned map from Pbg to Alice's per-step gate labels, and this works across different coupling and noise settings without you needing to retrain it.

Conclusion: Kai: Looking at the title Hardware-Agnostic Modeling of Quantum Side-Channel Leakage via Conditional Dynamics and Learning from Full Correlation Data, I think it points to a framework that tries to model side channels without being tied to specific hardware implementation details.

Mira: Exactly. It's not just about power traces or electromagnetic signals; it’s about analyzing the classical channel generated by coherent, history-dependent quantum interactions. The novelty lies in treating this temporal correlation structure Pbg(b) rather than conventional side-channel analysis.

Lev: And what this means for hardware is that we can develop operational constraints on how much information Eve can extract just by looking at these correlation statistics, even when the underlying physics are complex and time-dependent.

Kai: It suggests that the risk of a quantum side-channel depends jointly on coupling strength, noise levels, shot count, circuit depth, and how you choose to represent Eve’s measurement record.

Mira: So operationally, this paper gives us tools to distinguish between guessing individual gates from marginal data versus trying to reconstruct the full sequence using the full correlation data.

Lev: It's about providing a way to measure learnability that't independent of whether you have a perfect estimator for the underlying quantum state, which is pretty important when we're dealing with noisy systems.

More episodes

← Home