Bilinear Compressive Security

TL;DR

This work introduces Bilinear Compressive Security (BCS), a secure communication scheme in which a randomly drawn per-transmission filter $h$ is convolved with the secret measurement $Qx$ to produce $y=h*Qx$, enabling secure transmission even if $Q$ is reused across rounds. Bob performs blind deconvolution to recover $x$ (and implicitly $h$) from $y$ without knowing $h$, leveraging sparsity in $x$ and $h$; security against an attacker observing multiple $(x_k,y_k)$ pairs is analyzed under favorable filter models. The main theoretical result shows that, when $h$ is phase-symmetric, recovering $Q$ from $M$ transmissions with $s$-sparse $x_k$ requires $M$ to scale as $\max\left(n,(n/s)^2\right)$; in particular, for $s=1$ recovery is impossible, implying a significant security advantage over standard compressed sensing. Empirical experiments corroborate the theory: recovery fails for highly sparse messages, with a transition around $s\approx\sqrt{n}/2$ and a threshold near $M\approx (n/s)^2$, supporting the practicality of BCS in scenarios with sparse signals and variable channels.

Abstract

Beyond its widespread application in signal and image processing, \emph{compressed sensing} principles have been greatly applied to secure information transmission (often termed 'compressive security'). In this scenario, the measurement matrix $Q$ acts as a one time pad encryption key (in complex number domain) which can achieve perfect information-theoretic security together with other benefits such as reduced complexity and energy efficiency particularly useful in IoT. However, unless the matrix is changed for every message it is vulnerable towards known plain text attacks: only $n$ observations suffices to recover a key $Q$ with $n$ columns. In this paper, we invent and analyze a new method (termed 'Bilinear Compressive Security (BCS)') addressing these shortcomings: In addition to the linear encoding of the message $x$ with a matrix $Q$, the sender convolves the resulting vector with a randomly generated filter $h$. Assuming that $h$ and $x$ are sparse, the receiver can then recover $x$ without knowledge of $h$ from $y=h*Qx$ through blind deconvolution. We study a rather idealized known plaintext attack for recovering $Q$ from repeated observations of $y$'s for different, known $x_k$, with varying and unknown $h$ ,giving Eve a number of advantages not present in practice. Our main result for BCS states that under a weak symmetry condition on the filter $h$, recovering $Q$ will require extensive sampling from transmissions of $Ω\left(\max\left(n,(n/s)^2\right)\right)$ messages $x_k$ if they are $s$-sparse. Remarkably, with $s=1$ it is impossible to recover the key. In this way, the scheme is much safer than standard compressed sensing even though our assumptions are much in favor towards a potential attacker.

Figures (3)

• Figure 1: A graphical depiction of BCS. Note that $\mathcal{D}$ and $y$ is public information, but $x$ and $Q$ are secret. The cloud shall indicate a potential wireless channel which corroborates the security properties of BCS.
• Figure 2: Success rate vs. $s$ and $M$ for $n=50$ and $m=5$ (top) and $n=100$ and $m=3$ (bottom). Best viewed in color.
• Figure 3: Success rates in dependence of $s$ and $M$. Both axes are in log-scale.

Theorems & Definitions (11)

• Theorem 1
• Definition 1
• Theorem 2
• proof
• Definition 2
• Definition 3
• Lemma 3
• Remark 1
• Theorem 4
• Theorem 5