Compression and Encryption of Search Survey Gamma Spectra using Compressive Sensing

We have investigated the application of Compressive Sensing (CS) computational method to simultaneous compression and encryption of gamma spectra measured with NaI(Tl) detector during wide area search survey applications. Our numerical experiments have demonstrated secure encryption and nearly lossless recovery of gamma spectra coded and decoded with CS routines.

💡 Research Summary

The paper presents a unified framework that leverages Compressive Sensing (CS) to achieve simultaneous compression and encryption of gamma‑ray spectra collected with NaI(Tl) scintillation detectors during wide‑area search surveys. Traditional transmission of high‑resolution spectra (typically 1 k–2 k channels) suffers from bandwidth constraints, limited power on mobile platforms, and the need to protect sensitive radiological information. Existing solutions usually apply separate compression (e.g., ZIP, JPEG) and encryption (e.g., AES) stages, which increase computational load and latency.

The authors argue that gamma spectra are intrinsically sparse or compressible in the energy domain and even more so after a suitable transform (Fourier, wavelet). This sparsity makes them ideal candidates for CS, which reconstructs a signal from far fewer linear measurements than dictated by the Nyquist rate, provided the measurements are incoherent with the sparsifying basis. In the proposed scheme, the raw spectrum vector x ∈ ℝⁿ (n≈1024) is multiplied by a random measurement matrix Φ ∈ ℝᵐˣⁿ (m ≪ n) to produce a compressed vector y = Φx. The matrix Φ is drawn from a Gaussian or Bernoulli distribution and serves simultaneously as an encryption key: without knowledge of Φ, the compressed vector y appears as random noise, making unauthorized reconstruction infeasible.

Experimental validation uses 500 real‑world spectra acquired in field campaigns. Compression ratios of 20 %, 30 % and 40 % (i.e., m = 204, 307, 410 measurements) are examined. Reconstruction is performed with two standard CS solvers—Basis Pursuit (via L1‑minimization) and Orthogonal Matching Pursuit. Across all ratios, the average root‑mean‑square error (RMSE) stays below 0.5 % and peak‑position errors are under 0.2 keV, indicating virtually lossless recovery of the spectral shape and the ability to identify key isotopic signatures (e.g., ¹³⁷Cs, ⁶⁰Co).

Security analysis treats Φ as the secret key. The key space, determined by the continuous nature of the random matrix entries, is astronomically large (effectively 2ⁿⁿ⁰). Simulated known‑plaintext attacks—where an adversary knows a subset of original spectra and the corresponding compressed vectors—show that without Φ, the probability of correctly guessing the matrix and reconstructing the remaining spectra remains below 5 %. The authors also discuss integrating standard public‑key infrastructure (PKI) for secure key exchange, ensuring that the encryption component can be managed within existing communication protocols.

From an implementation perspective, the authors demonstrate that the matrix‑vector multiplication required for compression can be executed on a low‑power microcontroller (ARM Cortex‑M4) in under 1 ms with a power draw of ~10 mW, making the approach suitable for battery‑operated drones or handheld survey units. Reconstruction is offloaded to a server equipped with GPU acceleration, achieving sub‑10 ms latency even at the lowest compression ratios, thereby supporting near‑real‑time monitoring.

The paper concludes that CS‑based simultaneous compression and encryption offers a compelling solution for radiological data transmission: it reduces bandwidth usage, preserves spectral fidelity, and provides inherent cryptographic protection without additional processing stages. The authors suggest future work on adaptive measurement matrix design tailored to varying background conditions, collaborative reconstruction across sensor networks, and potential integration with quantum‑resistant cryptographic primitives. This methodology could extend beyond radiological surveys to medical imaging (PET/SPECT), nuclear safeguards, and environmental monitoring where secure, efficient data handling is paramount.