Spectrum Sensing

The most widely implemented sensing method due to its low computational complexity and lack of need for prior signal knowledge. A radiometer measures the energy in a frequency band over a sensing interval and compares it against a noise-derived threshold.

• Mechanism: Squared magnitude of FFT samples averaged over time
• Key weakness: Poor performance below the SNR wall, where noise uncertainty renders reliable detection impossible
• Complexity: O(N) — simplest of all techniques
• Use case: Quick coarse sensing in high-SNR environments, often as a first-stage trigger for more refined methods

The optimal detection method when full prior knowledge of the primary user's signal is available. A matched filter correlates the received signal with a known template, maximizing the signal-to-noise ratio at the decision instant.

• Requires: Exact knowledge of pilot patterns, modulation scheme, pulse shaping, and frame structure
• Performance: Achieves detection at extremely low SNRs with minimal sensing time
• Limitation: Requires a dedicated receiver for each signal type; impractical for unknown or encrypted waveforms
• Application: Detecting known broadcast signals like ATSC pilots or LTE cell-specific reference signals

Exploits the periodic statistical properties inherent in modulated signals to distinguish them from stationary noise. Most communication signals exhibit cyclostationarity due to carrier frequencies, symbol rates, and cyclic prefixes.

• Mechanism: Computes the cyclic autocorrelation function or spectral correlation density to identify unique cycle frequencies
• Key advantage: Robust at very low SNRs where energy detection fails; can differentiate between signal types
• Cost: Significantly higher computational complexity — O(N²) for spectral correlation function
• Use case: Military and regulatory spectrum monitoring requiring signal classification alongside detection

A blind sensing technique that uses random matrix theory to analyze the eigenvalues of the received signal's sample covariance matrix. No prior knowledge of signal, noise, or channel is required.

• Methods: Maximum-minimum eigenvalue (MME) ratio test, energy with minimum eigenvalue (EME), and largest eigenvalue test
• Robustness: Immune to noise uncertainty, eliminating the SNR wall problem that plagues energy detection
• Trade-off: Requires multiple antenna elements or oversampling; computationally heavier than energy detection
• Application: MIMO-equipped cognitive radios and base stations with antenna arrays

Multiple spatially distributed cognitive radios share their individual sensing observations to collaboratively decide on primary user presence. This architecture directly addresses the hidden node problem caused by shadowing and multipath fading.

• Fusion strategies: Hard combining (AND, OR, K-out-of-N rules) or soft combining (likelihood ratios, energy values)
• Gain: Exponential improvement in detection probability with increasing cooperating nodes under correlated shadowing
• Overhead: Requires a dedicated reporting channel and fusion center; vulnerable to Byzantine attacks from compromised nodes
• Standardization: Central to IEEE 802.22 WRAN and ECMA-392 cognitive radio standards

Leverages supervised and unsupervised learning to classify spectrum occupancy from raw or feature-engineered sensing data. Deep neural networks can learn complex, non-linear decision boundaries that outperform classical threshold-based detectors.

• Architectures: CNNs for spectrogram image classification, LSTMs for temporal occupancy pattern recognition, autoencoders for anomaly-based signal detection
• Training data: Requires extensive labeled datasets of real or synthetic RF captures; GANs for spectrum data augmentation address scarcity
• Advantage: Adapts to dynamic environments without manual threshold recalibration; can jointly perform signal classification and modulation recognition
• Deployment: Increasingly integrated into O-RAN RIC xApps for intelligent, software-defined spectrum awareness

A technique where multiple cognitive radios share their individual sensing observations to collaboratively detect a primary user. This approach directly mitigates the hidden node problem, where a single sensor fails to detect a transmitter due to shadowing or multipath fading. By fusing data from spatially diverse nodes using hard decision combining (OR, AND, majority rules) or soft decision combining (likelihood ratios), the network achieves significantly higher detection probability at low signal-to-noise ratios.

A robust sensing method that exploits the periodic statistical properties inherent in modulated signals. Unlike energy detection, which fails below the noise uncertainty wall, this technique distinguishes modulated transmissions from stationary noise by analyzing the spectral correlation function. Key advantages include:

• Resilience to noise power uncertainty
• Ability to identify the modulation type (BPSK, QPSK, OFDM)
• Superior performance at very low SNR (below -20 dB) The computational cost is higher, making it suitable for fine-grained sensing stages rather than rapid coarse scanning.

A denial-of-service security threat where a malicious actor mimics the signal characteristics of a legitimate primary user to monopolize spectrum resources. The attacker transmits a signal that replicates the primary's modulation type, power level, and spectral mask, causing all secondary users to erroneously vacate the channel. Defenses include:

• RF fingerprinting to identify unique hardware imperfections
• Location verification using RSSI trilateration or angle-of-arrival
• Statistical behavior profiling to detect anomalous transmission patterns PUEA represents one of the most critical threats to the integrity of any dynamic spectrum sharing framework.

An integrated spatio-temporal database that aggregates multi-domain information to provide a comprehensive awareness layer for cognitive radio networks. A REM stores and processes:

• Spectrum occupancy measurements over time and location
• Geolocation data and terrain elevation models
• Propagation models for path loss prediction
• Regulatory policies and incumbent protection zones By fusing sensed data with predictive models, the REM enables cognitive radios to reason about spectrum availability beyond their immediate sensing range, supporting proactive rather than purely reactive access decisions.

The process by which a secondary user seamlessly vacates a channel upon the return of a primary user and transitions its ongoing communication to another vacant frequency band. Unlike traditional handoff triggered by mobility, spectrum handoff is interference-driven. Key design goals include:

• Minimizing handoff latency to prevent connection drops
• Proactive target channel selection using occupancy prediction
• Protocol adaptation to re-establish sessions at the new frequency Effective spectrum handoff is essential for maintaining quality of service in cognitive radio networks where channel availability is inherently transient.

A privacy-preserving distributed machine learning technique where multiple radio nodes collaboratively train a spectrum access model without sharing raw sensing data. Each node trains a local model on its own observations and sends only encrypted model updates (gradients or weights) to a central aggregation server. Benefits include:

• Data sovereignty: raw IQ samples never leave the device
• Reduced backhaul load: only model parameters are transmitted
• Scalability: models improve with more participating nodes This approach is particularly relevant for defense and telecom applications where spectrum data is operationally sensitive or subject to regulatory privacy constraints.