Probability of False Alarm

Formally denoted as P_fa, the probability of false alarm is the conditional probability that the test statistic exceeds the detection threshold given the null hypothesis (H₀: signal absent). It is calculated as the integral of the probability density function of the test statistic under H₀ from the threshold to infinity. In energy detection over AWGN channels, P_fa is expressed using the Q-function or the complementary cumulative distribution function of a chi-squared distribution.

A high P_fa directly reduces secondary user throughput by causing the cognitive radio to erroneously back off from vacant spectrum. The fundamental tradeoff is:

• Low threshold → High P_d (good primary user protection) but also high P_fa (wasted transmission opportunities)
• High threshold → Low P_fa (more access) but also low P_d (increased interference risk) This is visualized on the Receiver Operating Characteristic (ROC) curve, which plots P_d against P_fa.

CFAR algorithms dynamically adapt the detection threshold to maintain a fixed, pre-defined P_fa despite fluctuations in ambient noise power. This is critical because noise uncertainty in practical receivers makes a static threshold unreliable. Common CFAR techniques include:

• Cell-Averaging CFAR: Estimates local noise power by averaging neighboring range bins or frequency cells
• Ordered-Statistic CFAR: Uses the k-th ordered sample to estimate noise, more robust in multi-target environments

In cooperative spectrum sensing, the global probability of false alarm (Q_fa) is a function of the local P_fa at each node and the fusion rule applied at the fusion center. For the K-out-of-N rule:

• A higher K value reduces Q_fa (more conservative) but also reduces the global probability of detection
• Soft decision fusion generally achieves a lower Q_fa for a given Q_d compared to hard decision fusion, as it preserves more information from the local test statistics
• Spectrum Sensing Data Falsification (SSDF) attacks can artificially inflate Q_fa, causing denial-of-service

The Neyman-Pearson (NP) criterion provides the optimal detection framework for spectrum sensing. It formulates the problem as:

• Maximize the probability of detection (P_d)
• Subject to a constraint that P_fa ≤ α, where α is the maximum tolerable false alarm rate

The NP lemma proves that the Likelihood Ratio Test (LRT) is the most powerful test for this constrained optimization. In practice, the LRT requires channel state information that is often unavailable, leading to suboptimal but practical alternatives like energy detection.

Noise uncertainty—the inherent imprecision in estimating ambient noise power—creates a fundamental limit on detection performance. Below a certain SNR wall, it becomes impossible to simultaneously achieve a desired P_d and P_fa, regardless of sensing duration. This phenomenon is particularly severe for energy detection, where a 1 dB noise uncertainty can require an SNR increase of several dB to maintain the same P_fa. Cyclostationary feature detection and eigenvalue-based blind sensing are robust alternatives that mitigate this limitation.

The statistical likelihood that a spectrum sensing algorithm correctly identifies a primary user signal when it is actually transmitting. This metric quantifies the primary user protection level. A high probability of detection is mandated by regulators to prevent harmful interference.

• Directly trades off against the Probability of False Alarm.
• The Neyman-Pearson Criterion formally maximizes this metric subject to a fixed false alarm constraint.
• A value of 0.99 means the algorithm misses only 1% of active transmissions.

A graphical plot illustrating the diagnostic ability of a binary classifier. For spectrum sensing, the ROC curve plots the Probability of Detection against the Probability of False Alarm as the detection threshold varies.

• The primary metric for evaluating and comparing sensing algorithms.
• The area under the ROC curve (AUC) summarizes overall performance; an AUC of 1.0 represents perfect classification.
• A steeper curve indicates a better trade-off, achieving high detection rates with very low false alarm rates.

An adaptive threshold-setting algorithm that dynamically adjusts the detection threshold based on estimated noise power. The goal is to maintain a fixed, pre-defined Probability of False Alarm despite fluctuations in the ambient noise floor.

• Essential for practical energy detection where noise power is non-stationary.
• Cell-Averaging CFAR estimates local noise by averaging neighboring range or frequency bins.
• Prevents the receiver from being overwhelmed by false positives in noisy environments.

An optimal detection framework that maximizes the Probability of Detection subject to an upper bound constraint on the Probability of False Alarm. This forms the theoretical basis for many spectrum sensing fusion rules.

• The solution is a Likelihood Ratio Test (LRT) that compares the ratio of probability density functions under the signal-present and signal-absent hypotheses.
• Provides a rigorous mathematical justification for setting a false alarm constraint as the primary design parameter.

The fundamental design conflict in cognitive radio where longer sensing durations improve detection accuracy (lowering both false alarm and miss probabilities) but reduce the time available for data transmission.

• A high Probability of False Alarm directly wastes transmission opportunities, reducing secondary user throughput.
• The optimal sensing time is found by maximizing the achievable throughput subject to a target detection probability.
• Frame structures must balance a sensing slot with a data transmission slot.

A hard decision fusion rule used in cooperative sensing. The fusion center declares a primary user present if at least K out of N cooperating sensing nodes report a positive detection.

• The global Probability of False Alarm is a function of the individual node false alarm probabilities and the chosen K value.
• OR Rule (K=1): Maximizes detection but also maximizes the global false alarm rate.
• AND Rule (K=N): Minimizes false alarms but degrades detection sensitivity.
• The majority rule (K=N/2) provides a balanced trade-off.