Fractional Equivocation in Partial Secrecy

• Fractional Equivocation is a quantitative metric that normalizes the uncertainty remaining about a secret message after eavesdropper observation.
• It generalizes perfect secrecy by linking operational decoding error bounds with rate–distortion theory via log-loss, providing actionable insights.
• The metric is applied to optimize system design in wireless networks through adaptive power allocation and refined secrecy–reliability trade-offs.

Fractional equivocation is a quantitative metric for partial secrecy in information-theoretic security, expressing the normalized uncertainty that remains about a secret message given the eavesdropper’s observation. It generalizes the classical notion of perfect secrecy and enables refined analysis of secrecy systems operating in the partial secrecy regime, where complete secrecy may not be achievable or required. Fractional equivocation possesses both a precise operational interpretation—as a lower bound on the eavesdropper’s decoding error probability—and a direct relationship with rate–distortion theory under log-loss distortion. This metric supports system design and optimization for secrecy in modern networks, particularly in wireless channels with fading, untrusted relays, and varying power allocation schemes (Osorio et al., 2019, Cuff, 2013).

1. Mathematical Definition and Interpretation

Fractional equivocation, denoted $\Theta$, is defined as

$\Theta = \frac{H(M|Z^n)}{H(M)},$

where $M$ is the confidential message, $Z^n$ is the eavesdropper’s entire observation over $n$ channel uses, and $H(\cdot)$ denotes entropy. Equivalently,

$\Theta = 1 - \frac{I(M; Z^n)}{H(M)},$

with $I(M; Z^n)$ the mutual information between the message and the eavesdropper’s observation (Osorio et al., 2019).

Fractional equivocation takes values in $[0,1]$, quantifying the fraction of the message’s entropy that remains hidden from the eavesdropper. In the large block length limit, if $\Delta$ denotes the fractional equivocation, the eavesdropper’s decoding error probability is lower bounded as

$\Theta = \frac{H(M|Z^n)}{H(M)},$0

so $\Theta = \frac{H(M|Z^n)}{H(M)},$1 ensures nearly perfect secrecy, while $\Theta = \frac{H(M|Z^n)}{H(M)},$2 reflects partial confusion (Osorio et al., 2019).

2. Connection to Distortion-Based Secrecy and Log-Loss

Fractional equivocation emerges as a special case in the distortion-based secrecy framework when the distortion metric is chosen as log-loss (Cuff, 2013). If $\Theta = \frac{H(M|Z^n)}{H(M)},$3 is the source and $\Theta = \frac{H(M|Z^n)}{H(M)},$4 the eavesdropper's side information,

$\Theta = \frac{H(M|Z^n)}{H(M)},$5

The normalized or fractional equivocation is

$\Theta = \frac{H(M|Z^n)}{H(M)},$6

which precisely matches the minimum expected log-loss endured by an adversary outputting the posterior PMF of the source (Cuff, 2013). Thus, equivocation can be seen as the minimum expected log-loss, and fractional equivocation normalizes this by the source entropy.

3. Operational Role in the Partial Secrecy Regime

In scenarios where some information leakage is tolerable, fractional equivocation quantifies the operational level of secrecy. For wiretap coding with codeword rate $\Theta = \frac{H(M|Z^n)}{H(M)},$7 and secrecy rate $\Theta = \frac{H(M|Z^n)}{H(M)},$8, and defining the instantaneous SNRs/SINRs $\Theta = \frac{H(M|Z^n)}{H(M)},$9 (legitimate receiver) and $M$0 (eavesdropper), the auxiliary random variable $M$1 is

$M$2

The instantaneous fractional equivocation $M$3 follows a piecewise-linear rule: $M$4 This metric reflects, for each channel realization, the minimum uncertainty from the eavesdropper’s perspective (Osorio et al., 2019).

The average fractional equivocation aggregates performance over channel variations: $M$5 where $M$6 is the density of $M$7 (Osorio et al., 2019).

4. Relationship to Information Leakage and System Design

Fractional equivocation serves as a core parameter for defining and analyzing alternative secrecy metrics. The average information leakage rate is

$M$8

capturing the average confidential bits per channel use leaked to the eavesdropper (Osorio et al., 2019).

Power allocation directly impacts fractional equivocation. Two principal strategies are:

• Equal Power Allocation (EPA): $M$9 for source, relay, and jamming powers,
• Optimal Power Allocation (OPA): computed to maximize $Z^n$0, minimize $Z^n$1, or achieve other secrecy-related targets subject to throughput constraints $Z^n$2 (Osorio et al., 2019).

Optimization of these parameters provides a tunable secrecy–reliability trade-off. For instance, lowering the threshold $Z^n$3 in secrecy outage metrics or optimizing $Z^n$4 under a rate constraint can adapt transmit/jamming powers and code rates to achieve specific service requirements—from partial secrecy to nearly perfect secrecy.

5. Relationship with Rate–Distortion Theory

Fractional equivocation fits naturally within the rate–distortion paradigm under log-loss distortion (Cuff, 2013). The rate–distortion function for source $Z^n$5 under distortion $Z^n$6 is

$Z^n$7

When $Z^n$8 is log-loss, $Z^n$9. With $n$0 as the public message rate and $n$1 as the secret-key rate, the adversary’s maximal equivocation is

$n$2

where $n$3. Normalization by $n$4 yields

$n$5

This formula interpolates between 0 and 1 depending on the resources allocated to secrecy and public rate (Cuff, 2013).

6. Practical Implications and Power Allocation Scenarios

The adoption of fractional equivocation allows system designers to balance secrecy and throughput efficiently. In wireless networks involving fading channels and untrusted relays, destination-based jamming (DBJ) and adaptive power allocation can significantly increase fractional equivocation relative to naive equal power schemes. Numerical methods, such as particle swarm optimization, are effective for finding optimal power allocation vectors and secrecy rates in these settings (Osorio et al., 2019).

Analysis demonstrates that OPA enhances $n$6 and reduces $n$7, particularly in regimes where partial secrecy is sufficient, and full secrecy is unnecessarily stringent. This aligns system-level metrics with application-level secrecy requirements.

7. Causal Disclosure and the Justification for Fractional Equivocation

In the distortion-based secrecy framework, the role of causal disclosure is to prevent artificial secrecy gains that could arise from withholding context rather than true encryption. Under causal disclosure, where the eavesdropper has access to the public message and all past source symbols or legitimate outputs, the equivalence between minimum expected log-loss and conditional entropy is assured (Cuff, 2013). This supports fractional equivocation as the principled metric for secrecy in dynamic systems, ensuring that it reflects true operational security properties rather than artifacts of system design.

---

Metric	Definition	Operational Meaning
Fractional Equivocation	$n$8	Normalized uncertainty at the eavesdropper
Average Fractional Eqv.	$n$9	Mean secrecy over channel realizations
Avg. Info Leakage Rate	$H(\cdot)$0	Average confidential bits leaked per channel use

---

Fractional equivocation provides a unified, operationally meaningful, and analytically tractable measure for evaluating and optimizing secrecy in modern communication systems. Its grounding in entropy and connection to distortion-based formulations support both theoretical analysis and practical design across a range of secrecy regimes (Osorio et al., 2019, Cuff, 2013).