# A Framework for Reputation Aware Uninorm-driven Consensus Algorithms for Blockchain Networks

Crypto.world

> Existing consensus algorithms tend to concentrate power: PoW rewards computational power, PoS rewards the largest stakes, and DPoS concentrates influence in a small set of delegates.

**arXiv**: 2607.20700

**Published**: 2026-07-22

**Authors**: Bruno Ramos-Cruz, Javier Andreu-Perez, David Richerby, Luis Martínez

## AI summary

## Overview

**Research area:** Blockchain consensus algorithms, specifically reputation-based validator selection, using fuzzy logic tools — intuitionistic fuzzy sets (IFSs) and uninorm aggregation operations (UAOs).

**Technical level:** Advanced. The paper assumes familiarity with fuzzy set theory (membership/non-membership functions, t-norms, t-conorms, uninorms) and with blockchain consensus mechanisms such as PoW, PoS and DPoS.

**Scope:** The paper proposes and formally defines a two-phase framework — an IFS-based reputation degree and a UAO-based reputation weight — for monitoring validator reputation over time and plugging it into the selection step of reputation-based consensus algorithms.

## What This Paper Is About

Existing consensus algorithms tend to concentrate power: PoW rewards computational power, PoS rewards the largest stakes, and DPoS concentrates influence in a small set of delegates. In reputation-based consensus algorithms, a validator that loses reputation typically has no mechanism to recover it, so it is permanently excluded from validation. The paper's goal is to model validator reputation in a way that captures uncertainty about how reputable a node really is, and that lets nodes recover from past failures while still penalising bad behaviour.

## Key Contributions

1. **A reputation-aware, uninorm-driven methodology for blockchain consensus** that combines the current round's reputation with the previous round's reputation weight, rather than selecting solely on the current round's score.
2. **An intuitionistic fuzzy framework for managing reputation uncertainty**, in which reputation is represented by a membership degree, a non-membership degree and an intuitionistic index expressing indeterminacy — motivated by the fact that knowledge about a validator's reputation is never complete.
3. **A reputation recovery mechanism** in which UAOs reinforce both positive and negative reputation evolution over time, allowing validators to rectify past failures instead of being permanently excluded.
4. **An analysis of suitable functions for the IFS and UAO components** (linear, logarithmic, square root, exponential, polynomial, logistic, hyperbolic, sigmoid) together with formal properties of the reputation degree and a modular selection algorithm that can replace stake-based probability mechanisms similar to PoS.

## Main Findings

- **Linear cost, no extra communication:** The framework "maintains linear computational complexity and does not introduce additional communication overhead beyond the underlying consensus protocol."
- **Successful Validation Rate definition:** A validator's rate in round *t* is the number of successful validations divided by *t*, since each validator has exactly one attempt per round.
- **Reputation degree is well-behaved:** Proposition 2 states that RepD(v_i, t) lies in (0, 1), increases monotonically with the membership degree μ_A(x), decreases monotonically with the non-membership degree ν_A(x), and is continuous on [0, 1].
- **Uncertainty reduces the reputation score:** The reputation degree is a certainty-adjusted version of the basic ratio r(x) = μ_A(x) / (μ_A(x) + ν_A(x)), scaled between a lower bound r₀ and r using a function s(π_A(x)) = π_A(x)^α.
- **Worked numerical illustration:** With SuccVR = 0.75, μ_A(0.75) = 0.70, ν_A(0.75) = 0.15, π_A(0.75) = 0.15 and r₀ = 0.30, the ratio μ_A(x) / (r₀(μ_A(x) + ν_A(x))) equals 0.70 / 0.26 ≈ 2.6923, giving RepD ≈ 0.3068 for α = 2, RepD ≈ 0.3480 for α = 1, and RepD ≈ 0.4402 for α = 0.5 — so higher α lowers the reputation degree and lower α raises it.
- **Motivating IFS example:** (μ_A = 0.8, ν_A = 0.1) represents an individual with substantial evidence of honest behaviour, whereas (μ_A = 0.08, ν_A = 0.01) represents an individual about whom little evidence exists, despite most of it suggesting honesty.
- **Symmetric-evidence case:** When μ_A(x) = c·ν_A(x) with c = 1, the reputation degree reduces to r₀(1/(2r₀))^(π_A(x)^α), which the authors describe as "some evidence that you are honest, but also some evidence that you are dishonest, which makes me less sure of your honesty."
- **Two-stage selection with a threshold:** Validators with w(v_i, t) ≥ τ form the candidate set V_τ; if V_τ is empty, the validator with maximum w(v_i, t) over V is selected, otherwise a validator is randomly selected from V_τ (uniformly or weight-proportionally).
- **Liveness and safety claim:** Proposition 1 states that if at least one validator satisfies w(v_i, t) ≥ τ, the selection step guarantees liveness by selecting a validator in finite time, and that because w(v_i, t) encodes long-term behaviour, the probability of selecting a faulty validator decreases exponentially with consecutive failures.
- **Experimental results are not reported in the provided text:** Section 5 is described as defining implementation features and presenting key results, but the full paper content supplied here is truncated before that section, so no quantitative benchmark, dataset size or comparison figures can be reported.

## Methodology in Plain English

Each validator's behaviour is scored in two stages. First, the framework counts how many of the validator's validation attempts succeeded, divides by the number of rounds, and feeds that rate into an intuitionistic fuzzy set. The fuzzy set says two things at once: how much evidence points to the node being honest, and how much points to it being dishonest. Whatever is left over — one minus the two — is treated as genuine uncertainty, and a node with lots of uncertainty gets pulled down toward a conservative baseline value r₀ rather than being trusted on thin evidence.

Second, the resulting short-term reputation degree is combined, using a uninorm aggregation operator, with the node's reputation weight from the previous round. A uninorm is like an "and/or" operator from fuzzy logic, except it lets the neutral point sit anywhere in [0, 1] rather than being fixed at 0 or 1, so the authors can tune how strongly good and bad history are reinforced. The output is a long-term reputation weight.

Selection then applies a threshold τ: only validators whose weight clears the threshold enter the candidate pool, and one is chosen at random from that pool rather than always taking the single highest scorer. Randomising inside the pool spreads participation while the threshold keeps unreliable nodes out. Because the selection step is modular, it can be dropped into existing reputation-based protocols and substitute for stake-based probability mechanisms.

## Why This Matters

**Impact on research:** The paper connects two normally separate strands — reputation-based consensus in blockchains and fuzzy-logic trust modelling (prior work by Carbó et al. and Quesada et al. is cited) — and adds a formal treatment of reputation *recovery*, which the authors argue classical reputation algorithms lack. It also frames reputation as an explicitly uncertain quantity rather than a crisp score.

**Real-world applications:**

- Consortium or permissioned blockchain networks where node reliability varies and permanent exclusion of participants is undesirable.
- Public blockchain validator selection where a threshold-plus-randomisation design is meant to counter concentration of power among high-stake or high-compute participants.
- Distributed ledger deployments in regulated or institutional settings (the journal is *Results in Engineering*), where auditability of validator behaviour over time matters.
- Systems needing a drop-in replacement for stake-weighted selection without adding communication rounds.

**Industry relevance:** Because the authors claim linear computational complexity and no extra communication overhead beyond the underlying consensus protocol, the framework is pitched as something that can be layered onto existing protocols rather than requiring a new network layer. The explicit knobs — the threshold τ, the bound r₀, the exponent α, and the choice of μ, ν and uninorm functions — give implementers a way to dial strictness against inclusivity.

## Future Directions

- Empirical validation of the framework against concrete consensus deployments: the paper's Section 5 is described as presenting implementation features and key results, but the supplied text is truncated, so the reported performance and any comparison baselines remain to be examined.
- Choosing among the function families listed for μ(x) and ν(x) (linear, logarithmic, square root, exponential, polynomial, logistic, hyperbolic, sigmoid) in a principled way, since the authors state the choice depends on the desired rate of reputation increase or decrease.
- Tuning and justifying the parameters r₀ (including the option r₀ = βr with 0 < β < 1), α, and the threshold τ, and establishing how each interacts with decentralisation and security guarantees.
- Formally characterising the claimed exponential decrease in the probability of selecting a faulty validator under adversarial behaviour, and testing the liveness and safety proposition beyond its stated assumptions.

## Target Audience

Researchers and graduate students working on blockchain consensus mechanisms, distributed-systems security, and computational intelligence applied to trust and reputation. It is also relevant to protocol designers and engineers who need a modular, reputation-based alternative to stake-weighted validator selection, and to fuzzy-systems researchers looking for an applied use of intuitionistic fuzzy sets and uninorm aggregation operators. Readers without a background in fuzzy logic will need to work through Section 2 before the methodology is accessible.

## Authors’ abstract

The operation of blockchain is governed by consensus algorithms (CA). Several consensus mechanisms require significant computational power, while others necessitate high amounts of stakes to select the participant to validate and verify the transactions in the block, leading to centralisation of power and participant exclusion. This paper proposes a novel methodology to address these issues in reputation-based consensus algorithms by studying the reputation behaviour of the…

[The full abstract on arXiv](https://arxiv.org/abs/2607.20700)

[Read the original paper](https://arxiv.org/abs/2607.20700)

---

Canonical: https://crypto.world/paper/a-framework-for-reputation-aware-uninorm-driven-consensus-algorithms-2607.20700
