Trust Transitivity

Trust is fundamentally directional and non-symmetric. If entity A trusts entity B, it does not logically follow that B trusts A. In a transitive chain (A→B→C), the derived trust A has for C is a function of A's trust in B and B's trust in C, but C has no implicit trust in A. This property distinguishes trust networks from simple social connections and requires directed graph models for accurate computation.

Trust does not propagate without loss. Each hop in a transitive chain introduces attenuation, where the derived trust value is discounted by a factor (often multiplicative). Key mechanisms include:

• Distance Decay: Trust diminishes with each successive edge traversal
• Confidence Decay: Uncertainty compounds across hops, widening error margins
• Context Decay: Trust in one domain (e.g., medical expertise) does not fully transfer to another (e.g., financial advice)

Without attenuation, a single compromised node could pollute the entire network.

When multiple transitive paths exist between two nodes, reputation systems must define a composition function to aggregate them. Common approaches include:

• Weighted Average: Combines trust scores from all paths proportionally to their individual confidence
• Maximum Path: Takes the strongest trust path, ignoring weaker signals
• Minimum Path: Conservative approach using the weakest link in the strongest path (bottleneck trust)

The choice of composition function dramatically affects resilience to Sybil attacks and collusion.

While trust can be transitive, distrust is generally intransitive. If A distrusts B, and B distrusts C, A does not necessarily trust C (the enemy of my enemy is not automatically my friend). Distrust propagation requires separate handling:

• Distrust edges often act as barriers that block trust flow
• Negative trust scores require signed graph algorithms rather than standard eigenvector methods
• Systems like Subjective Logic model distrust explicitly as a separate dimension from trust and uncertainty

Iterative transitive trust algorithms like EigenTrust and PageRank rely on convergence to a stable fixed point. The system repeatedly propagates trust scores through the network graph until values stabilize within a tolerance threshold. Key properties:

• Requires the trust matrix to be stochastic and irreducible
• Damping factors prevent rank sink and ensure convergence
• The Perron-Frobenius theorem guarantees a unique, positive eigenvector solution for well-formed trust graphs

Trust is rarely universal; it is scoped to specific contexts or domains. An entity trusted for code review may not be trusted for financial auditing. Transitivity must respect these boundaries:

• Ontology-anchored edges: Each trust relationship carries a domain tag (e.g., expertise:neuroscience)
• Context-aware propagation: Trust flows only along edges with compatible or overlapping scopes
• Sub-context inheritance: Trust in a broad domain (e.g., medicine) may partially propagate to narrower subdomains (e.g., cardiology) with controlled attenuation

The foundational application of trust transitivity in information retrieval. PageRank and TrustRank algorithms propagate authority from a curated set of seed pages through the hyperlink graph. A page linked by a highly trusted page inherits a portion of that trust, enabling search engines to rank content quality without manually reviewing every document. This transitive flow penalizes link farms and spam networks by ensuring bad actors cannot easily acquire authority.

In distributed file-sharing and blockchain networks, the EigenTrust algorithm calculates a global trust value for every peer by aggregating and weighting local trust scores transitively. A peer's reputation is not just its own history but the composite trust of those who trust it. This provides Sybil resistance by making it computationally expensive for malicious actors to create fake identities that gain meaningful influence in the network.

Trust transitivity enables end-to-end supply chain verification. A retailer trusts a distributor, who trusts a manufacturer, who trusts a raw material supplier. By chaining cryptographic attestations at each custody transfer, the final consumer can transitively verify the origin and authenticity of a product. This is implemented using Verifiable Credentials and distributed ledger technologies to create immutable provenance chains.

New entities in a network face the cold start problem—no interaction history means no trust score. Trust transitivity solves this through reputation portability and reputation attestation. A new merchant on a marketplace can import their established reputation from another platform where they have a high score. The new platform transitively trusts the old platform's scoring mechanism, giving the merchant immediate credibility.

A distributed reputation management algorithm for peer-to-peer networks that calculates a global trust value for each peer by analyzing the transitive trust relationships across the network. It uses an iterative matrix multiplication process similar to PageRank, where a peer's trust score is determined by the trust scores of those who vouch for it. The algorithm converges to a stable left principal eigenvector representing the stationary distribution of trust.

• Solves the Sybil attack problem in P2P file-sharing
• Requires a set of pre-trusted peers to anchor the computation
• Each peer maintains a local trust vector based on transaction satisfaction

A type of probabilistic logic developed by Audun Jøsang that explicitly models uncertainty and belief ownership, allowing reputation systems to represent trust as a composite of belief, disbelief, and uncertainty masses. Unlike classical probability, it enables the discounting and consensus operators that mathematically formalize trust transitivity.

• Uses a ternary opinion structure: ω = (b, d, u, a)
• The discounting operator (⊗) propagates trust along a chain
• The consensus operator (⊕) fuses parallel trust paths

A link analysis algorithm designed to combat web spam by propagating trust from a manually selected set of highly reputable seed pages to the rest of the web graph. It operates as a modified PageRank where the random surfer teleports only to trusted seeds, causing trust to attenuate with each hop away from the seed set.

• Inverse of spam detection: good pages propagate goodness
• Trust decay factor controls how far trust travels
• Seed selection is critical; typically uses human-curated directories

A specialized knowledge graph or network structure where nodes represent entities and directed edges represent trust or endorsement relationships, used to compute transitive reputation scores. Unlike simple scalar scores, the graph preserves the topology of trust, enabling path-based queries like 'find the most trusted path between A and B' and resistance to collusion attacks.

• Edges can be weighted by confidence or transaction value
• Supports graph neural network approaches for inductive scoring
• Enables detection of trust rings and reciprocal boosting

A mechanism in trust models that reduces the weight or value of historical behavioral data over time to ensure that the reputation score reflects an entity's most recent performance. In transitive trust chains, decay prevents stale endorsements from indefinitely influencing derived trust scores, ensuring the system adapts to entities that become compromised or negligent.

• Implemented via exponential time decay functions: w(t) = e^(-λt)
• Critical for systems where behavior is non-stationary
• Balances recency vs. long-term reliability signals