PageRank-Based Flow Trust Scoring

Why PageRank?

In dynamic home networks, not all flows are equally trustworthy. Traditional rule-based systems don’t adapt to context. Inspired by the algorithm used by Google Search, we apply PageRank to assign a **trust score** to each flow or host, based on its observed behavior and communication patterns.

Graph Construction

A directed graph is built where:

Nodes represent devices or flow IDs
Edges represent communication between them (frequency, byte volume)

This graph is continuously updated as flows are established.

Trust Score Computation

Using Python's networkx.pagerank(), we compute a normalized trust score for each node. High PageRank indicates:

Frequent, consistent, or central communication
Low anomaly risk

Conversely, low-scoring nodes may be:

New or infrequent participants
Actively communicating with suspicious or isolated hosts

How It Affects Flow Decisions

Trust scores influence:

Priority: Higher-trust flows may be given lower-latency paths
Rate Limiting: Suspicious flows may be throttled or logged more aggressively
Path Selection (Optional): Scores may be used in multipath routing decisions

Visualization

PageRank scores are logged in real-time and can be visualized using CLI output or JSON dumps. Below is a sample snapshot of trust scores for several devices:

{
  "10.0.0.2": 0.304,
  "10.0.0.3": 0.215,
  "10.0.1.2": 0.184,
  "10.2.0.2": 0.103,
  "10.3.1.2": 0.038
}