SybilShield-Core | Composite Trust Scoring for Sybil Attack Mitigation

Why this matters

Sybil resistance is not a niche blockchain concern, it is the structural assumption that "one peer, one vote" depends on. The same identity-spoofing pattern underwrites attacks on decentralized federated learning, oracle price feeds, and any system that trusts the apparent diversity of its participants. A framework that closes this gap without reintroducing a central authority strengthens the integrity of decentralized infrastructure that increasingly underpins finance, identity, and shared AI training.

Closes a gap neither existing defense solves. Proof-of-Work prices out small adversaries but not well-funded ones; trusted identity authorities reintroduce the single point of failure permissionless systems exist to avoid. The Composite Trust Score (CTS) detects Sybil behavior from observable evidence instead.
Outperforms single-signal detection. Fusing behavioral, graph-theoretic, and economic signals lifts detection 8.4-21.7 percentage points above SybilGuard, SybilLimit, and stake-weighted baselines at adversary fractions above 30%, evaluated on the same synthetic networks.
Protects shared machine learning, not just consensus. Projected into a federated learning aggregation layer, the CTS cuts poisoned-model backdoor accuracy from 84.7% to 11.3% at a 30% Sybil fraction, directly relevant to any organization running federated or oracle-fed ML on shared infrastructure.
Mitigation is reversible by design. Responses escalate gradually (monitor → throttle → quarantine) and recover as evidence changes, removing the permanent-ban incentive that lets attackers weaponize false reports against honest peers.
Decentralization-preserving. No external authority is consulted and no computationally wasteful puzzle is required; the score recalculates continuously from each peer's own behavior.
Open and reproducible. MIT-licensed reference implementation, simulation harness, and unit tests are public, so detection-rate and overhead claims can be independently verified.

How it works

Three independently maintained sub-scores combine into one Composite Trust Score (CTS) per peer, which then governs quorum weight, connection acceptance, and (optionally) federated learning aggregation weight.

Behavioral score

Continuously updated from block relay timing, mempool propagation, and voting deviations. Decays toward neutral with a 24-hour half-life, so static whitelisting is impossible: inactive peers must keep demonstrating good behavior.

Graph score

Derived from local clustering coefficient, SybilRank propagation from high-stake seed nodes, and subnet diversity. Penalizes the tight-cluster topology characteristic of Sybil subgraphs without requiring a certified honest seed set.

Economic score

Captures stake age and capital lockup. Used as a contributing signal rather than the primary defense, since well-capitalized adversaries can offset pure economic penalties through identity rotation.

Mitigation pipeline

A sigmoid mapping turns the composite score into quorum weight. Four response tiers, monitoring, throttling, peer-list removal, and full quarantine, apply escalating but reversible pressure as the score drops.

Attack coverage

Evaluated across four adversarial scenarios on synthetic P2P networks of 50 to 2,000 nodes, with adversary fractions from 10% to 45%.

Attack class	SybilShield-Core defense
Consensus masking	Behavioral + graph score drift triggers quorum weight reduction before false views can dominate a target's local chain
Eclipse partitioning	IsolationGuard subnet-diversity and per-subnet connection limits reject monopolizing connection attempts independent of the CTS
DFL gradient poisoning	CTS-weighted FedAvg attenuates the aggregation weight of peers whose gradients deviate statistically from honest submissions
Mempool / bandwidth exhaustion	Behavioral score penalizes flooding and GETDATA cycling, escalating to connection throttling and quarantine

Quick start

Python 3.10+ is the only prerequisite.

git clone https://github.com/sunilgentyala/SybilShield-Core.git
cd SybilShield-Core
pip install -e ".[dev]"

# Run a Sybil simulation (50-node network, 30% adversary ratio)
python scripts/run_simulation.py --nodes 50 --sybil-ratio 0.30 --scenario eclipse

# Run unit tests
pytest tests/unit/ -v

Stopping Sybil attacks without identity friction