Eclipse Attack Defense: P2P Sybil Resistance

Attack

Sybil fraction of candidate pool 40%

Peer table size k 6

Defenses

Network Status

SAFE — honest majority

Honest peers in table

6 / 6

Eclipse prob. P=f_eff^k

0.4%

Effective Sybil fraction

40.0%

Eclipsed / rounds so far

0 / 0

Victim node Honest peer Sybil peer

How it works

In a Sybil attack an adversary floods a P2P overlay with cheap fake identities. If a victim node fills its k-slot peer table with independent random picks from a candidate pool that is a fraction f Sybil-controlled, the classic result (Douceur, 2002) gives the probability every slot lands on an attacker identity — the victim is eclipsed, cut off from the honest network and fed a false view of it:

P(eclipse) = f_eff^k

f_eff = f · (1 − rep_gain) · (1 − div_gain)

Two real defenses lower the effective Sybil fraction the selection actually sees:

Reputation scoring — new identities start with zero trust and must survive many honest interactions before being admitted to a peer table. An attacker who keeps minting fresh identities to dodge bans never accumulates trust, so most Sybils are filtered before selection (modelled here as a ~65% cut to f).
Subnet / ASN diversity cap — real Sybils are cheap but IP space is not: an attacker usually controls only a handful of subnets. Capping peers-per-subnet (2 here) stops one cheap IP block from claiming most of the table (modelled as a further ~55% cut to f).

Neither defense is perfect on its own — this is why production systems (Kademlia-based DHTs, Bitcoin's peer selection, Tor's guard rotation) layer both. The simulator reselects the table on a timer and after any control change, applying the diversity cap as an actual per-cluster admission limit during selection, not just as a multiplier on the readout.

Center: victim's own node Bright ring: current peer table

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.