Pulse attack: not new, and even our headline number was already old news
Discharge B, charge R, repeat. This model had already been done in 2005 (Srivatsa et al.'s strategic oscillation); on 2026-08-06 we withdrew the "novel attack" claim.
#The model
The attack strategy is a "pulse":
discharge: B consecutive bad deliveries → charge: R clean deliveries → repeatParameterized as (B, R). Three existing strategies are special cases of it:
| Existing strategy | (B, R) |
|---|---|
| patient (sustained, slow) | (1, ∞) |
| whitewash (misbehaves all the time) | (∞, 0) |
| Continuous attack | (∞, 0) or (1, 1) |
#Prior work
Srivatsa et al. (2005, TrustGuard) §2 model I writes, verbatim:
"Or it could oscillate between building and milking reputation."
Their model I is a fixed-period square wave, which is exactly our (B, R) parameterization. We did not sweep models II–IV (exponential intervals / random goodness / sinusoidal ramp).
Hoffman et al. (2009, ACM Computing Surveys) §5.4 list it as a named attack category, oscillation attack, credited to Srivatsa 2005. Their version includes multiple identities taking shifts, which is stronger than our single attacker.
#Even our headline had been anticipated
#Withdrawn claims
On 2026-08-06 we formally withdrew two kinds of claims:
| Claim | Reason for withdrawal |
|---|---|
| The pulse attack is a new finding | Srivatsa 2005 model I describes it verbatim; Hoffman 2009 lists it as a named category |
| The seven mechanism components are novel mechanisms | Six already have equivalent prior work |
#What we swept
30 seeds × 600 rounds, each cell with 5 honest residents + 1 attacker, audit rate 0.2, reviewer accuracy 0.7, probation m = 2.
1,740 per-round files, 1,041,246 lines.
Eight experiments: E17 (optimal burst length), E18 (recovery time), E19 (lock the total misbehavior count, leaving only timing differences), E20 (blind spot × temporal structure), E21 (can a higher audit rate compensate), E22 (how many items leak before the first catch), E23 (damage concentration of selective bursts), E24 (which cell is cheapest).
#What happened to the six conclusions
| Original conclusion | Verdict |
|---|---|
| Temporal structure doesn't matter (1.81×) | Overstated |
| The blind spot matters far more than parameter choice | Overstated |
| Beyond a blind spot > 0.5–0.75, pulsing does worse than sustained | Refuted |
| Raising audits can't rescue a blind spot | Refuted |
| Caught once = removed permanently | Overstated |
| At equal budget, crude beats pulsing | Refuted |
#Two positive conclusions that were kept
The re-verification did not refute the whole report. Two held up:
#1. Detection probability is a single product
P(caught) = (1 − blind spot β) × sampling rate p × audit accuracyThe three factors are on the same axis. Measured support: at blind spot 0.5, audit 0.1→1.0 cut total hits by 82% (p = 5.5e-6); audit 0.5→1.0 exactly offsets a 50% blind spot (paired difference 0.03, p = 0.78).
#2. Interaction η² = 0.264
The blind spot × temporal structure interaction is 4 times the main effect of temporal structure (0.039).
#Three that are still not re-verified
| Reaction lag | With zero blind spot, one wave still leaks 2.53 items, and reaction takes 27.95 rounds on average (blind spot 0.5 → 3.95 / 34.53; 0.75 → 4.55 / 25.14) |
| Selective attack | Fewer total hits (1.93 < 2.37) but 4.8 times the high-value damage (1.93 vs 0.40; at blind spot 0, 100% vs 16.9%) |
| The real crossover | Measured at about 0.90. And for an attacker who picks its parameters, there is no crossover: pulse(20,2)/(20,1)/(50,1) at blind spot 1.0 got 81/84/87 hits, all higher than patient's 79 |
#The general lesson of this case
#Next
- The full details of the re-verification: adversarial-recheck
- The list of withdrawn claims: Changelog
- Evidence ledger: Evidence ledger
