NETWORK RESILIENCE // OFFICIAL SHA-256 SEALED RUNS
Network Resilience Attack Results
Three controlled Pi 1 netem packet-loss attacks. Each analysis window is exactly 1000 official terminal outcomes. Configured impairment, measured loss, delivery, retry recovery, terminal failures and latency are reported separately.
VERIFIED
RANDOM LOSS 10%
Random Packet Loss — 10%
Configured loss
10%
Measured loss
9.552%
Delivery rate
100%
Terminal failures
0
Retry events
93
Retry recoveries
87
Median RTT
129.5ms
P95 RTT
6,535.35ms
TARGETED LOSS 20%
Targeted Receiver Traffic Loss — 20%
Configured loss
20%
Measured loss
19.947%
Delivery rate
99.1%
Terminal failures
9
Retry events
244
Retry recoveries
193
Median RTT
132ms
P95 RTT
6,912ms
TARGETED LOSS 50%
Targeted Receiver Traffic Loss — 50%
Configured loss
50%
Measured loss
50.536%
Delivery rate
73.9%
Terminal failures
261
Retry events
739
Retry recoveries
326
Median RTT
546ms
P95 RTT
13,300.4ms
COMPARISON TABLE
| Attack | Configured | Measured | Delivered | Failed | Delivery rate | Retries | Recoveries | Median RTT | P95 RTT | Max RTT |
|---|---|---|---|---|---|---|---|---|---|---|
| Random Loss 10% | 10% | 9.552418% | 1000 | 0 | 100% | 93 | 87 | 129.5 ms | 6,535.35 ms | 13,460 ms |
| Targeted Loss 20% | 20% | 19.946524% | 991 | 9 | 99.1% | 244 | 193 | 132 ms | 6,912 ms | 13,483 ms |
| Targeted Loss 50% | 50% | 50.535540% | 739 | 261 | 73.9% | 739 | 326 | 546 ms | 13,300.4 ms | 13,767 ms |
DELIVERY VS FAILURE
RETRY EVENTS VS SUCCESSFUL RECOVERIES
CONFIGURED VS MEASURED PACKET LOSS (%)
LATENCY — MEAN / MEDIAN / P95 RTT (ms)
ATTEMPT DISTRIBUTION (1000 OUTCOMES EACH)
EVIDENCE & REPRODUCIBILITY
NETWORK RESILIENCE FINDINGS
- At 10% random loss, bounded retransmission recovered all official exchanges.
- At 20% targeted receiver loss, delivery remained 99.1%, with retransmission recovering most impaired exchanges.
- At 50% targeted receiver loss, delivery decreased to 73.9% as retry-budget exhaustion became significant.
- Median RTT remained comparatively low at moderate loss because many exchanges succeeded on the first attempt.
- P95/P99 latency increased sharply because timeout-driven retransmissions dominate tail latency.
- Empirical packet-loss measurements closely matched configured Linux netem values, independently validating the impairment setup.
- Reverse-path packet counts matched during the targeted-loss experiments, supporting that impairment was applied specifically in the sender-to-receiver direction.
- Cryptographic/security processing remained operational during impairment; observed failures were network/retry outcomes, not cryptographic authentication failures.