What the study found
A configuration’s effect depends on where, and against what, you measure it. One setting cut detector confidence substantially, yet independent detectors still flagged every session.
Summary
| Question | Comparison | Result | Outcome |
|---|---|---|---|
| H1, browser profile | Camouflaged against default, both multiplexing c2 | Camouflaged scored higher, not lower, under all 7 detectors | Not supported |
| H2, multiplexing | Multiplexing c2 against off | Confidence fell by 0.29 to 0.51; detection stayed at 100% at every calibrated threshold | Supported for confidence, not for classification |
| H3, resolution path | Alternate mapping against standard DNS | 30 of 30 in both | No difference: compatible |
The claim, exactly: a reduction in detector confidence under a controlled lab threat model. Not detector bypass, and not evidence of real-world censorship resistance.
H1: browser profile
Phase 17 compared the camouflaged and default browser profiles with multiplexing fixed at c2 and parallelism 1, 50 sessions each.
No detector scored the camouflaged profile lower. The traffic shows why: both produced a mean duration of 0.188 seconds and about 54 packets. The camouflaged profile also had higher overhead, and its PRAXIS score was 0.030 worse. Both completed every session.
H1 not supported Browser camouflage acts at the browser layer, and a flow detector does not see that layer. The archived results do not include a direct inferential test of this contrast, so the conclusion is “no demonstrated advantage,” not proof of zero effect.
H2: multiplexing
Phase 17 compared multiplexing off (B) with multiplexing c2 (D), default profile and parallelism 1, 50 sessions each. This comparison carries the most information in the study.
Under the original detector, mean probability fell from 1.000 to 0.487, a reduction of 0.513 (p < 0.001). It also fell under both holdout detectors, by 0.294 and 0.335. Every session still completed, and overhead dropped.
The size of the drop depends on the detector family: 0.38 for random forest, 0.16 for extra trees, and essentially zero for linear SVC, gradient boosting and logistic regression. Tree models split on count thresholds, so a 20% shift crosses split points. Linear models add up correlated features and barely move.
Lower confidence did not change the decision
| Detector | Mean probability | Threshold | Sessions flagged (of 50) |
|---|---|---|---|
| Original (selection) detector | 0.487 | 0.5 | 3 (6%) |
| Holdout, strict-unseen | 0.666 | Calibrated to 0–5% false positives | 50 (100%) |
| Holdout, baseline-known | 0.663 | Calibrated to 0–5% false positives | 50 (100%) |
| Random forest | 0.580 | 0.5 | 49 (98%) |
| Extra trees | 0.839 | 0.5 | 50 (100%) |
| Linear SVC, calibrated | 0.947 | 0.5 | 50 (100%) |
| Gradient boosting | 1.000 | 0.5 | 50 (100%) |
| Logistic regression | 0.996 | 0.5 | 50 (100%) |
The lowest holdout probability of any multiplexed session was 0.58, well above the strictest calibrated thresholds of 0.216 and 0.275. The detector used to pick the winner was fooled most of the time; independent detectors, set the way an operator would set them, caught everything. That gap is why the confirmation step exists.
H2 supported for confidence, not for classification The result holds for c2 multiplexing at parallelism 1. Earlier static multiplexing tests at c2, c4 and c8 showed no change in identifiability.
What changed in the traffic
| Feature | Multiplexing off | Multiplexing c2 | Change |
|---|---|---|---|
| Duration (s) | 1.143 | 0.188 | −83.5% |
| Packets | 68.8 | 53.8 | −21.9% |
| Bursts | 43.1 | 33.8 | −21.4% |
| Payload bytes | 15,349 | 15,548 | +1.3% |
| Capture size (bytes) | 21,493 | 20,321 | −5.5% |
Multiplexed sessions were compressed: much shorter, with about a fifth fewer packets and bursts, carrying the same payload. The traffic became a different shape, not an invisible one. In the Phase 29B artifact control, a detector trained only to tell the two configurations apart separated them perfectly (AUC 1.00), even with duration and packet counts removed and even from the first 20 signed lengths alone.
H3: resolution path
Phase 21 held the selected configuration’s transport constant and ran 30 curl sessions in each of three resolver modes, all to one internal hostname. Port 53 was blocked for the blocked mode; the alternate mapping came from a local hosts-file entry.
Standard DNS and the alternate mapping both succeeded in 30 of 30 sessions: risk difference 0, Fisher p = 1.000. The blocked mode failed in all 30, which confirms the block worked. The Wilson 95% interval for each working mode is 0.886 to 1.000.
H3: no difference Swapping the resolution path did not degrade the configuration. No difference is not proof of equivalence: a drop of up to about 11 points cannot be ruled out. This is a compatibility test with a hosts-file mapping, not evidence of censorship resistance, and detectability under the alternate path was not measured.
“Looks benign” depends on the reference
Phase 31 rescored the confirmation sessions against five benign references. The selected configuration won under all five, but whether it moved toward or away from benign traffic depended on which reference was used.
| Benign reference | Distance change, D against B | Score, D | Lowest score |
|---|---|---|---|
| MAWI day 1 | +0.110 | 0.433 | D |
| MAWI day 2 | +0.111 | 0.433 | D |
| MAWI, both days | +0.105 | 0.432 | D |
| CIC-IDS 2017, mapped | −0.120 | 0.372 | D |
| MAWI and CIC-IDS, balanced | +0.012 | 0.416 | D |
Against packet-derived MAWI traces, the selected configuration moved away from benign traffic; against mapped CIC-IDS 2017, it moved toward it. Benign similarity is a relation to a reference, not a property of the traffic. D won on detector confidence and overhead, not on similarity.
Conclusions
- Layer matters. Browser-profile camouflage did not lower network-level detector confidence under any of 7 detectors.
- Configuration matters. Static multiplexing changed nothing; c2 at parallelism 1 cut confidence by 0.29 to 0.51, depending on the detector.
- Confidence is not classification. Detection stayed at 100% at every threshold calibrated to public false-positive rates.
- Evaluation choices matter. Effects depend on detector family and benign reference; single-model results can mislead.
- The resolution path is compatible. The selected transport stayed fully operational over an alternate mapping, 30 of 30 against 30 of 30.
All five hold within the passive, non-timing, laboratory threat model. A circumvention defense offers meaningful protection only when its advantage persists across aligned detection layers, detector families, operating thresholds and representative reference traffic.
Limitations
| Limitation | What it limits |
|---|---|
| Training confound: proxy sessions came from the lab, benign negatives from public captures | 100% detection may partly reflect the capture environment rather than proxy use. The B against D comparison is less affected, since both came from the same lab. |
| Near-replicate sessions: scripted workload, fixed destinations, duration spread of 0.003 s | p-values and effect sizes show separation, not operational effect size |
| Static, shape-only observer | No claim against timing-aware, handshake-aware or adaptive censors |
| Confirmation design: parallelism fixed at 1; randomized run order not documented | H2 confirmed for c2 at parallelism 1 only; drift not ruled out |
| Narrow resolver test: one hostname, hosts-file mapping, one block type | Shows compatibility only, not censorship resistance or detectability |
Next steps
- Same-environment negatives. Train detectors on benign traffic captured in the same lab, so they cannot lean on environment differences.
- Realistic variability. Variable workloads and many destinations instead of near-replicate sessions.
- Factorial transport design. Confirm multiplexing and parallelism together, not at parallelism 1 only.
- Adaptive adversaries. Censors that retrain on the configurations they observe (Abolfathi et al., 2024).
- Real decentralized naming. Production systems under poisoning, NXDOMAIN injection and null routing, measuring availability, latency and detectability.