tor   Tor (onion routing)

During the June 2025 Iran shutdown, circumvention tool performance diverged sharply by transport design. Psiphon's multi-protocol architecture sustained 1.5 million concurrent users—roughly one-third of its normal Iranian base. Lantern's "proxyless" protocol (domain-fronting via CDN, ~40% of Lantern's Iranian traffic) showed moderate success. Tor usage collapsed during the blackout but bridge connections surged and rebounded quickly after lifting. BeePass (serving 500k+ daily users at shutdown onset) used live A/B testing of port/obfuscation-prefix combinations to probe the censors' blocking parameters in real time. The Ceno Browser's P2P network grew from 600 active peers on June 13 to ~8,000 by July 11, indicating that decentralized fallback paths stayed up even during peak blocking.

2025-iran-shutdown-measurement §Tool Performance evaluation

HOWLR successfully detects 83% of the 1,902 real-world prefix hijack incidents observed in high-confidence Cloudflare Radar data from May 2026. The detection method fails when the hijacked prefix contains no authenticatable witnesses, and also cannot defend against routing adversaries capable of transparently forwarding non-targeted traffic within the prefix.

2026-doumanidis-howlr-client-driven-approach §4, §7 evaluation

HOWLR's T4 variant (scanning only port 443) reduces worst-case prefix scan time from 260 minutes to 4 minutes while missing fewer than 2 witnesses in 83.6% of prefixes; the worst-case drop was 13 witnesses in a high-density prefix, still leaving 15 usable witnesses. Over a 6-day monitoring period, 96% of the 30,217-witness set maintained 99+% uptime.

2026-doumanidis-howlr-client-driven-approach §5, §6 defense

HOWLR provides Light protection (≥3 witnesses in prefix) to 89% of Tor relay prefixes and Strong protection (≥8 witnesses from ≥2 CAs) to 64%, based on sampling 200 random relays. By contrast, DNS server prefixes receive Light protection in only 40% and Strong protection in only 12% of cases, reflecting their tendency toward residential or sparse hosting.

2026-doumanidis-howlr-client-driven-approach §6 evaluation

BGP hijacking remains a practical threat to circumvention infrastructure: BGPWatch reported 5,673 possible hijack events in 2024 alone, and approximately 43% of ASes currently do not enforce route origin validation, leaving RPKI-covered prefixes still vulnerable. An attacker can statistically divert 50% of traffic by announcing an equally specific prefix with the genuine origin and itself prepended as next hop.

2026-doumanidis-howlr-client-driven-approach §2.2 detection

NATA (Non-invasive Active Traffic-correlation Analysis) injects low-frequency bandwidth waveforms (sinusoidal, square-wave, triangular) into Tor TCP connections at an upstream gateway without endpoint compromise, payload decryption, or Tor-browser modification. BM-Net, a selective state-space classifier trained on the exit-side observations, achieves a 99.65% binary detection F1 score distinguishing watermarked from natural traffic on a 20,000-trace real-world dataset.

2026-fan-activeflowmark-assessing-tor §IV, §VI.D, Table III detection

BM-Net achieves a 99.65% binary detection F1 score for distinguishing bandwidth-watermarked Tor flows from natural traffic, outperforming all evaluated baselines (next best: TikTok at 75.96% F1). The accuracy gap stems from active perturbation imposing a deterministic low-frequency throughput constraint rather than relying on subtle natural metadata, making the detection task fundamentally easier than passive website fingerprinting.

2026-fan-activeflowmark-assessing-tor §VI-D, Table III–IV detection

BM-Net achieves a 99.65% binary detection F1 score distinguishing watermarked from natural Tor flows, and a 97.5% macro-F1 score for fine-grained modulation classification across sinusoidal, square-wave, and triangular patterns. The fine-grained test set contains 201 held-out samples collected from ten clients across five geographic regions (Europe, North America, Australia, Southeast Asia, East Asia), with training traces including traffic collected under WTF-PAD and Walkie-Talkie defenses.

2026-fan-activeflowmark-assessing-tor §VI-D, Table III, Table V evaluation

Active bandwidth perturbation has an inherent detectability–stability trade-off: overly aggressive low-rate phases cause Tor SENDME-based flow control stalls, retransmissions, timeouts, or circuit replacement before sufficient correlation evidence is collected. The paper selects a 30-second modulation period and an empirically determined minimum shaping rate; the usable shaping range varies with relay load, path length, TCP congestion control behavior, and Tor multiplexing.

2026-fan-activeflowmark-assessing-tor §VII-A detection

Tornettools-based simulations with a 1%-scaled Tor network show that with 1 adversary-controlled exit relay (148 Mbps), the exit-observation probability pexit ≈ 2.13%; with 5 adversary-controlled relays, pexit exceeds 10%. Tor's bandwidth-weighted path selection means high-bandwidth malicious exits attract a disproportionate share of circuits, and repeated observations over multiple circuits compound correlation risk multiplicatively even at modest relay counts. The simulation scaled historical Tor consensus data to 1% of the real network (80 relays), and repeated observations across T independent windows compound as 1 - (1 - Pcorr)^T.

2026-fan-activeflowmark-assessing-tor §VI-A, Table I evaluation

BM-Net achieves a 97.5% macro-F1 score for fine-grained classification of three modulation geometries (sinusoidal, square-wave, triangular) from noisy exit-side Tor observations using only 201 labeled test samples collected across cross-continental Tor paths. Residual errors concentrate between natural traffic and square-wave modulation, as abrupt low-rate transitions are partially smoothed by Tor multiplexing and network jitter. Per-class F1 is 95.7% for square-wave, the hardest class, versus 99.0% each for sinusoidal and triangular.

2026-fan-activeflowmark-assessing-tor §VI-D, Table V detection

NATA requires no endpoint compromise, no Tor-browser modification, and no payload decryption; it operates solely from (1) an upstream gateway controlling Tor TCP connections via standard Linux tc/wondershaper rate-limiting and (2) one or more adversary-controlled exit relays passively recording packet traces. The shaper identifies Tor connections using flow-level metadata (client IP, relay IP, port, transport protocol), meaning the adversary needs only ISP or AS-level vantage, not host-level access. The authors name this Shaper-Sniffer architecture NATA, for Non-invasive Active Traffic-correlation Analysis.

2026-fan-activeflowmark-assessing-tor §III-A, §IV-A detection

Padding-based client-side defenses including WTF-PAD and Walkie-Talkie are insufficient against active bandwidth perturbation: they reshape packet timing and burst structure but cannot remove the upstream rate limit imposed by the gateway shaper. BM-Net trained on a defense-aware dataset containing both undefended and WTF-PAD/Walkie-Talkie traces still achieves 99.65% F1, and the paper explicitly notes that 'client-side padding and burst reshaping may alter the logical traffic pattern, but they do not directly remove the rate limit imposed by the upstream bottleneck.'

2026-fan-activeflowmark-assessing-tor §VI.E.2, §VII detection

TrafficMoE achieves 97.65% accuracy and F1-score on the ISCX-Tor2016 dataset, substantially outperforming all baselines including the best pretraining-based competitor FlowletFormer (91.16% F1), by separately modeling protocol headers and encrypted payloads via dual-branch sparse Mixture-of-Experts rather than treating them as a unified byte stream.

2026-he-trafficmoe-heterogeneity-aware-mixture §IV-B, Table II detection

Classification from the first 5 packets × 320 bytes (1600-byte burst) achieves near-perfect accuracy across Tor (F1=0.9990), VPN (F1=0.9871), malware (F1=0.9954), and IoT attack traffic (F1=0.9966), with IP addresses masked and only header and initial payload retained. The earliest portion of each packet provides sufficient discriminative information for a classification decision made within the first kilobyte of a flow.

2026-kulatilleke-mambanetburst-direct-byte-level §III-A, §V-A, Table II–III detection

FreeUp achieves 86.68% AUC on CIC-IoT2023, 85.44% AUC on DoHBrw2020 (malicious DNS-over-HTTPS tunneling), and 95.53% AUC / 93.22% F1 on ISCX-Tor2016 (Tor anonymous traffic), outperforming all nine baselines by more than 3% AUC on the first two datasets. The ISCX-Tor2016 result demonstrates that frequency-decoupled ML classifiers can detect Tor-like anonymous traffic with high confidence under zero-positive (unsupervised) training.

2026-lian-decompose-understand-fuse §V-B, Table I evaluation

FreeUp operates under a zero-positive (unsupervised) learning paradigm — trained exclusively on normal traffic with no labeled anomaly examples — yet achieves 95.53% AUC on Tor traffic and 85.44% AUC on DNS-over-HTTPS tunneling detection. This demonstrates that frequency-aware anomaly detectors generalize to novel circumvention protocols without requiring any labeled attack data, eliminating the labeling bottleneck that previously limited ML-based censorship detection.

2026-lian-decompose-understand-fuse §III, §V-B detection

CAPTCHAs co-occurred with 'Resource Inaccessible' in 70% of CAPTCHA reports and appeared in 23% of all 'Resource Inaccessible' reports; overall 14% of the 119 reports involved one or both problems. Two CAPTCHA failure modes were identified: excessive repetitive CAPTCHAs and broken CAPTCHA servers that made the underlying website permanently inaccessible. The 'Unusual traffic detected from your computer' Google error appeared in 5% of all reports.

2026-micallef-reportor-facilitating-user §6.1.2 evaluation

17% of ReporTor reports cited broken content; investigation found that several websites returned HTTP 403 errors through Tor Browser but loaded normally in Firefox, revealing deliberate differential treatment of Tor traffic masquerading as technical failure. Blocked resources included advertising platforms (e.g., t.co) and JavaScript files handling cookie-consent dialogs, and 8% of reports involved authentication failures where initial page load succeeded but subsequent auth steps were silently refused.

2026-micallef-reportor-facilitating-user §6.1.3 evaluation

Two mechanistically distinct blocking categories account for Tor exit-node inaccessibility: explicit blocks (deliberate CDN/WAF configuration, e.g., Akamai Bot Manager renders AirBnB inaccessible over Tor) and dynamic blocks (abuse-detection systems that flag Tor exit-node IPs because pooled traffic from diverse users raises apparent abuse scores, triggering rate-limiting or blocking despite no explicit Tor policy). Cloudflare does not block Tor by default, but its aggressive IP scoring results in disproportionate blocking in practice.

2026-micallef-reportor-facilitating-user §6.1.1 evaluation

'Resource Inaccessible' was the most frequently reported issue (61% of 119 submitted reports) during a month of naturalistic Tor Browser browsing, followed by CAPTCHAs (18%), Broken Content (17%), Other Issues (13%), and Timeouts (5%). These categories document the operational failure modes that degrade everyday Tor Browser usability beyond protocol-level censorship.

2026-micallef-reportor-facilitating-user §6.1, Figure 3 evaluation

The privacy properties of Tor Browser structurally preclude automated telemetry collection, creating a persistent blind spot for diagnosing user experience failures at scale. ReporTor demonstrates that anonymous voluntary in-situ reporting — transmitted over the Tor network and stored in a password-protected onion-service database — can substitute for telemetry: 119 reports over one month from five expert users sufficed to reproduce approximately half of reported issues exactly and identify root causes for most of the remainder.

2026-micallef-reportor-facilitating-user §1, §7.1 evaluation

CNN-based passive traffic analysis failed to deanonymize I2P services when transferred from a controlled lab to the public I2P network. Lab-trained models produced mostly unusable results: the 'Without port' variant misclassified Class 2 packets at 71.6–88.4× the true count, and the 'Without payload' variant was only marginally better (12.8–13.2× false positives), demonstrating that lab-learned patterns do not generalize to real-world I2P traffic.

2026-rohrer-convolutional-neural-networks-deanonymisation-i2p §V Fourth Experiment / Table VIII detection

Fano's inequality establishes a theoretical lower bound on deanonymization error probability as a function of anonymity set size |Θ|, prior uncertainty H(X), and mutual information leakage I(X;Y). For a network of N sufficiently large nodes with uniform routing, Pe ≥ (log N − 1) / log(N−1), approaching 1 (perfect anonymity). The authors found that closed-form estimation of I(X;Y) from I2P traffic features was analytically intractable, requiring ML approximation — and that ML also failed in practice.

2026-rohrer-convolutional-neural-networks-deanonymisation-i2p §III-A evaluation

Three conditions are jointly necessary and sufficient for a Fragging attack: (C1) the adversary can drop packets while knowing their sender—met by a malicious gateway/entry node in both Nym and Tor; (C2) the adversary can detect a missing fragment—met by Nym's explicit fragment-set identifiers and by Tor's encryption-state desynchronization on cell drop; and (C3) the adversary can learn the recipient from the surviving fragments—met in both systems because other fragments of the set arrive intact.

2026-schadt-breaking-partially-fixing §4.3 detection

Fragging-like attacks on Tor allow an adversary controlling the entry node to encode at least a 1-bit confirmation signal by dropping a single stream cell and observing circuit teardown (caused by encryption-state desynchronization between client and exit relay). If the adversary knows the expected cell count, they can select which cell to drop and thereby encode log(n) bits of information to the exit relay.

2026-schadt-breaking-partially-fixing §4.2 detection

Fingerprinting acceptance averaged 3.90 for cybersecurity applications, 3.73 for law enforcement, and 3.00 for user experience on a 5-point scale (n=734); privacy-protective behaviors (Privacy Behavior Index, p<0.01) and prior use of Tor or Brave (p<0.01) were the strongest negative predictors of acceptance across all three domains. Privacy concern level (Westin Index) alone did not significantly predict fingerprinting awareness, whereas behavioral privacy practices did.

2026-schramm-analyzing-societal-awareness §4.2.1, §4.2.2 evaluation

Across 734 participants, fingerprinting awareness (mean 2.4 on a 1–4 scale) was significantly lower than cookie awareness (mean 2.86), confirmed by Wilcoxon signed-rank test at p < 0.001. Only 22% of participants had used a privacy-focused browser (Tor or Brave), and the Spearman correlation between cookie and fingerprinting awareness was only ρ = 0.40, indicating that tracking-technology literacy does not transfer automatically between mechanisms.

2026-schramm-analyzing-societal-awareness §4.1.1, Figure 2 evaluation

Among 734 participants, 70.4% were categorized as Privacy Pragmatists willing to trade data for convenience, while only 25.9% were Privacy Fundamentalists; prior use of Tor or Brave (~22% of the sample) and a higher Privacy Behavior Index were the strongest positive predictors of fingerprinting awareness (p<0.01 each), while self-reported privacy concern alone had no significant relationship with awareness. The structural minority of Fundamentalists represents the population disposed to tolerate the usability costs of strong countermeasures.

2026-schramm-analyzing-societal-awareness §4.1.2, §3.3, Figure 10 evaluation

Countermeasures against fingerprinting can paradoxically increase a user's identifiability when deployed by a small population: spoofing-introduced attribute inconsistencies and unusual configurations (e.g., Tor Browser's letterboxing applied to an uncommon monitor size) produce a rarer, higher-entropy fingerprint than baseline. This 'information paradox' means that countermeasure effectiveness is a property of the anonymity set size, not of individual tool configuration.

2026-schramm-analyzing-societal-awareness §2.2 detection

Brave's built-in strict fingerprinting protection was discontinued in January 2024 because of low adoption and frequent web breakages caused by restricting JavaScript APIs; browser vendors are broadly reluctant to implement robust built-in countermeasures for the same compatibility reason. Spoofing-based extensions (HTTP-header spoofers, Canvas Defender) offer only partial protection and can degrade site functionality through overlapping layouts, unreadable fonts, or bot misclassification.

2026-schramm-analyzing-societal-awareness §2.2, §3.1.4 evaluation

Joint multi-task training with a combined loss L_joint = L_site + λ·L_pers shows that increasing λ from 0 to 2 raises mixed-site persona accuracy from approximately 45% to approximately 80% while website accuracy declines only from approximately 90% to approximately 75%, demonstrating a wide regime where an attacker can gain strong persona inference at modest cost to existing WFP capability.

2026-song-personafingerprint-measuring-persona §4.4, §5.7.2, Figure 5 detection

Using only 1,000-packet windows of signed packet lengths and inter-arrival times (no payload, no URLs, no cookies), a passive adversary achieves approximately 84% accuracy at inferring behavioral persona in a mixed-site open-world setting spanning 10 modern websites and 15 canonical personas plus an open-world class. Per-site persona macro-F1 typically ranges from about 0.78 to 0.91 across representative platforms including Bilibili, eBay, Yahoo, Zhihu, and LinkedIn.

2026-song-personafingerprint-measuring-persona §5.3, §5.5, Abstract detection

In open-world evaluation, an average of 34.4% of traffic from unseen personas is misattributed to a specific canonical persona (MisAttr@OW), and misattributions concentrate heavily: on average 58.7% of misattributed windows fall into just the top-3 canonical personas, rising to 66.8% and 69.3% on Bilibili and Zhihu respectively. The classifier correctly rejects unseen personas as OW with an average F1 of only 65.6%.

2026-song-personafingerprint-measuring-persona §5.4, Table 2 detection

Attack accuracy scales steeply with persona-labeled training data: mixed-site open-world persona accuracy rises from 55.0% at 500 windows/persona to 65.0% at 1,000, 76.0% at 2,000, and 84.0% at 5,000 windows/persona across 10 sites (results consistent across 3 random seeds with std ≤1.0%). LLM-driven browsing agents make large-scale persona-labeled traffic generation practical for adversaries.

2026-song-personafingerprint-measuring-persona §5.6, Table 3 evaluation

A site-only WFP encoder trained without any persona labels already encodes substantial persona information: attaching a lightweight MLP probe to its frozen representations recovers persona accuracy roughly 20–30 percentage points above a random-encoder baseline across all 10 sites (e.g., approximately 53% vs. 21% on Amazon, 49% vs. 27% on YouTube, using the same probe architecture and training budget).

2026-song-personafingerprint-measuring-persona §4.5, §5.7.1, Figure 4 detection

Six widely deployed VPN and circumvention tools—OpenVPN, WireGuard/NordLynx, NordWhisper, Orbot (Tor on Android), Lantern, and Psiphon—all failed to block internal IP inference, connection-state detection, and TCP reset injection under identical adversarial conditions on fully patched Android 16. Application-layer obfuscation in Lantern and Psiphon did not prevent TCP-layer disruption; Orbot's VPN-style encapsulation of Tor traffic was bypassed via the same tunnel-level side channels.

2026-tolley-architectural §5.2, §5.4–5.5 evaluation

Positive user reports (reporting successful connections) cannot reliably indicate whether a bridge is accessible: censors possess all technical capabilities of regular users and can submit fabricated positive reports after blocking a bridge. Restricting positive reports to high-reputation users would exclude most honest users early in a bridge's lifetime, when each Tor Lox bridge is initially distributed to only 10 users globally.

2026-vecna-troll-patrol-anonymous §4 defense

Troll Patrol's anonymous negative-reporting protocols, implemented as an extension to the Tor Project's Lox system, impose only a small overhead over existing Lox protocols: Report Submit adds a 1,300-byte request (vs. 877 bytes for Check Blockage) with a 2.72 ms client computation time; existing protocols increase by approximately 100–200 bytes per request due to two new credential attributes. All new protocol timings are sub-5 ms on the client side.

2026-vecna-troll-patrol-anonymous §6, Table 3 defense

Troll Patrol uses zero-knowledge proofs over anonymous credential attributes to track per-user false report counts (f) and pending session state (p) without linking reports across sessions or revealing user identity. A user proves in zero knowledge that f plus any new false reports does not exceed threshold m before submitting; users who exceed m lose reporting ability without any identity disclosure or impact on their reputation within the broader bridge distribution system.

2026-vecna-troll-patrol-anonymous §5.3, §5.4 defense

A plug-and-play Boundary Preserving Aggregation Module (overlapping window partitioning with joint packet- and burst-level features, W=20ms, stride=10ms) consistently improves existing WF baselines without architectural modification: applied to DF, AUC rises from 0.780 to 0.901 and P@5 from 0.315 to 0.545; applied to ARES'25, P@5 rises from 0.869 to 0.900 in the open-world 5-tab setting. The module's consistent gains across all three tested baselines confirm that fixed non-overlapping window segmentation is a structural vulnerability in prior WF pipelines.

2026-yuan-demux-boundary-aware-multi-scale §V-G, Figure 5 detection

DEMUX achieves a P@5 of 0.943 and MAP@5 of 0.961 in the closed-world 5-tab multi-tab website fingerprinting setting, outperforming the strongest prior baseline (ARES'25) by 9.2 and 6.2 percentage points respectively. ARES'25's P@K degrades from 0.900 at 2-tab to 0.851 at 5-tab (a drop of 4.9 pp), while DEMUX improves from 0.926 to 0.943 over the same range, expanding the absolute margin from 2.6 to over 9 points.

2026-yuan-demux-boundary-aware-multi-scale §V-B, Table II detection

In the open-world 5-tab setting — where each trace contains one unmonitored site, substantially increasing noise and class imbalance — DEMUX achieves AUC of 0.998, P@5 of 0.951, and MAP@5 of 0.966, while ARES'25 achieves 0.988/0.869/0.911. DEMUX's advantage widens in the open-world setting (the P@5 gap grows from 2.6 pp to 8.2 pp versus closed-world), confirming that state-of-the-art WF attacks are not defeated by open-world conditions or unmonitored co-browsing traffic.

2026-yuan-demux-boundary-aware-multi-scale §V-C, Table III detection

The Traffic Aggregation Matrix (TAM) representation used by the RF baseline — which counts directional packet counts over fixed time slots rather than tracking per-packet sequences — shows unexpectedly strong robustness under TrafficSliver, achieving P@2 of 0.702, substantially exceeding all other CNN-based methods under that defense. Var-CNN similarly achieves P@2 of 0.826 under TrafficSliver despite mediocre no-defense performance, suggesting that tolerance to partial packet loss is architecturally separable from peak single-observer accuracy.

2026-yuan-demux-boundary-aware-multi-scale §V-D, Table IV evaluation

Under the TrafficSliver defense — which splits traffic across multiple Tor entry nodes so no single observer sees more than a partial fraction of packets — TMWF collapses to a P@2 of 0.399 and ARES'23 to 0.429, while DEMUX retains a P@2 of 0.940, exceeding the next-best competitor by 2.5 points. WTF-PAD and FRONT are substantially weaker defenses, with most methods maintaining near-baseline performance under WTF-PAD.

2026-yuan-demux-boundary-aware-multi-scale §V-D, Table IV evaluation

A Tor relay serving as a Bento server and hosting a CenTor instance for 10,000 simultaneous clients experienced only approximately 0.4% performance degradation compared to a vanilla relay; running 20 concurrent CenTor instances on a single relay caused roughly 1% degradation. Shadow-aware routing in a low-relay-density region (US, 933 of 6,666 relays in shadow) produced 0.7% relay degradation while delivering 3.6% client performance improvement.

2025-arora-improving-performance-security §5.2 evaluation

CenTor, a CDN for Tor onion services deployed as a Bento network function, achieved approximately 56.4% reduction in download time compared to standard Tor under a geographically-aware shadow configuration, while location-unaware CenTor (reducing hops without shadow restriction) achieved a 34.5% reduction. The key mechanism is eliminating the server-side 3-hop circuit by deploying replicas in non-anonymous mode, cutting the default 6-hop onion service path to 3 hops.

2025-arora-improving-performance-security §5.1 evaluation

CenTor protects origin onion service operators from DoS and deanonymization by routing all client traffic through geographically distributed Bento replicas running inside SGX-based Trusted Execution Environments (TEEs). The original operator can go fully offline after deploying static content; replicas enforce confidentiality and integrity of hosted content with ephemeral per-enclave encryption keys, preventing malicious Bento node operators from inspecting or modifying content even if they control the underlying hardware.

2025-arora-improving-performance-security §3.1, §5.3 defense

CenTor's anonymity scoring function quantifies the privacy cost of geographic shadow selection using six parameters (client density, AS-level and country-level entropy, relay density, exit density, guard density). Prior work establishes that reducing the client anonymity set by 20x—retaining at least 5% of total Tor users—still provides strong anonymity; accordingly, CenTor recommends minimum thresholds of CD, EL, EC ≥ 0.05 and RD, ED ≥ 0.2 for safe shadow operation.

2025-arora-improving-performance-security §4.1, §4.2, Table 2 defense

Interactive communication on Tor incurs latencies more than 5x greater than direct Internet paths; onion services compound this by creating a 6-hop circuit (3 client-side plus 3 server-side). Shadow simulations at 100% Tor network scale (752,338 active clients, 6,666 nodes) showed that deploying 3,500 and 6,000 Bento servers caused only 4.4% and 9.6% client-side performance degradation respectively, demonstrating that programmable middlebox overlays are feasible at Tor scale.

2025-arora-improving-performance-security §1, §5.2 evaluation

The Ahmia search engine provided the most onion addresses (18,069 in a single day, ranging 18,000–22,000 week-to-week), outperforming five other sources combined (36,028 total across six engines). However, Ahmia's intentional exclusion blacklist contains 46,000+ hashed addresses, and crawling onion services for 20 days yielded 48,745 unique v3 addresses, 11,809 of which were on Ahmia's blacklist — meaning any index-based collection systematically misses a significant share of the onion ecosystem by design.

2025-h-ller-evaluating §4.1, §4.3 evaluation

Combining six onion search engines/repositories plus clearnet search engines, Tor2web-style DNS leakage, and 20 days of self-run crawling (2.9 million pages), the authors assembled 482,614 unique v3 onion addresses — the largest known collection. Verifying against HSDir blinded public keys showed the collected addresses accounted for 25% of observed blinded keys but were responsible for 66% of all successful service descriptor downloads, confirming a heavy-tailed usage distribution.

2025-h-ller-evaluating §4, §5 evaluation

Strict anonymity requirements in Tor onion services render conventional DoS mitigation strategies inapplicable, forcing the Tor community to revive the client-puzzle approach as the only feasible admission-control mechanism compatible with unlinkability guarantees.

2025-lee-onions-got-puzzled Abstract defense

OnionFlation attacks succeed by inflating puzzle difficulty without causing detectable congestion at the targeted service, meaning the attack leaves no noticeable traffic-volume signature at the victim — standard congestion-based anomaly detection cannot identify the attack in progress.

2025-lee-onions-got-puzzled Abstract detection

The OnionFlation attack family artificially inflates the required client-puzzle difficulty for all connecting clients without causing noticeable congestion at the targeted onion service, rendering any existing onion service largely unusable at an attack cost of a couple of dollars per hour.

2025-lee-onions-got-puzzled Abstract evaluation

The paper offers practical guidance for Tor onion service operators aimed at balancing mitigation of OnionFlation attacks against maintaining service availability for legitimate clients, recognizing that no complete solution exists within the current puzzle architecture.

2025-lee-onions-got-puzzled Abstract defense

The Tor client puzzle mechanism contains a fundamental architectural trade-off: the system is forced to choose between inflation resistance (preventing attackers from artificially raising puzzle difficulty) and congestion resistance (preventing the service from being overwhelmed), but cannot achieve both simultaneously — a root-cause vulnerability acknowledged by the Tor Project.

2025-lee-onions-got-puzzled Abstract evaluation

Client-puzzle DoS mitigation has been adopted in an official Tor protocol update and is in active use by several major onion services. An ethical live-network evaluation of OnionFlation attacks confirmed the vulnerability on the production Tor network, and the Tor Project has acknowledged the findings. The mechanism was revived and shipped by the Tor community in response to real-world DoS attacks against onion services.

2025-lee-onions-got-puzzled Abstract deployment

PWA-based circumvention tools that display their name or any identifying string in the browser URL bar or page title expose that identifier to all six Chinese browser vendors' telemetry servers, since all six browsers collect page titles and full URLs. Browser SDKs with READ_PHONE_STATE and elevated permissions can monitor PWA activity at the OS level in ways not possible with standard browsers, making browser selection as security-critical as the circumvention tool itself for the Tor Browser threat model.

2025-rodriguez-revisiting §5, §6 defense

CenPush is implemented and evaluated specifically for Tor bridge distribution, replacing the existing polled bridge-line fetching with push delivery. The design is presented as a general mechanism applicable to any circumvention tool that needs to push fresh proxy addresses to clients — not just Tor bridges — whenever censors block the tool's normal update channel.

2025-sharma-cenpush §3, §5 deployment

Scanning 0.91B unique SANs extracted from 3.7B certificates across 17 CT logs revealed 3,330 unique .onion addresses configured by 26,937 domains. After six months, only 2,101 onions (63%) remained reachable, of which 1,505 (72%) had matching clearnet index pages, constituting the effectively enumerable target set for a targeted OLF adversary.

2025-syverson-onion-location-measurements-fingerprinting §4.1 evaluation

Local onion association—periodically downloading the full set of onion associations from a CT-log-based API and performing each lookup locally—produces a traffic pattern from the guard's perspective that is indistinguishable from generic onion service access, eliminating both the OLF fingerprint and the DNS-based Website Oracle attack vector. This approach requires no per-connection clearnet exit circuit and imposes negligible overhead given the current ~1,500 stable O-L site count.

2025-syverson-onion-location-measurements-fingerprinting §2.2.4, §6 defense

OLF reduces an adversary's target anonymity set from roughly 10,000 active onionsites to the ~1,500 stably available O-L sites—nearly an order of magnitude. Because O-L requires an exit circuit with a DNS lookup, a DNS-based Website Oracle further collapses the false-positive rate, making OLF effectively a closed-world attack on the enumerated O-L site list.

2025-syverson-onion-location-measurements-fingerprinting §2.3, §3 detection

Circuit fingerprinting from a guard-relay position achieves ≥99.9% accuracy with FPR ≤0.1% for all four Tor circuit types (general, HSDir, introductory, rendezvous) using the Deep Fingerprinting classifier on the first 512 cells, despite Tor's deployed partial defenses. Onion-Location fingerprinting (OLF) combining these circuit classifiers then achieves 98.81–99.87% accuracy (FPR 0.16–1.23%) distinguishing O-L sessions from ordinary clearnet or onion-only visits.

2025-syverson-onion-location-measurements-fingerprinting §5.1–5.2 detection

Automatic Onion-Location redirect was disabled in Tor Browser 13.0.12 as a direct result of this research, because automatic redirect forces the distinguishable clearnet-then-onion circuit pattern on every visit without user awareness. Manual O-L remains in Tor Browser but is still fingerprintable with the same near-perfect accuracy since the exit→onion circuit sequence is identical whether the redirect is automatic or manually triggered.

2025-syverson-onion-location-measurements-fingerprinting §2.2, footnote 1; §6 evaluation

In laboratory benchmarks, the best UPGen-generated protocol achieves 252 ms TTFB latency (vs 212 ms Obfs4, 313 ms TLS) and 4.25 Gbit/s throughput per core (vs 4.65 Gbit/s Obfs4, 9.42 Gbit/s TLS). The worst-case UPGen protocol (4.5 RTT handshake) reaches 677 ms TTFB but 3.70 Gbit/s throughput. In large-scale distributed Tor simulations, the choice of UPGen protocol had no statistically significant effect on end-to-end Tor flow performance.

2025-wails-censorship §5, Table 7 evaluation

Extrapolating empirical FPRs using Wang's base-rate-adjusted precision formula (𝜋_r), the best HTTPS-only approach can sustain precision 0.5 at recall 0.5 only up to a world size of 37.5M videos; precision 0.1 at recall 0.5 extends to 337.5M — still short of YouTube's ~10 billion catalog (Table 3). For Tor, the corresponding limits are 4.8M and 42.9M, making dragnet surveillance of unselected users on large platforms effectively infeasible at any acceptable precision with current techniques.

2025-walsh-improved-open-world-fingerprinting §4.5.1, Tables 3–4 evaluation

When a fingerprinting model is trained on traffic collected from one geographic vantage point and tested on traffic from a different continent, the HTTPS-only open-world FPR at 0.5 recall increased by factors ranging from 2.8x (EU-West-2) to 50.3x (Africa) relative to the same-vantage baseline — despite 60-way closed-world accuracy remaining above 0.99 across all vantage-point pairs (Table 5). For Tor traffic the effect was weaker but still reached 25.2x (Asia-Pacific Southeast-1), showing path diversity also disrupts Tor-based fingerprinting.

2025-walsh-improved-open-world-fingerprinting §5.1, Table 5 evaluation

The paper establishes, for the first time in a large open-world scenario (64,000 unmonitored test videos), that HTTPS-only video stream fingerprinting is significantly easier than Tor-based fingerprinting because DASH adaptive bitrate selection introduces a second-order network-condition effect: clients request entirely different video segments at different quality levels depending on path conditions, causing traffic traces from different geographic vantage points to diverge at the application layer even when network conditions are nominally similar. This makes NOTA and synthetic training sample techniques less effective on Tor data due to inherent trace noisiness.

2025-walsh-improved-open-world-fingerprinting §4.5, §5.1 evaluation

Tor provides substantial and measurable protection against video stream fingerprinting: the best-case FPR at 0.5 recall is 0.0000063 for Tor versus 0.0000008 for HTTPS-only connections, roughly an 8x increase. Translating to world sizes, at 0.5 recall and 0.1 precision the maximum viable platform catalog is 42.9M videos over Tor versus 337.5M over HTTPS-only (Tables 3–4), confirming Tor degrades adversary capability even after an assumed prior website-fingerprinting step that identifies video platform visits.

2025-walsh-improved-open-world-fingerprinting §4.5.1, Tables 3–4 evaluation

Snowflake has been deployed in Tor Browser and Orbot for several years and served as a significant circumvention tool during the Russia 2021 network disruptions and Iran 2022 protests. The paper documents a history of deployment and blocking attempts, providing empirical evidence that the ephemeral WebRTC proxy design has sustained availability under real censor pressure across multiple high-profile events.

2024-bocovich-snowflake §1, §6 deployment

Against the state-of-the-art DeepCoFFEA flow-correlation attacker, FC-DeTorrent reduces the true positive rate at a 10^-5 false positive rate to approximately 0.12 — less than half that of the next-best defense Decaf (TPR ≈ 0.29) — while using 97.3% bandwidth overhead, without delaying any real traffic packets.

2024-holland-detorrent §4.2, §6, Table 5 evaluation

DeTorrent exhibits strong diminishing returns in the bandwidth-performance tradeoff: increasing the dummy-download budget from N=1,000 to N=3,000 reduces Tik-Tok accuracy by ~19.1 percentage points, while a further increase from N=5,000 to N=7,000 yields only an additional 4.9-point reduction (accuracy floor near 20.8% at ~210% overhead). At the lowest tested budget (~40% overhead) Tik-Tok accuracy is still only 52.8%.

2024-holland-detorrent §6.2, Figure 6 evaluation

DeTorrent is implemented as a Tor pluggable transport on top of the WFPadTools/Obfsproxy framework and deployed against live Tor traffic; a modest VPS with 4 GB RAM and 2 vCPUs running at under 50% CPU utilization can defend five simultaneous connections in real time with no GPU required. Performance drops only 0.7% when the generator is trained on one dataset partition and tested on another.

2024-holland-detorrent §5.4 deployment

DeTorrent reduces closed-world Tik-Tok attack accuracy from 93.4% to 31.9% on the BE dataset — 10.5 percentage points better than the next-best padding-only defense (FRONT at 42.4%) — and reduces Deep Fingerprinting accuracy from 94.3% to 30.0%, at a bandwidth overhead of 98.9%. On the larger DF dataset, Tik-Tok accuracy falls from 97.7% to 79.5%.

2024-holland-detorrent §6.1, Table 4 evaluation

The authors propose a 'shim' pluggable transport that splits client traffic across N PT connections using unmodified existing PT bridges as proxies and a gateway bridge that correlates streams back into a Tor circuit via the Turbo Tunnel reliability pattern. This architecture enables all existing and future PTs to benefit from traffic splitting without modifying each PT's client or server code individually.

2024-lorimer-extended §3 defense

When a user splits traffic across N paths, a censor observing a single path sees only a partial trace, substantially reducing the accuracy of classifiers trained on complete network traces. Prior Tor traffic-splitting work (TrafficSliver, CoMPS, multipath Tor studies) has validated this defense against website fingerprinting outside the PT context.

2024-lorimer-extended §1, §4.2 defense

The first multi-perspective study of the circumvention-tool ecosystem surveyed 12 leading CT providers collectively serving over 100 million users, plus CT users in Russia and China. Beyond technical blocking challenges, the study found that funding constraints, usability problems, misconceptions (users and providers hold inaccurate beliefs about each other's capabilities), and misbehaving players (tools operated by adversarial actors) are equally significant threats to the ecosystem's health — and are largely unaddressed by the academic research community.

2024-xue-bridging Abstract, §4, §6 evaluation

DeTorOS enables provable geographic avoidance for Tor onion services by running a TEE-backed Bento function as a trusted middlebox: both the client and the onion service upload their respective 3-hop circuit halves to this enclave, which computes the never-once or never-twice avoidance proof without revealing either party's circuit to the other.

2023-arora-detor-onion §3 defense

Computing a never-once avoidance proof for a 6-hop onion-service circuit takes an average of 64.85 seconds — incurred once at connection setup — because the system must collect round-trip timing measurements across all six relays before running the geographic proof; SGX execution overhead is nominal, and the paper notes that lower-RTT circuits (more likely to be DeTorOS-compliant) reduce subsequent data-transfer latency.

2023-arora-detor-onion §5.3 evaluation

Never-twice provable avoidance succeeds for 72.4% of sampled source-destination pairs on 6-hop onion-service circuits, compared to approximately 98% on the original 3-hop DeTor circuits; the degradation arises because the additional hops increase round-trip time, making it harder to rule out forbidden-region traversal via speed-of-light bounds.

2023-arora-detor-onion §5.2 evaluation

DeTorOS's security relies on the honest-but-curious model: if the onion service refuses to participate or lies about its circuit, the client receives no avoidance guarantee. The paper explicitly flags this as an open limitation and notes it cannot be closed without either requiring a TEE on the onion service side or fundamental protocol changes.

2023-arora-detor-onion §6 defense

Tor's built-in country-exclusion mechanism is unreliable: circuits configured to exclude US Tor nodes only actually bypassed the US 12% of the time, motivating provably-avoidant circuit construction.

2023-arora-detor-onion §1 evaluation

Following the invasion, Psiphon user counts and VPN usage in Russia increased many-fold and correlated with specific censorship events, while multiple access paths to Tor (direct connections, bridges, pluggable transports) were progressively blocked. Despite this surge, circumvention tools reached only a small fraction of all Russian Internet users, indicating that aggressive multi-vector blocking and lack of user awareness left most people unable to access censored resources.

2023-ramesh-network §6 evaluation

Relying on third-party email providers to verify users was demonstrated by Ling et al. to leave Tor's BridgeDB vulnerable to censors capable of creating multiple accounts, enabling bridge enumeration via sock-puppet attacks at scale. Active and passive detection techniques — including traffic flow analysis, DPI, website fingerprinting, and active probing — have been demonstrated in prior work to reveal Tor bridges, making Tor inaccessible for the majority of users in some regions.

2023-tulloch-lox §2.2.1 detection

Switching source IP via VPN, Tor, or HTTP proxy is the primary victim-side mitigation because residual censorship is tuple-keyed; however, if the proxy entry node's path also crosses the censor, the attacker can redirect the attack at the proxy itself. On the censor side, null-routing middleboxes could eliminate the vulnerability by validating TCP sequence/acknowledgment numbers before dropping traffic, or by replacing null routing with an explicit block-page response.

2021-bock-your §VII defense

Survey data indicates 31% of Chinese Internet users use VPN services compared to Tor's approximately 2 million daily users globally, and centralized non-anonymous systems like Lantern and Psiphon dominate adoption over anonymity-focused tools. The paper argues this demonstrates that the majority of censored users prioritize blocking resistance over anonymity, supporting a separation-of-properties design principle.

2020-nasr-massbrowser §III-B policy

Anonymization and circumvention tools (VPNs, Tor, etc.) are among the three most commonly blocked content categories across all commercial filters surveyed, alongside pornography and gambling. This holds across diverse products including Fortinet, Cisco, and government-deployed firewalls in Iran, Saudi Arabia, and Bahrain.

2020-raman-measuring §V-A-1, Fig. 7 evaluation

Meek over Azure CDN successfully established Tor circuits from China in all tests; meek over Amazon was inconsistent and often failed mid-circuit. Meek requires TLS on the bridge — without it the GFW blocks the bridge within minutes and purges it from the blacklist, suggesting a separate meek-specific detection and blocklist is maintained.

2018-dunna-analyzing §5.1 evaluation

obfs4 successfully established Tor circuits on the authors' own unpublished bridge relays but failed to connect to any public obfs4 bridge, consistent with the GFW having scraped and blacklisted public bridge addresses. This demonstrates that address confidentiality is a prerequisite for obfs4's effectiveness, independent of its obfuscation properties.

2018-dunna-analyzing §5.1 evaluation

Because a disproportionate number of Tor exit nodes are located in the EU, GDPR-motivated blanket blocking of EU IP ranges creates collateral access restrictions for Tor users globally. This illustrates that privacy-protective legislation and censorship-circumvention infrastructure can have directly competing effects when server-side enforcement is implemented via coarse geographic IP filtering.

2018-tschantz-bestiary §7 policy

Across all tested countries, circumvention and anonymization tools are the most consistently blocked category: www.hotspotshield.com is blocked in 5 of 13 detected censoring countries, and three Tor Project properties (bridges.torproject.org, www.torproject.org, ooni.torproject.org) each appear in the top-10 most broadly blocked domains. Collateral damage is also documented — Iran blocks psiphonhealthyliving.com as a substring match for the psiphon.ca circumvention domain.

2018-vandersloot-quack §6.2, Figure 12 evaluation

The paper identifies that Shadowsocks can also serve as a transport layer for Tor and VPN connections, meaning a Shadowsocks flow detector functions as a first-stage classifier that unmasks compounded anonymity systems. The authors explicitly cite this as a motivation for detection.

2017-deng-random §II detection

Approximately 10% of respondents (n=23) held uncertain or incorrect beliefs about which actor was responsible for a given block, systematically conflating government censorship with geoblocking, paywalls, and platform-side restrictions. This misidentification cascaded into inappropriate tool selection and inaccurate risk assessment: users who could not distinguish state blocking from licensing restrictions could neither choose the right circumvention tool nor accurately gauge the legal jeopardy of accessing the content. Respondents specifically requested a pre-visit blocking-actor classification tool.

2017-gebhart-internet §5.2.3, §5.4.1, §6.2.1 evaluation

Of 229 Thai Internet users surveyed, 63% (n=144) had attempted to circumvent censorship, and of those, roughly 90% (n=132) reported success using VPNs (32.64%), proxies (32.64%), or Tor (23.61%). Failures were isolated to proxies (n=2), VPNs (n=2), and alternative searches (n=3), indicating that existing circumvention tools were technically adequate but that availability and comprehensibility—not raw capability—were the binding constraints on user success.

2017-gebhart-internet §5.2.1, Table 2 evaluation

Relay-based circumvention severely degrades ad relevance: across Alexa top-500 uncensored sites, the overlap between ad sets fetched via Tor and the direct-path ground truth averaged only 28%, with near-zero overlap for sites serving geo-targeted ads. For blocked sites, only ~16% of ads shown via Tor were in the user's language.

2017-javaid-online §3.2, Table 3 evaluation

ADVENTION's split-path design — fetching publisher content via relay and ad requests via the direct path — raises average ad-set overlap from 28% (Tor) to 70%; combining ADVENTION with Intelligent Relay Selection (language-matched relay) further increases average overlap to ~80%. For blocked sites, ADVENTION with IRS raised ad relevance from ~16% to 100%.

2017-javaid-online §5.1, §5.3, §5.4 evaluation

ADVENTION provides up to 47% improvement in average page load time (PLT) compared to Tor, because ad requests — which are often on the critical rendering path — are served over the direct channel rather than through the relay. The exact improvement depends on webpage structure and bottleneck resources.

2017-javaid-online §5.1, Figure 4 evaluation

The authors recommend 'smart automation' for bridge selection: the client first connects via a hard-to-censor bridge, then contacts a central Tor server over that Tor connection to identify the best available bridge for the user's location and network conditions, then reconnects using that bridge — eliminating the manual trial-and-error that caused 79% of attempts to fail. This is contrasted with 'naive automation' (sequential blind retry) which avoids UI friction but wastes time on non-working bridges.

2017-lee-usability §8 defense

Participants spent 64–78% of their total connection time on the progress/waiting screen (not in the configuration UI), and the simulated censorship environment was the dominant predictor of connection time (Kruskal–Wallis χ² = 80.5, df = 2, p < 10⁻¹⁵). In E3, each failed bridge attempt added several minutes of timeout before the user could retry, compounding the overall latency.

2017-lee-usability §7.3.3, Table 4 evaluation

79% of total user attempts (363 of 458) to connect to Tor in simulated censored environments failed. In the most heavily censored condition (E3, requiring a meek or custom bridge), only 50% (10/20) of participants using the original interface connected, and even with the redesigned interface only 68% (13/19) succeeded within 40 minutes.

2017-lee-usability §7.3, Table 2 evaluation

A redesigned Tor Launcher interface significantly increased success rates (Pearson χ² = 2.808, p < 0.047) and reduced median connection time in E3 from 40:08 to 20:25 (Mann–Whitney Z = −1.84, p < 0.0328, r = 0.172); configuration time also dropped significantly (Z = −3.28, p < 0.0005, r = 0.307). Changes included eliminating yes/no bridge and proxy question screens, adding auto-detection for proxies, consolidating options, and surfacing meek bridges as a fallback recommendation.

2017-lee-usability §6.2, §7.3.1–7.3.3, Table 2 defense

DeTor circuits have significantly lower end-to-end RTTs than standard Tor circuits because high-RTT paths cannot satisfy avoidance proofs, effectively self-selecting for shorter routes. Bandwidth distributions are similar to standard Tor. However, intentional packet-delay defenses proposed for Tor (to defeat timing attacks) would increase effective δ and reduce DeTor proof coverage, creating a tension between delay-based anonymity defenses and RTT-based geographic avoidance.

2017-li-detor §6.2.3, §6.2.1 (δ discussion) evaluation

Never-once avoidance succeeds for 75% of source-destination pairs that do not already terminate in the US (a highly routing-central country) at δ=0.5, and for nearly all pairs avoiding less central countries. Russia is the hardest case at ~35% success (δ=0.5) due to proximity to the dense European node cluster. The median successful source-destination pair has over 1,000 valid DeTor circuits when avoiding the US and 500 when avoiding China.

2017-li-detor §6.2.1, §6.2.4 evaluation

Never-twice avoidance — ensuring no country appears on both the entry leg (source→entry) and exit leg (exit→destination) of a Tor circuit — succeeds for 98.6% of source-destination pairs not in the same country, using only client-side RTT measurements. This directly defeats traffic-correlation deanonymization attacks that require an adversary on both legs of the circuit simultaneously.

2017-li-detor §4.2, §6.3.1 defense

DeTor proves geographic avoidance using speed-of-light RTT constraints rather than Internet topology maps. If the measured end-to-end RTT satisfies (1+δ)·Re2e < Rmin, where Rmin is the theoretical minimum RTT that would include any point in the forbidden region, then packets provably could not have traversed that region — even against adversaries who forge traceroute and BGP responses.

2017-li-detor §3.1, §4.1 defense

Tor's built-in country-exclusion feature provides only the illusion of control: among circuits configured to exclude the US, only 12% could be identified as definitively avoiding US territory. The remaining 88% of 'trusted' circuits fail to deliver a proof of avoidance, meaning standard Tor policy and provable security diverge sharply.

2017-li-detor §1, §6.2.1 evaluation

Measured packet loss rates under GFW censorship (Feb–Apr 2017, client at Tsinghua University/CERNET): Tor with meek obfuscation suffers 4.4% average PLR; Shadowsocks (AES-256-CFB) suffers 0.77% PLR; native VPN (PPTP/L2TP) and OpenVPN both achieve ~0.21% PLR. For comparison, the same tools accessed from a US vantage point show PLR below 0.1%, confirming the excess loss is GFW-induced. The GFW's DPI and active probing techniques specifically target Tor and Shadowsocks protocol signatures.

2017-lu-accessing §4.3 evaluation

Bridges that carry clients are highly stable: their median lifetime is 116 days (~4 months) and 84% never change IP address, with 90% having at most one IP change. This means current censor policies that remove bridge IP blocks every 25 hours are far more conservative than necessary — an adversary could sustain blocks for months without significant collateral damage.

2017-matic-dissecting §V-B, Figure 3 evaluation

Default bridges — whose IP addresses are hardcoded in the Tor Browser Bundle — carry 91.4% of all bridge clients globally in April 2016, and 86.1% in Iran and 69.2% in Syria. Because these addresses are trivially obtainable from the Tor Browser Bundle configuration files, a censor can block the vast majority of bridge users in a country at any time.

2017-matic-dissecting §V-E, Table II evaluation

Four OR ports (443, 8443, 444, 9001) account for 82% of all active public bridge fingerprints as of April 2016, down from 95% in March 2013 but still concentrated. Scanning just three of these ports (443, 8443, 9001) is sufficient to deanonymize 71% of all active public bridges. Additionally, CollecTor's published per-bridge usage statistics allow a censor to rank bridges by client count per country and identify the highest-impact OR ports to scan next.

2017-matic-dissecting §V-D, Figure 6, Table IV detection

Tor's vanilla TLS certificate presents a distinctive pattern (SubjectCN=www.[random].com; IssuerCN=www.[random].net using base32 random strings), which never changes across certificate rotations every 2 hours. Using this pattern against Censys and Shodan scan data without running any active scans, the authors discovered 694 private bridges and 645 private proxies, and deanonymized the IP address of 35% of public bridges with clients (23% of all active public bridges) in April 2016.

2017-matic-dissecting §II-B, §VI-A, Table V detection

CloudFlare platform policy creates outsized blocking: 80% of CloudFlare-hosted websites discriminate against at least 60% of studied Tor exits, while Amazon- and Akamai-hosted sites show high policy diversity. Social networking and shopping sites are the most aggressive discriminators — 50% block over 60% of studied exits — while search engines are least aggressive, with 83% blocking fewer than 20% of exits.

2017-singh-characterizing §6.4 evaluation

Conservative exit policies (Reduced-Reduced, which additionally blocks SSH, Telnet, and IRC ports beyond the default) have no statistically significant correlation with IP blacklisting rates or abuse complaint volume. Web-traffic accounts for 98.88% of all connections on Reduced-Reduced exits, confirming that ports 80/443 are the primary abuse vector and that port-restriction does not meaningfully reduce exposure.

2017-singh-characterizing §5.5 evaluation

7% of 84 commercial IP blacklists proactively blacklist Tor exit relay IPs as a matter of policy: the Snort IP and Paid Aggregator blacklists listed newly deployed relay IPs within 3 hours of their first appearance in the Tor consensus and maintained the listing for the entire relay lifetime. In total, 88% of all Tor exits appear on at least one commercial blacklist, compared to 9% of VPNGate and 69% of HMA VPN endpoints.

2017-singh-characterizing §5.2, §5.3, §5.6 detection

Real Tor users browsing the Alexa Top 1M websites via deployed exit relays experience failed HTTP requests at rates of 15.8–33.4% and failed HTTPS handshakes at rates of 35.0–49.6%, representing severe service degradation compared to non-Tor browsing (Table 8).

2017-singh-characterizing §7, Table 8 evaluation

20.03% of Alexa Top 500 website front-page loads showed discrimination against Tor exit users. Exercising search functionality on compatible sites raised discrimination by 3.89% (to 21.33%), while exercising login functionality raised it by 7.48% (to 24.56%), demonstrating that headless front-page-only crawlers significantly underestimate the true blocking rate Tor users face.

2017-singh-characterizing §6.4, Table 6 evaluation

Tor usage in Turkey spiked sharply during the initial days of the July 2016 coup—when ISPs were actively throttling Twitter—but declined steadily in subsequent months back toward pre-coup baselines, consistent with post-coup suppression being driven by chilling effects rather than sustained network-level blocking.

2017-tanash-decline §4.1, Figures 2–3 evaluation

A university closed survey of 64 Pakistani users found that 51% evade censorship using VPNs (Hotspot Shield being the most prominent), 25% use web proxies, 17% use Tor/onion routing, and approximately 7.2% use CDNs, mirror sites, search-engine caches, or web-based DNS lookup services.

2016-aceto-analyzing §V-B evaluation

China's firewall never blocked a bridge before its public Tor Browser release, despite bridges being discoverable earlier via bug-tracker tickets and source code commits. The four bridges distributed only in Orbot (not Tor Browser) remained unblocked throughout the experiment, indicating the GFW monitors end-user software releases rather than upstream repositories or alternative distribution channels.

2016-fifield-censors §4 Results detection

The Great Firewall of China blocked newly published obfs4 Tor Browser default bridges after delays of 7, 2, 18, 11, and 36 days following the first public software release, and up to 57 days after bridges were first discoverable via bug-tracker ticket filing. Iran showed no blocking of the same default bridges across the entire five-month measurement period.

2016-fifield-censors §4 Results, Table 2 evaluation

Winter and Lindskog [157] (2012) documented that the GFW used TLS SNI inspection in combination with IP/port filtering and TCP disruption to block Tor, as recorded in the survey's Table 1. This is one of the earliest published accounts of the GFW applying SNI-based blocking specifically to a circumvention protocol, demonstrating that the GFW correlated multiple detection signals rather than relying on any single technique.

2015-aceto-internet Table 1 detection

The GFW sends protocol-specific probe payloads tailored to each circumvention tool: Tor bridges receive a TLS ClientHello mimicking Tor's own; obfs2/obfs3 servers receive random-looking payloads; Shadowsocks servers receive random bytes. A server that responds differently to these crafted probes versus innocent traffic (e.g., by sending a valid protocol handshake in response to a probe) reveals itself and is subsequently blocked.

2015-ensafi-active-probing §5 detection

The GFW blocks Tor primarily by dropping SYN/ACK segments entering China from blacklisted IP/port pairs, not by dropping SYN segments leaving China. Of 142,802 CN→Tor-Relay measurements, 81.52% were Server-to-client-dropped versus only 0.55% Client-to-server-dropped. Blocking Tor directory authorities also showed substantial Client-to-server drops (19.61%), suggesting authorities may be treated differently.

2015-ensafi-analyzing §4.1, Table 1 detection

In experiments using 200 back-to-back fetches of the YouTube homepage (≈360 KB), HTTPS produced lower page load times than Tor in most cases because Tor circuits do not optimize for performance and often select longer paths. Tor's page load times varied widely as circuits changed approximately every 10 minutes, producing a heavy tail in the latency distribution.

2015-nisar-case §3.2 / Figure 2a evaluation

CloudTransport Cirriform in tunnel and proxified-Tor modes achieves performance comparable to Tor with Obfsproxy across Web browsing (Alexa Top 30 front pages), 300 KB SCP uploads, 10 MB YouTube uploads, and 5-minute 480p streaming video. Bandwidth overhead per message is 350–400 bytes for Amazon S3, with HTTPS adding an extra 2–3% overhead. Per-page browsing costs are as low as $0.00100¢ (Cumuliform on S3), with idle-polling costs of $0.185/day plus $0.34/day/connection for Cirriform on S3.

2014-brubaker-cloudtransport §5, Table 3, Table 4 evaluation

Port 9001 (Tor) ranked third among all blocked ports in Syria, behind only ports 80 and 443. Proxy SG-48 was responsible for a disproportionate share of Tor censorship — blocking Tor traffic for multiple consecutive days — while other proxies in the same deployment did not, indicating per-proxy policy specialization or traffic steering of suspected circumvention flows to dedicated blocking infrastructure.

2014-chaabane-censorship §4 (Ports), §5.2 detection

The GFW blocks Tor primarily via stateless SYN/ACK dropping based on the server's source IP address and port (server-to-client direction, 73.04% of CN,Tor-dir cases). Two specific Tor directory authorities account for 98.8% of client-to-server (null-routed) blocks and 72.7% of error cases, indicating selective deeper blocking of specific IP addresses beyond the common return-path filter.

2014-ensafi-detecting §5 detection

Over 5 days of measurement, 73.04% of connections from Chinese clients to Tor directory servers were blocked server-to-client (stateless SYN/ACK dropping), 16.73% were blocked client-to-server (null routing), and only 0.63% were unblocked. Of all censored Tor directory server connections measured across all regions, 98% originated from Chinese clients.

2014-ensafi-detecting §5, Table 1 evaluation

The paper sketches a decentralized DHT-based communication protocol where all payloads are encrypted in TLS and explicit redirection enables a form of onion routing. Because the censor cannot distinguish censored from non-censored streams, it is forced into a binary choice: block all protocol traffic (overblocking) or allow all of it.

2014-tan-censorship §2.2 defense

Comprehensive Internet-wide scanning enables cross-IP tracking of users and devices by correlating stable cryptographic identifiers — TLS certificates or SSH host keys presented by home routers and cable modems — with public geolocation data across DHCP lease changes, defeating the anonymity assumption behind dynamic IP addresses.

2013-durumeric-zmap §4.6 detection

An FTE-tunneled Tor circuit using intersection, manual, and auto HTTP formats successfully traversed the Great Firewall of China from a VPS inside China to a server in the United States on port 80. A persistent tunnel polling a censored URL every five minutes remained active for one month until VPS account termination, with no blocking observed.

2013-dyer-protocol §6 deployment

Default Tor connections to a private bridge inside China were detected by the Great Firewall via active probing: an initial connection succeeded, followed by a probe from a Chinese IP address approximately 15 minutes later that performed a TLS handshake and then blacklisted the (IP, port) combination. Subsequent connection attempts resulted in a successful SYN followed by spoofed TCP RSTs terminating both the client and bridge connections.

2013-dyer-protocol §6 detection

Pseudonymity is insufficient for dissent networks: social-network profile information can be correlated with external data to deanonymize users, and fixed-infrastructure networks enable localization attacks even without explicit identity. The authors argue that true anonymity—or at minimum strong deniability where usage is non-incriminating and activity is difficult to trace—is required to protect participants.

2013-hasan-building §5.2 defense

Existing censorship-resistant systems share a fundamental vulnerability: they require the user to know a finite set of entry points (bridge addresses, rendezvous points, or ISP-level collaborators) that a censor can enumerate by impersonating a legitimate user. China has blocked the majority of Tor bridges since 2010 and Iran blocked all encrypted traffic in 2012, demonstrating this attack is operationally deployed at scale.

2013-invernizzi-message §1 detection

Tor, which has minimal commercial footprint and a distinctive network signature, was blocked throughout China using tailor-made GFW countermeasures and lost approximately 85% of its Chinese users as a result. In contrast to GoAgent and VPNs, China's censors can block Tor without significant economic collateral damage, making it uniquely vulnerable despite its strong privacy properties.

2013-robinson-collateral §2 Tool Usage, §3 Collateral Freedom Matters evaluation

In a DHT-based censorship-resistant name system, poisoning attacks (injecting invalid mappings) are neutralized by requiring signature verification on stored values; eclipse attacks (isolating specific mappings from the network) require replication across multiple DHT nodes. Critically, decentralizing lookups from a single ISP resolver to a DHT shifts query visibility from ISPs to arbitrary peers, requiring per-query encryption keyed to secrets known only to the querying client to limit adversaries to confirmation attacks.

2013-wachs-feasibility §4.4 defense

Knowing a user's bridge assignment narrows the adversary's anonymity set to the small group sharing that bridge, deanonymizing Tor users even when the bridge itself is not compromised; rBridge addresses this using 1-out-of-m Oblivious Transfer, Pedersen commitments, and non-interactive zero-knowledge proofs so the bridge distributor learns nothing about which bridges a user holds.

2013-wang-rbridge §5, §5.1 defense

Tor's traffic contains a characteristic prevalence of 586-byte packets (Tor's 512-byte cells plus TLS header overhead) that form a strong flow-level fingerprint detectable from a few dozen captured packets. ScrambleSuit's packet length morphing eliminates this signature and shifts the distribution toward MTU-sized packets, but the authors note that a censor using the VNG++ classifier — which relies on coarse features like connection duration, total bytes, and burstiness — would still require only a marginal increase in ScrambleSuit's overhead to defeat.

2013-winter-scramblesuit §4.3.1, §5.1, Figure 10 detection

Flash proxies successfully relayed Tor traffic from within China in December 2011, but the test relied on a simple HTTP-based rendezvous blockable by IP address; the authors identify rendezvous — getting just a few bytes (the client's IP address) out of the censored region — as the bottleneck that determines whether the entire proxy system remains operational.

2012-fifield-evading §4.4, §6 deployment

Flash proxies provide mean throughput of 79.7 KB/s when uninterrupted — comparable to direct Tor (69.5 KB/s) — but throughput drops to 56.6 KB/s (20–40% lower) when proxies alternate on 8-second duty cycles, with most variance attributable to Tor circuit reconstruction overhead rather than transport switching.

2012-fifield-evading §4.2 evaluation

OONI pairs client-submitted test reports with data independently collected at the OONIB backend TestHelper, providing both connection endpoints' viewpoints in a single unified report. The backend is designed to be run by anyone and exposed both over HTTPS and as Tor Hidden Services to resist simplistic denial-of-service and reduce fingerprint-ability of the reporting infrastructure.

2012-filast-ooni §5 Architecture evaluation

OONI's threat model assumes an adversary capable of country-wide traffic manipulation who may actively fingerprint and identify measurement probes. Prior measurement tools (e.g., ONI's rTurtle) used easily fingerprinted centralized DNS and HTTPS traffic, which the authors flag as a pattern to avoid. The authors acknowledge that anti-fingerprinting measures will likely reduce measurement accuracy — a trade-off unresolved at publication.

2012-filast-ooni §3 Threat Model; §10 Limits and future work detection

With 512 PlanetLab nodes each advertising 50 KB/s as malicious Tor middle routers, the theoretical catch probability that at least one bridge circuit traverses a controlled node reaches P(512, 50, 30) ≈ 99% after only 30 circuits. In real-world validation, the 21st circuit created by a bridge client traversed one of the 512 controlled PlanetLab nodes, matching theory. The result generalizes: the 30-circuit exposure threshold applies to any adversary whose nodes' aggregated bandwidth reaches the equivalent of 512 × 50 KB/s = ~25.6 MB/s.

2012-ling-extensive §IV-B, §V-B, Fig. 7, Theorem 4 detection

Tor's bandwidth-weighted path selection creates a structural amplification: 60% of middle routers selected across 430 circuits had bandwidth above 1 MB/s, yet only 10% of all Tor routers exceed 1 MB/s. This skew means that an adversary advertising a single high-bandwidth middle node achieves selection probability far exceeding its proportional count in the network, making high-bandwidth Sybil nodes highly cost-effective for bridge discovery.

2012-ling-extensive §V-B, Fig. 8, Fig. 9 detection

The paper identifies three countermeasure classes against bridge discovery: (i) CAPTCHA on email/HTTPS distribution (limited by automated solving services); (ii) uniform random middle-node selection, which defeats bandwidth-Sybil attacks but degrades Tor throughput by routing through low-bandwidth nodes; (iii) DHT-based P2P architecture where no central server holds all bridge IPs, making systematic enumeration infeasible—though DHT systems introduce Sybil and eclipse-attack vulnerabilities of their own.

2012-ling-extensive §VI defense

Large-scale email and HTTPS enumeration of Tor bridges using 500+ PlanetLab nodes and 2,000 Yahoo accounts discovered 2,365 distinct bridges over approximately one month. The bridge https server rate-limits distribution to 3 bridges per 24-bit IP prefix per day, and the email server to 1 reply per account per day; these controls are circumvented by sourcing requests from hundreds of distinct prefixes. Bridge distribution follows a weighted coupon collector model proportional to bridge bandwidth, not uniform probability.

2012-ling-extensive §III-B, §III-C, §IV-A, §V-A detection

A single malicious Tor middle router advertising 10 MB/s bandwidth discovered 2,369 distinct bridges in 14 days. The catch probability is determined solely by the aggregated bandwidth M = k·b of malicious middle routers regardless of how that bandwidth is distributed across nodes: three routers at 10 MB/s each achieve strictly greater catch probability than 512 nodes at 50 KB/s each. This means a well-resourced single node is equivalent to or surpasses hundreds of low-bandwidth Sybil nodes.

2012-ling-extensive §IV-B, §V-B, Theorem 3 detection

SkypeMorph achieves a goodput of 33.9 ± 0.8 KB/s (naïve shaping) and 34 ± 1 KB/s (enhanced Traffic Morphing) versus 200 ± 100 KB/s for a normal Tor bridge, with overhead of ~28% compared to 12% for normal Tor. The two traffic-shaping methods perform statistically identically (KS p > 0.5), but the overhead grows during silent periods because the transport must transmit padding to maintain Skype's constant bitrate even when the Tor buffer is empty.

2012-moghaddam-skypemorph Table 1, §8 evaluation

A naive-Bayes website-fingerprinting classifier achieves AUC > 0.94 against vanilla Tor for 8 of 9 Alexa top-ten sites (e.g., Wikipedia 0.9991, YouTube 0.9947). Against StegoTorus-HTTP, AUC drops to ≤ 0.75 for 7 of 9 sites (YouTube 0.4125, Facebook 0.5413, Google 0.6928), which the authors argue is too low for practical perimeter-scale deployment where near-perfect precision is required to avoid error floods.

2012-weinberg-stegotorus §5.2, Table 2 evaluation

A blocked Tor bridge becomes reachable again after approximately 12 hours if Chinese scanners are unable to reach it continuously. In the authors' experiment, one bridge (port 23941) whitelisted to their Chinese VPS via iptables was unblocked within 12 hours despite remaining actively used, while an unrestricted bridge (port 27418) stayed blocked indefinitely.

2012-winter-great §4.2 detection

Of 2819 public Tor relays in the February 2012 consensus, only 47 (1.6%) were reachable via TCP from within China. After three days, only 1 of those 47 remained reachable. The GFC blocks relays by IP:port tuple rather than by IP to minimize collateral damage to co-hosted services.

2012-winter-great §4.1, §4.3 detection

Over 3295 active-probing scans observed across 17 days, 51% (1680) originated from a single IP address (202.108.181.70), while 98% of the remaining 1615 addresses were unique. All scanner IPs belong to three Chinese ASes: AS4837 (65.7%), AS4134 (30.5%), and AS17622 (3.8%). TTL analysis of 85 connections shows the scanner IPs are likely spoofed by the GFC—post-scan ping TTLs differed by +1 from during-scan TTLs.

2012-winter-great §4.5 detection

The GFC identifies Tor connections via a unique TLS ClientHello cipher list sent by the Tor client. Once DPI boxes detect this fingerprint on outbound traffic, active scanning is initiated within minutes: scanners connect to the suspected bridge, attempt to build a Tor circuit, and if successful the IP:port tuple is blocked. This two-stage pipeline (fingerprint → confirm → block) allows dynamic bridge blocking without pre-enumeration.

2012-winter-great §2, §4.1 detection

Tor DPI fingerprinting by the GFC is applied exclusively to egress traffic (from inside China to the outside world). Simulated Tor connections between domestic Chinese nodes and between external nodes connecting inward to a Chinese VPS attracted zero active scans across multiple experimental runs, indicating the detection infrastructure is positioned on the border for outbound flows only.

2012-winter-great §4.4 detection

Graduated censorship — limiting the suppression rate to remain within the typical weekly variance band — evades the weekly-interval detector entirely. The paper acknowledges that detecting slow-ramp blocking requires extending the observation window beyond seven days.

2011-danezis-anomaly-based §6 detection

Per-jurisdiction user counts are modeled as a Poisson process; the detector infers the 99.99th-percentile credible interval for the underlying rate λ from the observed count via a Gamma-Poisson approximation rather than a Gaussian assumption, correctly treating small-jurisdiction zero-user days as non-anomalous.

2011-danezis-anomaly-based §3–§4 defense

The detector constructs its 'typical ratio' baseline exclusively from the 50 largest jurisdictions, then discards outliers beyond four inter-quartile ranges of the median before fitting N(m,v). This ensures a jurisdiction undergoing active censorship cannot bias the global model and mask its own anomaly.

2011-danezis-anomaly-based §4 defense

A censor can defeat the anomaly detector without triggering an alert by replacing blocked user traffic with synthetic requests from adversary-controlled machines, keeping per-jurisdiction connection counts within the typical range. The paper explicitly identifies this as an unaddressed active-attack vector.

2011-danezis-anomaly-based §6 detection

The deployed system uses 7-day intervals and a baseline built from the 50 largest Tor jurisdictions; a jurisdiction's user-count ratio is flagged when it falls outside the 99.99th percentile of the fitted Normal distribution N(m,v), yielding an expected false-alarm rate of approximately 1 in 10,000 per jurisdiction-week.

2011-danezis-anomaly-based §5 evaluation

Cloud-based onion routing confronts censors with a collateral-damage dilemma: blocking a cloud provider's IP prefixes requires blocking all co-hosted services (Amazon EC2 hosted over 1 million instances sharing common IP prefixes in 2010), while allowing the traffic means circumvention succeeds. Rotating IP addresses—by retiring and spinning up new VM instances or via DHCP/gratuitous ARPs—reduces the window a blocked address remains in service, forcing censors into a perpetual cat-and-mouse game across all major cloud providers simultaneously.

2011-jones-hiding §2.3 defense

In controlled benchmarks using TorPerf, the best COR circuit achieved a median file download time 7.6× faster than Tor across 50 KB, 1 MB, and 5 MB files (100 repetitions each). COR was also several times faster than Tor for downloading full web pages across the top 10 Alexa domains, even when COR relays were serving 50 simultaneous connections.

2011-jones-hiding §4.1, §4.2, Figure 3 evaluation

COR does not solve the bootstrapping problem: a user's first connections to the COR bootstrapping network are vulnerable to the same IP-enumeration and blocking attacks as public Tor directory connections. To mitigate directory-partitioning attacks, directory retrieval is always performed through an existing COR circuit, and directories return only a random subset of available nodes rather than the full list—but this subset-delivery design is itself exploitable by a malicious directory that can fingerprint users via uniquely-assigned relay subsets.

2011-jones-hiding §2.3, §3.1, §5 defense

COR circuit construction enforces four properties to prevent single-entity de-anonymization in a limited-provider setting: (1) entry and exit ASPs must differ; (2) entry and exit CHPs must differ; (3) the same ASP's relays must not surround another ASP's relay without an intervening hop of a distinct ASP; and (4) at least two relays per traversed datacenter so an adversary with only perimeter visibility cannot trivially correlate ingress/egress.

2011-jones-hiding §2.4 defense

Running a COR network matching Tor's 2011 aggregate bandwidth (estimated at 150 MB/s end-user demand, ~376 TB/month) would cost approximately $61,200/month on Amazon EC2 at July 2011 pricing. A single EC2 node at 17¢/hour plus bandwidth charges can relay approximately 110 Mbps and support up to 100 concurrent users at ~1 Mbps each; m1.large and c1.medium instances handled 100+ concurrent connections while t1.micro struggled beyond 10.

2011-jones-hiding §2.5, §4.3, Figure 3(right) evaluation

By integrating onion routing at the ISP/AS boundary (exactly one onion router per AS hop), the specific relay used is neither specified in the protocol nor addressable by end hosts, eliminating the enumerable relay list that makes overlay Tor blockable. The end host only knows ISP public keys, not individual relay addresses.

2011-liu-tor §5 defense

Encrypting traffic at the application layer still discloses communicating parties to every ISP along the path; overlay anonymization is subject to blacklisting of exit nodes and traffic analysis. The paper argues that effective privacy requires building anonymity into the network routing layer itself, with the necessary tradeoff being hardware cost and routing inefficiency for privacy-requiring circuits.

2011-liu-tor §1, §8 defense

Tor-like anonymizing overlays are easily censored because they rely on centralized, publicly visible relay lists; governments can blacklist Tor nodes or monitor all Tor exit traffic so that traffic analysis can reveal the source. Traffic to or from Tor 'essentially advertises itself as probably worth tracking.'

2011-liu-tor §1–§2 detection

The design guarantees that as long as an end host can reach any non-censoring ISP, it can trampoline to any service; the anonymity properties make it difficult for ISPs to selectively block flows without cutting off the end host from the outside world entirely. Wikileaks-like services require only one willing authority for name resolution, not universal cooperation.

2011-liu-tor §7.1 defense

When RIAA filed suit against more than 30,000 individual filesharers, users migrated toward anonymous channels, small-world networks of vetted peers, ephemeral pointers, and user-generated IP blacklists for spoofed-peer detection. The University of Washington demonstrated IP-to-person attribution is unreliable — a networked laser printer received a DMCA takedown notice.

2011-seltzer-infrastructures §1.3 defense

The U.S. 'five strikes' program had major ISPs reduce bandwidth of accused subscribers; challenging required a $35 fee with only one permitted defense category ('unauthorized use of account'). Users responded by routing traffic through VPNs and anonymizing networks such as I2P to bypass ISP-level monitoring entirely.

2011-seltzer-infrastructures §1.4.2 policy

Tor bridges that always accept incoming connections enable a three-phase 'bridge aliveness attack': an adversary collects bridge descriptors at scale, correlates bridge uptime timestamps with pseudonymous post timestamps to narrow the candidate set (winnowing), then confirms identity via circuit-clogging and timing attacks. Because bridge descriptors remain valid indefinitely and the BridgeDB rate-limits only to one descriptor set per /24 prefix per week, an adversary with botnet or open-proxy access can hoard enough bridges for the winnowing phase to succeed.

2011-smits-bridgespa §1, §1.1 detection

A passive observer of BridgeSPA traffic sees only a TCP connection timeout on failed authorization or a successful TLS connection on success—exactly what they would observe with an unmodified Tor bridge. The ConnectionTag is indistinguishable from the normally-random ISN and timestamp fields in Linux 2.6, so no new observable artifact is introduced. However, BridgeSPA does not address the separate problem that Tor traffic itself remains fingerprint-distinguishable from HTTPS; this is an orthogonal concern.

2011-smits-bridgespa §4, §6.2.1 defense

BridgeSPA encodes a 32-bit SHA256-HMAC ConnectionTag derived from a time-limited MACKey into the TCP SYN packet's ISN (lower 3 bytes) and TCP timestamp (lower 1 byte) fields—values that are uniformly random in Linux 2.6 and therefore carry the tag innocuously. Bridges silently drop unauthorized SYN packets without returning any response, preventing aliveness queries. MACKeys rotate every 1–7 days (bridge-configured), so hoarded descriptors become stale within the epoch.

2011-smits-bridgespa §4, §5 defense

An active man-in-the-middle adversary can hijack a live BridgeSPA TCP SYN by intercepting the ConnectionTag-bearing packet and racing to complete the bridge connection before the client's timestamp rounds to a new minute. Mitigating this requires the client to re-send the full (non-truncated) ConnectionTag after TLS is established, causing the bridge to act as a cover service (e.g., IMAP over TLS) until validated—but this mitigation is undermined by the fact that Tor bridge TLS certificates are currently distinguishable from other service certificates.

2011-smits-bridgespa §6.2.2, §7.5 detection

At the time of writing, the Tor network had no publicly announced exit nodes located on the Chinese mainland, making direct Tor-based measurement of GFW filtering unavailable. The paper generalizes this: heavily filtered countries show systematically low availability of relay services, precisely where measurement need is highest.

2011-wright-fine-grained §4.2 evaluation

Adding dummy traffic to any anonymity mechanism yields the corresponding kind of unobservability: 'A mechanism to achieve some kind of anonymity appropriately combined with dummy traffic yields the corresponding kind of unobservability.' DC-nets achieve sender anonymity and MIX-nets achieve relationship anonymity; with dummy traffic both achieve the corresponding sender and relationship unobservability respectively.

2010-pfitzmann-terminology §8 defense

Pseudonymity uses persistent identifiers other than real names, enabling accountability while providing partial unlinkability; however, use of the same pseudonym across different contexts enables linkability: the attacker can link all data related to a pseudonym. Unlinkability of two messages requires that the attacker cannot sufficiently distinguish whether they share a sender or recipient; for a scenario with n senders, this holds iff the probability of common authorship is sufficiently close to 1/n.

2010-pfitzmann-terminology §4, §9, §11 defense

The paper establishes a strict property hierarchy: unobservability ⇒ anonymity, and sender/recipient anonymity ⇒ relationship anonymity. Unobservability is strictly stronger than anonymity because it additionally requires undetectability against all uninvolved subjects — the IOI's very existence must be hidden — while anonymity only hides the subject's relationship to the IOI.

2010-pfitzmann-terminology §7 defense

A circuit-clogging attack against bridge operators—using median-normalized latency correlations—achieved an AUC of 0.884 and an equal error rate of 0.2 when distinguishing the victim bridge from innocent bridges in PlanetLab experiments with 180 victim and 180 disjoint runs. With 10 repeated clogging experiments and a majority-vote threshold, the false positive (and false negative) rate drops below 0.033, confirming a bridge operator's identity with high confidence given a candidate set of ≤4.4 bridges from the winnowing stage.

2009-mclachlan-risks §3.4 detection

An 'unfair queuing' mechanism that partitions CPU time between bridge-operator circuits and bridge-client circuits using a time-allocation parameter τ=0.9 reduced the circuit-clogging AUC from 0.884 to 0.520 (median-normalized) and 0.412 (mean-normalized)—indistinguishable from random guessing—in 20 PlanetLab experiments. The mechanism eliminates latency interference between the two circuit types without requiring the bridge to ever refuse connections, but introduces up to 1−τ performance loss for client traffic.

2009-mclachlan-risks §4.3 defense

Cross-referencing the online/offline status of 87 monitored bridges against 186,935 Wikipedia users' edit sessions showed that 95.7% of users with 50 or more sessions matched zero bridges after winnowing. For users with 180 or more sessions (a surrogate for long-term pseudonymous activity), only 89 false positives remained among 2,329 users—a false positive rate of 0.000439—meaning that even if 10,000 Tor clients volunteer to bridge, on average only 4.4 bridges remain after the winnowing stage.

2009-mclachlan-risks §3.3 detection

The hybrid two-stage design's architectural vulnerability is that circumventing either stage independently defeats the system: end-users can tunnel via Tor or JAP to bypass both stages entirely, while content providers can serve different content to IWF crawlers versus real users, exploiting the fact that only 33% of IWF hotline reports were substantiated as potentially illegal. The system's precision is entirely contingent on content-provider cooperation, which cannot be assumed.

2006-clayton-failures §4, §6 defense

The paper evaluates all major circumvention techniques available in 2003 and concludes that only application-layer proxies (HTTP, SOCKS, JAP, peek-a-booty) and IP tunneling can defeat all three blocking layers (IP filtering, DNS tampering, filtering proxies) simultaneously. Encryption alone cannot circumvent IP or DNS blocking; HTTPS hides URL paths but not the destination host; DNS-over-HTTPS/DNSSEC can detect but not defeat DNS tampering without a third-party resolver.

2003-dornseif-government §2.4 defense

Active-server document anonymity is achieved by routing decryption through a randomly chosen ephemeral 'decrypter' node: the storer holds only ciphertext {h}k while key k is delivered separately to the decrypter via onion routing. Neither the storer nor any other single node can reconstruct the plaintext share, so a storer cannot identify the document it is hosting even by attempting to retrieve it.

2002-serjantov-anonymizing §2–§3 defense

The paper proposes a forwarder/storer role split in which forwarders hold only an anonymous return-address pointer to the storer, and deliberately forget the storer's identity upon receiving a publication acknowledgment. Because forwarders neither hold content nor retain storer addresses post-publication, coercing a forwarder after publication yields no actionable information about where shares are held.

2002-serjantov-anonymizing §2–§3 defense

Publius provides source anonymity once content is published but offers no connection-based anonymity at upload time. A network-layer eavesdropper between the publisher and the servers, or a server's connection log, can reveal the publisher's IP address. The paper explicitly states that Publius must be combined with a mix-network or crowd-anonymity tool (e.g., Crowds, Onion Routing) to protect publisher identity during the upload phase.

2000-waldman-publius §5.4 evaluation