TLS ClientHello fingerprinting

A June 2024 internal MESA Team survey ('针对审查系统的科学研究及探测技术调研报告') catalogs the academic censorship-measurement toolkit (OONI, Augur, Satellite, Quack/Hyperquack, GFWatch, GFWeb, middlebox weaponization studies, traceroute-based middlebox localization, device fingerprinting) and separately reviews circumvention-tool countermeasures, explicitly naming Lantern alongside Psiphon, Tor Meek and Signal as tools using uTLS-style TLS ClientHello mimicry and domain-fronting.

The same T1/NTC node config (ntcconf/t1conf/main.conf, K18/Astana/Kazakhtelecom) exposes concrete DPI enforcement toggles: a SYNACK_OR_RST switch and SEND_INJECT_PKT flag governing active TCP-response/packet-injection behavior; ASN- and IP-based blocklists (ASN_MAAT, IPD_DYN_MAAT, IPD_STATIC_MAAT) refreshed from MAAT/Redis on a 1-second effect interval; a dynamic blacklist with a 180s timeout; explicit protocol-blocking switches for BitTorrent (DHT/uTP) and eMule (Kad); and TLS metadata harvesting including certificate SAN fields (NTC_SSL_COLLECT, collect_san_sw=1) streamed to a Kafka topic.

Exported AppSketch/MAAT signature-object JSON shows named per-VPN detection rules combining multiple independent vectors: CyberGhost's WireGuard variant is matched by a hex UDP-payload pattern (04000000*), destination port 1337, and ip.proto=17 together (signature cyberghostvpn_wireguard); Windscribe's OpenVPN control channel is matched by two offset-anchored hex byte patterns plus an exact 86-byte payload length (windscribevpn_openvpn_payload); and ExpressVPN is matched independently via a literal JA3 hash (ssl.analysis.ja3), FQDN strings (expressapisv2.com, www.expressvpn.works), and a large enumerated IP list (700+ individual ip.dst entries observed before the 200KB text-extraction cutoff).

The same 224-column TSG per-session log schema carries a full TLS- interception field set — proxy_pinning_status, proxy_intercept_status, proxy_passthrough_reason, proxy_cert_verify, proxy_intercept_error, sc_rsp_raw and sc_rsp_decrypted (raw vs. decrypted server response content) — plus ssl_esni_flag and ssl_ech_flag (explicit ECH/ESNI-usage flags), ssl_ja3_hash/ssl_ja3s_hash, and ssh_hassh (SSH client fingerprinting), confirming certstore-style MITM interception, ECH/ESNI detection, and TLS/SSH fingerprinting are all first-class fields logged on every session, not experimental add-ons.

MAAT/AppSketch signatures "hotspotvpn_ja3" and "ultrasurfvpn_update_behavior" identify Hotspot VPN and Ultrasurf specifically via static JA3 TLS ClientHello hashes (e.g. f49621211538d12435b8498f195d0c31 for Hotspot VPN; 706ea0b1920182287146b195ad4279a6 and 1f0d8bb4fab11dc335dec2e69da0c36e for Ultrasurf), in Ultrasurf's case combined with an AND-condition on a set of legitimate-looking cover domains (skype.com, msn.com, microsoft update domains) — showing the DPI engine defeats Ultrasurf's domain-mimicry cover by fingerprinting the underlying TLS stack instead of trusting the SNI/host.

A 2020 MESA Lab monthly report describes building a sapp plugin that extracts packet-sequence features specifically "for DoH (DNS-over-HTTPS) service discovery," alongside a broader CSTNET DoH measurement-report effort and configuring a DNS-to-DoH gateway -- confirming sapp is used to fingerprint DoH traffic via statistical sequence features rather than plaintext DNS content.

MESA Lab researcher notes on an encrypted-video-identification project describe adding SSL-layer information output to more precisely trace a flow's true source/identity, and considering reinforcement learning so the identification model adapts as network conditions change, working within/around sapp's plugin limitations.

A MESA Lab task tracker records the assignment "TSG: determine how many clients are behind an IP address," and a companion technical design document details the method: identify distinct TLS clients sharing one public IP using a <JA3 fingerprint, server domain, server IP> 3-tuple (JA3 alone collides across different apps), then use TLS Session Ticket reuse/lifetime sequences per identified client to detect multiple concurrent devices (i.e. NAT) behind that IP.

Internal MESA Lab reading notes dissect the USENIX 2024 paper on fingerprinting obfuscated proxies via encapsulated TLS handshakes, highlighting its protocol-agnostic packet-size-3-gram-plus-Mahalanobis-distance-over-bursts classifier, which the paper's own mid-size-ISP deployment reliably fingerprinted across shadowsocks, vmess, trojan, and vless-family configurations at false-positive rates the notes explicitly say the annotator estimates the GFW would find operationally acceptable (<0.6%). The notes flag the technique's main gaps as: no public source code, sharply reduced true-positive rate under connection multiplexing (10-30% vs. 60-80% unmultiplexed), and no evaluation against UDP/QUIC.

An internal app-fingerprint engineering document (identical content filed under two separate artifact ids) catalogs precise per-action (login/create/upload/ download/share/delete) DNS, SNI, and certificate-CN signatures for identifying use of specific cloud-collaboration apps from traffic alone, covering foreign platforms Zoom and Office365 alongside domestic apps (Huawei WeLink, Feishu, DingTalk, WeCom, Shimo Docs, CSTCloud drive), using tools like myssl.com for active certificate/IP verification of each signature.

A raw TSG session-log export shows the full production log schema: per-flow fields include subscriber_id, imei, imsi, phone_number and apn alongside ssl_ja3_hash/ja3s_hash, ssl_esni_flag, ssl_ech_flag, quic_sni, dtls_ja3_fingerprint, and proxy_pinning_status/proxy_intercept_status/ proxy_cert_verify fields for the MITM proxy path. This confirms TSG's session logging natively joins network-flow identity to subscriber identity and captures TLS fingerprint/ESNI-ECH state and MITM-interception outcome in the same record.

TSG's TLS ClientHello parser (MESA_Platform/ssl GitLab component) explicitly parses the ec_point_format extension from ClientHello, confirming deep TLS extension-level fingerprinting beyond simple SNI extraction.

An internal risk memo, "Signal审查规避模式阻断风险说明" (Signal circumvention-mode blocking risk explanation), documents that after the operator pushed a "Deny Signal APP" policy, users with Signal's censorship-circumvention (domain fronting via Google infrastructure) enabled could still send messages/files; packet capture showed the fronted traffic used Google SNIs (clients3/4.google.com, inbox.google.com, android.clients.google.com, www.google.com), and the team built a targeted FQDN+JA3+payload signature specifically to block this fronted traffic while flagging a residual risk of false-positive blocking of legitimate Google services.

The Galaxy query-gateway platform's (galaxy/platform/galaxy-qgw-service) statistics/policy-metrics schema had a JA4 fingerprint field added (ticket TSG-23812), confirming TSG's analytics/policy layer captures JA4 -- a modern client TLS/QUIC fingerprint -- as a first-class, queryable field platform-wide, beyond legacy JA3.

Raw TSG session_record export logs dated Nov 2024, tagged device_group 'YGN-MYTEL' (Yangon, Mytel), directly corroborate the leak's M22=Myanmar/Mytel/Yangon site-codename mapping with live production data: individual SSL sessions are classified by app as 'Hotspot Shield VPN' (destination get.adobe.com, likely a fronting/CDN endpoint) and allowed under a named whitelist rule 'whitelist_102024', geolocated client-side as Myanmar.Yangon.Yangon.

An internal research writeup ("研究点二：基于GRU神经网络的共享接入IP检测技术") builds a GRU/CNN model over sequences of TLS JA3/SNI/session-ticket and HTTP cookie fingerprints, trained on 6 days / 155GB of mirrored traffic captured from an internal gateway named "华严网关" (Huayan Gateway), to determine whether a single source IP represents one device or several devices sharing that IP (e.g. behind a NAT/proxy), reporting precision 0.844 / recall 0.874 / F1 0.859.

Two IIE graduate-research repos (cuiyiming/gradproj, a 2019-2020 master's thesis project citing NDSS'17 TLS-interception-measurement and TLS-client-identification papers; daxiaoxu/xmr_bsexpr2, a 2022 project with GRU-based sequence classifiers over TCP/DNS flow JSON and deleted docs on TLS1.3 and Tencent's proprietary MMTLS protocol) document the ML feature-engineering methodology (TLS certificate length, handshake message sequences, JA3-style statistics, Markov-chain packet-size/timing models, GRU sequence models) that plausibly underlies production classifiers (e.g. stellar's later JA4/JA4S support, MESA_sts's randomness checks).

The tango/kni (Kernel Network Interface) repo shows TSG's SNI/TLS-ClientHello-based interception logic (kni_entry.cpp) receiving dedicated 2024 fixes for handling TLS ClientHello fragmented across multiple packets ("ssl chello frag") and a "chello first packet hit intercept policy" bug, on top of a long-standing (2020) "dynamic bypass" feature that replaced raw-ClientHello matching with JA3 TLS fingerprints, plus counters for SSL certificate-pinning detection.

A censorship-circumvention survey presentation (English-language, apparently used for internal briefing/training on evasion techniques within this ecosystem) explicitly lists Lantern by name, alongside Snowflake, as a circumvention tool discoverable by censors via TLS ClientHello fingerprinting, citing NDSS 2019 "The Use of TLS in Censorship Circumvention," and separately catalogs active- probing techniques (port-scan plus protocol-specific probes) used to discover OpenVPN, probe-resistant proxies (obfs4/shadowsocks/OSSH/MTProto), and traditional VPN servers.

The Transformer_master.h header (part of the Maat/sapp DPI framework, dated 2023-05-04) defines a region-keyed DNS response-forgery subsystem (MSG_OPT_DNS_CHEAT_TYPE/RCODE/STRATEGY/RECORD/TTL, DNS_FAKE_INFO/DNS_FAKE_IP tables, TF_get_dns_response_strategy_id(user_region)) used specifically by the block ("FD") action path, plus a dedicated function to classify whether a detected L7 protocol is a VPN (TF_is_L7_vpn_prot), and native extraction of TLS JA3/JA3S client and server fingerprints.

A June 2024 MESA Lab internal survey ("针对审查系统的科学研究及探测技术调研报告") explicitly states its purpose is to catalog academic/public censorship-measurement and circumvention research (OONI, Augur, Satellite, GFWatch, Citizen Lab, CensorBib, FOCI/IMC/NDSS/CCS/USENIX Security papers) in order to find and patch GFW/censorship-system vulnerabilities before outside researchers exploit them. It systematically covers circumvention protocols/tools (Shadowsocks, VMess, Trojan, decoy routing, Parrot-style mimicry, CovertCast, Slitheen++, ESNI/ECH, uTLS) and notes GFW blocked ESNI (not ECH) since July 2020, plus historical TLS-fingerprint blocking of meek by a Cyberoam firewall (2016).

A raw production SSL/TLS session log from sapp's monitoring pipeline (timestamps dated 2021-08-23, client IPs in domestic Chinese carrier ranges) shows the per-connection logging schema includes dedicated fields for a captured certificate chain (INDIVIDUAL_CERT_FILE, MIDDLE_CERT_FILE, ROOT_CERT_FILE, CHAIN_CERT_FILE) and for any injected packet (INJECTED_PKT_FILE), alongside SNI and TLS version -- confirming sapp's live SNI-logging and MITM-cert-capture instrumentation was operating against real user traffic to services including huobi.com, steamcommunity.com, and dropbox.com.

An IIE researcher's repo hswfp ('server-side fingerprinting experiment') packages written experiment steps and 'server-side fingerprinting experiment-related code' as a self-contained study, indicating active internal R&D into fingerprinting server/service implementations beyond the productized JA3/JA4 work already shipped in the SSL plugin.

The same SSL plugin added explicit detection of the Encrypted Client Hello extension (type 0xFE0D) with test cases in June 2023, and separately implements JA3/JA3S (2020) and JA4/JA4S (2024) TLS client/server fingerprint calculation as first-class features -- so TSG-class DPI can flag 'ECH is in use' even without decrypting it, and independently fingerprints TLS stacks via JA3/JA4.

MESA_Platform's "quic" module (built and packaged as stellar-on-sapp/sapp RPMs) implements SNI/User-Agent extraction from both cleartext GQUIC (versions 23-59) and encrypted IETF QUIC RFC9000 ClientHello payloads, and supports a QUIC SNI whitelist -- i.e. the DPI pipeline decrypts/parses encrypted QUIC handshakes to recover the destination domain rather than being blocked by QUIC's encryption.

The tango/tfe ("TFE") engine ships dual root-CA trust stores (tango-ca-trust-ca.pem / tango-ca-untrust-ca.pem, mesalab-ca.pem) for TLS interception plus a JA3-fingerprint-based certificate-pinning detector with a configurable "Dynamic Bypass"; a 2021 bug ticket documents testing this against Firefox without a root cert installed visiting twitter.com, confirming live-target validation of the MITM/pinning-detection pipeline described under the certstore product.

Standard workflow (recurring across AdGuard, JumpJump, Hula, BigMama, and Ace VPN tickets) for adding a new VPN app to TSG's blocklist: capture traffic on iOS/Android/Windows in a dedicated test environment, extract SNI first, fall back to enumerating server IPs when SNI proves ineffective or the app rotates it, then re-verify weekly/periodically with automated 'dial testing' (拨测) that checks whether newly discovered IPs still connect ('未穿透' = not yet penetrated/blocked).

A MESA research-log entry details a NAT/shared-connection identification methodology combining TCP/IP fingerprinting (IP-ID, TTL, DF, window size, MSS, TCP-option ordering -- p0f-style), HTTP User-Agent diversity, TLS/SSL JA3 fingerprint diversity, and traffic statistical features (concurrent-TCP-connection count, idle-time jitter, upstream/downstream ratio stability, DNS query frequency) per endpoint over rolling time windows, with detection methods spanning direct UA inspection, threshold statistics, ML classifiers (random forest/SVM), and per-window entropy jumps across the fingerprint features.

TSG's QUIC-parsing layer decrypts QUIC ClientHello payloads and parses a user-agent-like parameter from the decrypted content, deployed at the Ethiopia (E21) DIR-IGW site. A missing bounds check on this field caused a watchdog-timeout crash, confirmed via a MESA_Platform/quic GitLab commit (git.mesalab.cn/MESA_Platform/quic).

A K18 customer specifically requested TSG support for TCP-segmented TLS ClientHello messages, confirming this was previously an inspection gap; Geedge shipped a fix ("Hotfix-Inspection-TLS-PQC-Extension") in 2024, and the same hotfix name indicates TSG's TLS inspection also tracks post-quantum-crypto ClientHello extensions.

Multi-engineer forensic investigation (pcap, JA3 hashing, TSG session-log correlation) at an Ethiopia site into tv.cctv.com being intermittently unreachable concluded the blocking was NOT done by TSG itself but by a separate, more-client-proximate network censorship system. That system dropped ClientHello-stage TLS1.2 sessions matching cctv.com's SNI, had degraded filtering ability once the client renegotiated to TLS1.3, and at one point misclassified cctv.com traffic as TikTok.

A debugging ticket for sites "Old Airport-PE" and "Bole-IGW" (naming consistent with Addis Ababa, Ethiopia) confirms the customer verified Psiphon3, YouTube, Facebook, and Telegram blocking as working well during independent testing, and reveals the operational session-correlation method used when packet captures and system logs must be cross-referenced across NAT: JA3_HASH + server IP + client public IP + SSL SNI.

Geedge extracted and analyzed Betternet VPN's underlying protocols (WireGuard, a proprietary "Hydra" protocol using spoofed-domain TLS 1.2, and IKEv2), and found the Hydra protocol could be "completely blocked" via JA3 TLS fingerprinting; IKEv2 was blocked via UDP port 500/4500 policy, and thousands of VPN server IPs were extracted as a supplementary IP blocklist.

NTC_SSL_COLLECT passively parses every TLS handshake in monitored traffic and streams the SNI, hex-encoded client and server cipher-suite lists (a JA3/JA3S-equivalent fingerprint), and full leaf-certificate fields (serial number, algorithm, issuer/subject CN, org, and country, validity window, and optionally the complete SAN list) to Kafka topic 'ntc_ssl_collect_log' for every TLS session, tagged with the same PROTO_ID/APP_ID/OS_ID/BS_ID/WEB_ID/BEHAV_ID classification label used across the T1 plugin family.

A MESA Lab repo (daxiaoxu/xmr_bsexpr3) contains standalone detection/fingerprinting scripts specifically targeting QQ (Mobile and Windows clients) and WeChat, alongside a script named PSK_LifeTime.py, suggesting TLS session-resumption / PSK-ticket lifetime is being explored as a traffic-fingerprinting feature for identifying or tracking these apps' encrypted sessions.

A MESA Lab traffic-fingerprinting dossier builds DPI detection material for third-party remote-access/tunneling software (Sunlogin/花生壳, TeamViewer, and a self-hosted OpenVPN test deployment) by cataloguing each tool's domains/IPs, TLS SNI values, HTTP User-Agent and Host headers, and UDP ports, with traffic captured and labeled per discrete user action (login, add device, start remote session, quit, logout) — exactly the raw material an app-identification signature system would be built from.

A 'feature_extract_plugin' repo built for MESA Lab's sapp platform (using the same '*_entry.cpp' plugin-registration pattern seen elsewhere in sapp's plugin family) implements dedicated SSL/TLS parsing and statistics entry points (sslparse_entry.cpp, sslstat_entry.cpp) alongside a generic stream-statistics entry (stmstat_entry.cpp), indicating sapp has an extensible plugin architecture specifically for extracting TLS-connection features -- consistent with TLS fingerprinting / classifier feature extraction -- beyond the base DPI engine.

An internal MESA Lab research survey dedicated to Snowflake reviews five external papers on fingerprinting/blocking Snowflake and WebRTC-based transports, including a 2023 CSCWD paper that fingerprints Snowflake's broker-contact requests (disguised as HTTPS) via packet size, direction, timing, and network speed to distinguish them from ordinary web requests, and a 2020 paper claiming 100% identification of Snowflake DTLS handshakes against Facebook Messenger/Hangouts/ Discord WebRTC traffic in a closed-world test — annotated throughout with the MESA reviewer's own critical commentary on each method's weaknesses (e.g. sensitivity to user geography/network, only tested at host/LAN vantage points rather than backbone).

Internal spec for "spacemap-service", a passive-DPI-fed reconnaissance database and web UI that catalogs every observed website's IP/domain, geolocation, ASN, DNS server, CDN vendor, TLS certificate fields (issuer/CN/SNI), OS and web-framework fingerprint (via Wappalyzer), and WHOIS/ICP registration data, exposing reverse-lookup search by any single attribute (e.g. all sites behind a given cert issuer, ASN, or CDN vendor).

A leaked SQL migration renames/normalizes a session-attribute-dictionary entry to `ssl.analysis.sni_absent` (backed by virtual_table `ATTR_SSL_NO_SNI`), confirming the detection pipeline maintains an explicit, first-class boolean signal for "TLS ClientHello with no SNI present" — a classic tell for ECH, domain fronting, and custom TLS stacks that omit SNI to evade SNI-based blocking.

MESA Lab's "stellar" SSL/TLS decoder (ssl_decoder, packaged in test fixtures as "stellar-on-sapp") ships a dedicated regression test case captured against an Ethiopia (E21) target host, alongside TLS 1.3 ESNI, encrypted ClientHello (ECH), and TachyonVPN-client test pcaps, showing the decoder is validated both against E21-deployment-style traffic and emerging TLS privacy extensions.

A November 2024 commit to the stellar DPI engine's TLS decoder adds JA4 and JA4S TLS-handshake fingerprint calculation ("tls decoder support calculating JA4/JA4S fingerprint"), showing Geedge's core inspection engine has adopted the modern JA4/JA4S fingerprinting scheme (successor to JA3/JA3S) for client/server TLS fingerprinting as of late 2024.

A previously-undocumented sapp plugin named 'Stellar' (repo stellar/quic_decoder, packaged as stellar-on-sapp, depending on sapp-devel and OpenSSL) implements 'quic_deprotection' — QUIC packet-protection removal — as a component distinct from the older MESA_Platform/gquic plugin, sharing the same test corpus of GQUIC/IETF-QUIC pcaps (RFC9000, chlo-fragment, no-SNI cases). This is a candidate new product entry for the taxonomy: a second, apparently newer QUIC-inspection engine alongside gquic.

A DPI session-plugin ("stellar" framework) SSL/TLS decoder fully parses live ClientHello/ServerHello handshakes, computing JA3/JA3S fingerprints (MD5 over version+cipher-suites+extensions with GREASE values stripped per the standard JA3 spec), extracting plaintext SNI, explicitly flagging whether a ClientHello carries ESNI (extension type 0xFFCE) or ECH (extension type 0xFE0D), and decoding full X.509 certificate chains (issuer/subject RDN sequences, SAN, validity window, serial, pubkey, signature algorithm) in real time. It also recognizes TLCP 1.0 (China's national "国密"/GmSSL TLS variant) as a distinct protocol version alongside SSLv2 through TLS1.3.

chongming/traffic_replay is a Flask/tcpreplay-based QA harness that replays curated pcaps against a TSG instance to exercise its decoders, including SSL fixtures specifically named ssl_ech_false.pcap, ssl_no_sni.pcap and ssl_tls13.pcap alongside esni.pcap, openvpn.pcap, and major-platform captures (facebook/google/twitter/ youtube). A companion update_tsg_ca.py/tsg_ca.txt installs a TSG-issued CA cert on the test client, confirming the QA environment exercises TLS interception as part of routine detection testing.

Two independently-leaked TSG session-log CSV exports share an identical 224-field schema whose proxy_* columns (proxy_action, proxy_pinning_status, proxy_intercept_status, proxy_cert_verify, proxy_intercept_error) log the outcome of TLS interception per session, alongside ssl_ja3_hash/ssl_ja3s_hash, ssl_esni_flag, ssl_ech_flag, and quic_sni fields — confirming MITM/cert-pinning-bypass instrumentation and TLS/QUIC fingerprinting are built into TSG's standard traffic-logging pipeline, not a bolt-on feature.

A leaked TSG session-log CSV schema (device XXG-TSG-BJ) includes built-in per-session fields for TLS interception status (proxy_intercept_status, proxy_cert_verify, proxy_pinning_status, proxy_passthrough_reason) alongside JA3/JA3S TLS fingerprints and explicit ssl_esni_flag/ssl_ech_flag columns, confirming TSG logs whether a session was MITM'd and whether it used ECH/ESNI.

The full TSG session-log CSV schema header discloses the complete set of per-session fields the platform records, including TLS JA3/JA3S fingerprint hashes, explicit ssl_esni_flag/ssl_ech_flag columns, QUIC/DTLS SNI and JA3 fields, and direct subscriber-identity fields (subscriber_id, imei, imsi, apn, phone_number) captured on every logged session alongside ordinary network 5-tuple data.

An internal field-mapping schema for TLS/SSL session logging (cross-referenced against multiple internal system versions, one explicitly labeled "tsg") shows JA3 and JA3S TLS ClientHello/ServerHello fingerprint hashes are captured as standard logged fields alongside SNI, negotiated cipher suites, and the full server and mutual-TLS client certificate chains (issuer, subject, serial, validity window, raw public key).

TSG's self-diagnostic suite (tsg/tsg-diagnose) imports the full badssl.com test-certificate corpus (superfish, eDellRoot, mitm-software, captive-portal, revoked, weak-key variants, etc.) to continuously validate its own TLS-interception logic in production, adds test cases for TLS ClientHello fragmentation handling by the intercept engine, and documents that firewall DENY blocking was changed from sending only a FIN to sending FIN-then-RST to close connections.

tsg_master computes and logs JA3 TLS ClientHello fingerprints (feature present since at least 2020, field renamed common_ja3_fingerprint -> common_ja3_hash; toggle GENERATE_JA3_FINGERPRINT added 2023) and, since 2023, also outputs OS-fingerprint log fields (common_client_os_name, common_server_os_name), plus improved SSL/TLS detection when the ClientHello is split across multiple TCP segments.

A confidential ("内部资料 注意保密") 2020 MESA architecture-group report catalogs infrastructure recon (ICP registration, whois, DNS, FOFA scans, TLS-cert lookups) and packet-capture/SNI fingerprinting for seven video-conferencing services -- DingTalk, Zoom, Feishu/Lark, 科技云会, Teambition, Slack, and WebEx -- including specific SNI values observed for Zoom's China CDN deployment (e.g. cn01www3.zoom.com.cn, zoomawscn5281151165zc.zoom.com.cn).

A confidentiality-marked ("内部资料 注意保密", 架构组/Architecture Group) MESA Lab research report systematically profiles 11 video-conferencing/collaboration platforms (DingTalk, Zoom, Feishu/Lark, WeCom, Tencent Meeting, WebEx, Huawei WeLink, Slack, Teambition, 科技云会, 小鱼易连) with SNI lists, TLS certificate fingerprints, DNS/whois infrastructure, and an explicit "passive traffic analysis feasibility" (被动流量分析可行性) verdict per app — direct precursor engineering work for per-app DPI detection signatures.

Live TSG session logs from an internal test bed (device 21426003 / XXG-TSG-BJ) show active Deny actions firing against named commercial VPN apps -- Super Unlimited VPN, BetternetVPN, TouchVPN, LetsVPN -- via dated signature patches (e.g. BetternetVPN_Patch01_20240812 through Patch06_20240920, SuperUnlimitedVPN_Patch05 through Patch10 within six weeks), and LetsVPN's block fires on a JA3 TLS fingerprint (LetsVPN_by_ja3) even when the SNI/Host is spoofed to www.bytedance.com.

Internal fingerprinting research report builds SNI/TLS/certificate profiles and a repeatable packet-capture methodology (separately labeled login/logout/ping/remote-desktop/ssh activity captures, split by TCP vs. UDP) for Sunlogin (向日葵/Oray), TeamViewer, and OpenVPN — explicit precursor engineering work for building per-app/per-protocol DPI detection signatures, including for OpenVPN over both its TCP and UDP transport modes.

Beyond the Lantern-specific rules, this batch contains ~40 individually-dated JSON signature files from the same MAAT/AppSketch rule format, each targeting one named VPN/circumvention client by FQDN, destination IP, TLS JA3 hash, or certificate issuer: Signal, Orbot, Cloudflare WARP (including a MASQUE-specific variant combining an app_id classifier with an FQDN condition), ExpressVPN (matched by a literal JA3 hash plus a separate FQDN rule), ProtonVPN/ProtonMail, TurboVPN, QuarkVPN, LetsVPN, BetternetVPN, Browsec, GeckoVPN, HulaVPN, JumpJumpVPN, MouseVPN, NotVPN, SecureVPN, StarkVPNReloaded, SuperUnlimitedVPN, TowerVPN, TrustzoneVPN, USAVPN, VPNIndia/VPNLite/VPNPro/VPNTurkey, FlyVPN, ACE VPN, F1 Rockets VPN, and the Chinese gaming accelerator biubiu加速器. Dates cluster tightly (e.g. many on 2024-08-09, others 2024-09/10/11), indicating an active, continuously-updated per-app signature production pipeline rather than a static list.

A large set of individually-dated per-product signature files shows at least 31 distinct named commercial VPN apps under active, ongoing detection engineering (SecureVPN, LetsVPN, BetternetVPN, SuperUnlimitedVPN, TunnelBear, AtlasVPN, TurboVPN, UrbanVPN, GeckoVPN, VPNUnlimited, BitdefenderVPN, QuarkVPN, FlyVPN, HulaVPN, SurfVPN, BulletVPN, BeePassVPN, SnapVPN, WhitehatVPN, LunaVPN, SuperNetVPN, GitiVPN, CleanerAntivirusVPN, V2VPN/V2Netvpn, 7VPN, and country-branded VPNIndia/VPNCanada/VPNIndonesia/VPNKorea apps), each with multiple detection vectors per app (destination-IP ranges, FQDN, DNS QNAME, TLS cert issuer, JA3, WireGuard-protocol matching, ISAKMP/IKE matching, raw TCP payload); one file shows the resulting policy rule object wired directly to an explicit deny action.

Across dozens of individual signature-rule JSON files in this batch, at least 40 distinct named commercial VPN products carry dedicated detection signatures (by IP src/dst, FQDN, JA3 hash, or protocol-specific payload): BeePassVPN, BetternetVPN (separate isakmp/ja3/WireGuard sub- signatures), BigMamaVPN, BravePrivateVPN, ExpressVPN (JA3 + UDP-payload variants), FlyVPN, hidemevpn (OpenVPN UDP payload), JourneyVPN, JumpjumpVPN, LetsVPN (JA3), Psiphon3 ("psiphon3vpn_serverip"), QuarkVPN, SecureVPN, SuperUnlimitedVPN, TurboVPN, VPNHero, VPNTurkey, and others, each dated/versioned (e.g. "_20240812") indicating an ongoing signature-maintenance pipeline rather than a one-off ruleset.

Dozens of near-identical MAAT/AppSketch-style signature-definition files in this batch each target one named commercial VPN app/service (ProtonVPN, HotSpot VPN, SuperUnlimitedVPN, jumpjumpVPN, BetternetVPN, NotVPN, VPNLite, SymlexVPN, etc.), matched via FQDN suffix, destination IP list, and/or TLS JA3 hash. The same "SuperUnlimitedVPN" signature recurs with new signature_ids/dates roughly monthly (2024-06, -09, -10, -11), showing this is a continuously-maintained, actively-updated VPN-detection pipeline rather than a one-off snapshot.

Leaked per-app JSON signature-rule exports show the detection engine layers multiple independent signals per VPN product: Windscribe VPN is matched via fixed-offset hex-byte keywords inside the OpenVPN UDP handshake payload (offset 73/depth 78 and offset 81/depth 85); "Super Unlimited VPN" (the SuperUnlimitedVPN app) gets three parallel signatures — a JA3 TLS-fingerprint match, an FQDN match, and an OpenVPN c2s-payload-prefix/length match; Hide.me VPN is matched on OpenVPN UDP payload; FastVPN is matched purely by known source/destination IP; and "V2VPN" (Google Play com.v2ray.v2vpn, a V2Ray-based VPN app) has its own dedicated signature — evidencing a deliberately layered (protocol-fingerprint + TLS-fingerprint + FQDN + IP) detection strategy per circumvention product.

The same M22-tagged VPN-finder plugin set implements active UDP probing to elicit and harvest server IP addresses from CyberGhost VPN's protocol behavior, and separately fingerprints Windscribe VPN via TLS certificate features — two distinct active-probing/fingerprinting techniques feeding the same central VPN-blocklist knowledge base, with an explicit configurable active-scan packet rate.

A MESA Lab research note dissects the Tor Project's webtunnel pluggable transport (HTTPT-based) and reports a hands-on pcap analysis of 3 public webtunnel bridges: the observed TLS ClientHello/cipher-suite fingerprint and SNI matched an ordinary HTTPS connection to the bridge's cover domain, with no certificate visible in the captured handshake -- i.e. these deployments withstood the lab's basic passive TLS-layer fingerprinting attempt at the time of writing.

A single AppSketch/MAAT signature bundle for Windscribe VPN fuses four independent detection techniques in production: an OpenVPN-handshake payload signature (fixed hex bytes at a specific byte offset/depth plus an exact packet-length check), a TLS certificate subject_organization_name match on the literal string 'Windscribe', a JA3 TLS-ClientHello fingerprint hash, and DNS query-name matching against 100+ individually enumerated obfuscation/CDN-rotation domains (windscribe.com plus many totallyacdn.com and whiskergalaxy.com hostnames).

An internal methodology memo for the '新疆省口项目' (Xinjiang provincial-gateway project) and '联通IDC项目' (Unicom IDC project) states both rely on 现有的系统（TSG与CN） (the existing TSG and CN [CyberNarrator] systems) to Block/Monitor/identify mainstream apps, and describes a largely manual signature-engineering workflow -- download the target app, capture traffic with 科莱(Colasoft Capsa)/Wireshark, extract fingerprint fields (http.host, http.user_agent, quic.sni, ssl.handshake.extensions_server_name) by hand, and register a per-app signature -- extending CyberNarrator's documented role beyond Psiphon-IP-harvesting/Pakistan subscriber correlation into general domestic app-blocking alongside TSG.