TSG's DPI engine (sapp) runs a pluggable protocol-decoder framework called 'stellar' (packaged as 'stellar-on-sapp'); its HTTP decoder module (TSG-20446) explicitly parses and can intercept HTTP CONNECT-method tunnels and HTTP/1.1 Upgrade (HTTP/2, WebSocket) transitions, so traffic tunneled or upgraded inside HTTP is visible to and selectively handled by TSG's classification pipeline rather than passing through as an opaque TCP stream.
feat: TSG-20446, support http tunnel with CONNECT method. ... Add BODY_START message; add http_url_decode(); disable CONNECT tunnel.
Defense implications
- Avoid disguising a transport as an HTTP CONNECT tunnel or HTTP Upgrade (WebSocket/H2) against Geedge/TSG-protected links without additional obfuscation -- TSG's 'stellar' HTTP decoder explicitly parses these transitions rather than treating them as opaque byte streams.
Cited artifacts
mesalab_git—stellar/http_decoder.bundle(git-commit-log)
Provenance pointers only — no leak content is rehosted. Raw material: Enlace Hacktivista.
Related findings
An internal TSG operations/troubleshooting manual lays out TSG's full traffic pipeline (NIC -> mrzcpd/marsio capture driver -> sapp DPI engine -> firewall/proxy(KNI->TFE)/active-defense/WAN-NAT policy branches) and shows engineers using maat_redis_tool to pull the live Redis-synced blocking policy tables (TSG_SECURITY_COMPILE, TSG_OBJ_IP_ADDR, TSG_OBJ_APP_ID) and filter them by numeric policy/object ID to debug why a block rule isn't firing.
The internal 'MAAT网络流处理配置统一描述框架' engineering manual (v3.1.20, author 郑超, 2021) documents MAAT's Redis-synced rule-compilation framework and its RuleScan/Hyperscan-based pattern-matching engine (libmaatframe.so / librulescan), and records that RuleScan's fast-scan feature caused a production outage on 2019-03-19 and has been disabled ever since.
An internal MESA Lab report (2021-06-25) documents a live test in which a second sapp instance ('sapp B') receives fully decrypted plaintext HTTP traffic via a Unix domain socket from a third-party TLS decryption platform, while sapp A separately captures raw ciphertext via the mrzcpd driver. 24 hours of the decrypted logs (319,569 HTTP records) show the top intercepted destinations are Facebook, Twitter, Google/YouTube and Instagram, with client IPs traced to residential China Telecom/Unicom/Mobile subscribers in Guangdong, Zhejiang and other provinces.
Raw structured DNS event logs (dated 2021-08-23) captured from what appears to be a GFW-adjacent DNS monitoring/injection pipeline use a schema purpose-built for DNS response forgery ('CHEAT_TYPE', 'CHEAT_RCODE', 'CHEAT_STRATEGY', 'CHEAT_RR', 'INJECTED_PKT_FILE' fields) alongside per-query geolocation; sampled records show lookups for facebook.com and tiktokv.com originating from residential China Telecom/Unicom/Mobile subscriber IPs in Guangdong, Zhejiang, Anhui and other provinces, resolving against both domestic and foreign (8.8.8.8, OpenDNS) resolvers.
A 2023 MESA Lab monthly report describes the 'TF' project's active-defense work: test cases for serial ('串联') HTTP hijack/tamper and parallel ('并联') DNS race-injection ('DNS抢答'), performance tuning that scaled active-defense capacity from 3 to 5 units, converting two existing TSG boxes to active-defense mode, and rewriting the active-defense flow-control logic to no longer depend on sapp.
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.