60 lines
6.5 KiB
JSON
60 lines
6.5 KiB
JSON
{
|
||
"title": "A potential technique to deanonymise users of the TOR network",
|
||
"type": "REPORTED",
|
||
"description": "The GCHQ technical report documenting the entry-exit traffic correlation attack against Tor — providing the complete mathematical specification, test methodology, experimental results and software reference implementation, with distribution to NSA R1, NSA R4, NSA CES, CSE, DSD, SEQUOIA, ICTR, JTRIG, TEA, GTE and multiple OPC divisions.\n",
|
||
"long_description": "This is UK TOP SECRET STRAP1 COMINT GCHQ technical report OPC-M/TECH.B/61, dated 13 June 2011, produced by OPC-MCR. It is the formal technical paper underlying the OPC-MCR presentation and the MIP GCWiki project page, providing the complete mathematical and experimental documentation of the entry-exit traffic correlation attack against Tor.\n\nThe distribution list is significant: copies were sent to NSA R1, NSA R4, NSA CES, CSE (Canada), DSD (Australia), SEQUOIA, ICTR, JTRIG, TEA, GTE, OPC-HQ, OPC-ALTO, OPC-CNE and OPC-MCR — fourteen named recipients spanning the Five Eyes cryptanalytic and network exploitation community, confirming the attack was considered mature enough for wider dissemination across partner agencies.\n\nThe paper provides the complete technical specification of the attack. Tor's standard protections make circuit tracing essentially impossible: traffic is encrypted in multiple layers, split into standard-size cells, multiplexed across circuits at each router, and subjected to per-circuit rate-limiting store-and-forward buffering that flattens timing features. The attack circumvents all of these by not attempting to follow traffic through the network at all, instead correlating cumulative packet counts at two observation points simultaneously — the exit node and the guard-to-client link — using linear regression.\n\nThe mathematical approach works as follows: time is binned into one-second intervals; for each interval a pair (Ei, Gi) is formed where Ei is the cumulative exit node packet count and Gi is the cumulative guard node packet count; the Pearson product-moment correlation coefficient r is computed between these sequences; a linear model Gi = α + βEi is fitted and only matches where 0.5 < β < 2 are retained; and the traces are slid against each other to account for unknown timing offsets. The r² threshold of 0.998 in the server-to-client direction correctly identifies four out of five test circuits with zero false positives across eight bearer-hours of bulk SIGINT data.\n\nTest data was collected in a carefully controlled experiment approved by JTRIG, using a virtual private server as an intermediate HTTP proxy with a modified User-Agent string to prevent other users from accidentally using the proxy, with packet loggers running on both the client machine and the VPS. Four ten-minute web browsing sessions were conducted covering news browsing, Tor website access, a large file download and forum browsing. Bulk data from four internet bearers was provided by GTE using their high-speed data recorder, with microsecond-accurate timestamps but no packet size, protocol or port information.\n\nThe paper states 13 JTRIG links were visible out of 8294 used. The report also explicitly notes that NSA had already demonstrated demultiplexing of traffic at a Tor exit node, citing a redacted NSA reference, confirming the two agencies were sharing Tor exploitation research.\n\nThe conclusion identifies the two data feeds still required for operational deployment: second-accurate packet logging at controlled Tor exit nodes with circuit identifiers, and second-accurate packet logging of guard-to-client sessions from SIGINT or from running guard nodes. At time of writing JTRIG were investigating exit node data collection and ICTR-FSP were trialling guard node data from research bearers. The paper recommends a first operational test case of deanonymising JTRIG's own Tor usage, both because truth data exists and for information security purposes.\n\nIn plain terms, this document describes a method GCHQ developed to identify who is using Tor — the anonymity network designed to hide a user's location on the internet. Tor works by routing traffic through multiple encrypted hops, making it very difficult to connect an internet user's real IP address to what they are doing online. GCHQ's technique sidesteps the encryption entirely by exploiting a statistical pattern: even though Tor scrambles and re-packages traffic as it passes through the network, the overall rhythm of data flowing in and out tends to match at both ends of the circuit. By simultaneously watching traffic at a Tor exit node they control and watching traffic between the user and the first Tor relay through bulk SIGINT collection on internet cables, GCHQ can mathematically correlate the two streams and identify the user's real IP address. The technique worked on four out of five test cases with no false identifications across eight hours of bulk internet traffic. The document confirms the attack was considered mature enough to share across all Five Eyes partners and was being moved toward live operational deployment, with JTRIG investigating how to collect exit node data and a separate GCHQ team trialling guard node collection from internet bearers.",
|
||
"tags": [
|
||
"Tor",
|
||
"ICTR",
|
||
"GTE",
|
||
"TEA",
|
||
"SEQUOIA",
|
||
"CES",
|
||
"CSE",
|
||
"DSD",
|
||
"TDSD",
|
||
"de-anonymization",
|
||
"traffic correlation",
|
||
"traffic analysis"
|
||
],
|
||
"document": {
|
||
"creator": "GCHQ",
|
||
"date_created": "2011-06-13",
|
||
"classification": [
|
||
"UK TOP SECRET",
|
||
"STRAP 1",
|
||
"COMINT"
|
||
],
|
||
"dissemination": [],
|
||
"banner": "UK TOP SECRET STRAP1 COMINT",
|
||
"genre": "report",
|
||
"programs": [
|
||
"JTRIG",
|
||
"SQUEAL",
|
||
"DISCOVER"
|
||
],
|
||
"countries": [],
|
||
"people": []
|
||
},
|
||
"publication": [
|
||
{
|
||
"publisher": "Der Spiegel",
|
||
"date_published": "2014-12-28",
|
||
"title": "Prying Eyes: Inside the NSA's War on Internet Security",
|
||
"link": "https://www.spiegel.de/international/germany/inside-the-nsa-s-war-on-internet-security-a-1010361.html",
|
||
"reporters": [
|
||
"Appelbaum, Jacob",
|
||
"Gibson, Aaron",
|
||
"Grothoff, Christian",
|
||
"Müller-Maguhn, Andy",
|
||
"Poitras, Laura",
|
||
"Sontheimer, Michael",
|
||
"Stöcker, Christian"
|
||
],
|
||
"contributors": [],
|
||
"doc_url": "https://cdn.prod.www.spiegel.de/media/c10c8842-0001-0014-0000-000000035543/media-35543.pdf"
|
||
}
|
||
],
|
||
"pdf_sha256": "c8491827ffabe5817d4579d21255eb128c184abd6e71fc99a8cef7f5279eacb1",
|
||
"pages": 18
|
||
}
|