SLIDE 1
Tor: Internet anonymity tool
2
Improving Tor using a TCP-over-DTLS Tunnel Joel Reardon Ian - - PowerPoint PPT Presentation
Improving Tor using a TCP-over-DTLS Tunnel Joel Reardon Ian - - PowerPoint PPT Presentation
Improving Tor using a TCP-over-DTLS Tunnel Joel Reardon Ian Goldberg Google Zurich University of Waterloo 18th USENIX Security Symposium August 12th, 2009 Tor: Internet anonymity tool 2 Problem 3 Tor Network 4 Tor: circuit construction
SLIDE 2
SLIDE 3
Problem
3
SLIDE 4
Tor Network
4
SLIDE 5
Tor: circuit construction
5
SLIDE 6
Tor: circuit construction
6
SLIDE 7
Tor: circuit construction
7
SLIDE 8
Tor: circuit construction
8
SLIDE 9
Tor
Privacy for usable, low-latency communication. However it can be slow, and that discourages casual usage.
9
SLIDE 10
Where is the observed latency?
10
SLIDE 11
Tor’s Datapath
- Input
Buffer Output Buffer Socket Socket Input Buffer Router Output Buffer Output Buffer Input Buffer Onion Router Input Buffer Router Output Buffer Output Buffer Input Buffer Onion Router Server Socket
Onion Proxy
Proxy
Datapath
11
SLIDE 12
Output buffers do introduce some latency
2 4 6 8 10
Time (min)
2 4 6 8 10 12 14 16 Datalength (KB)
(c) Buffer Length over Time
2 4 6 8 10
Time (min)
50 100 150 200 250 300 Waiting Time (milliseconds)
(a) Waiting Times over Time
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Waiting Time (milliseconds)
20 40 60 80 100 Percentile
(b) Waiting Times CDF
2 4 6 8 10 12 14 16
Buffer Length (KB)
20 40 60 80 100 Percentile
(d) Buffer Length CDF
Output Buffer Size and Latency 12
SLIDE 13
This occurs when the socket is unwritable
13
SLIDE 14
A brief outline of TCP
TCP is designed to reliably send streams of data using packets Congestion controls throttles sending to maximize throughput while avoiding packet drops.
14
SLIDE 15
Of what are TCP output buffers composed?
100 200 300 400 500 600 Time (seconds) 5 10 15 20 25 30 35 40 Socket Output Buffer Size (KB)
Socket Output Buffer Size and Unacknowledged Packets Socket Output Buffer Size Unacknowledged Packets Unwritable Socket 15
SLIDE 16
TCP Congestion Control (C/C) is to blame.
16
SLIDE 17
Tor’s multiplexing of circuits over TCP
b b a a a, b OR OR OR OR OR OR
If C/C is applied to a, then it is also applied to b This is suboptimal; TCP is designed to throttle individual connections based on whether they witness a packet drop—proportional to their traffic.
17
SLIDE 18
An example of cross-circuit interference
100 200 300 400 500 600 700 Time (seconds) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Data (MB)
Buffer Sizes across Circuits
18
SLIDE 19
Experiment to observe interference by bulk senders
Circuit Delay Circuit delay while peer under load 200 400 600 800 1000 1200 1400 1600
Delay (ms)
19
SLIDE 20
Packet Dropping / Reordering
OR OR TCP Stream (over network) Readable
Kernel TCP
Buffered / Waiting
20
SLIDE 21
We want to use a separate TCP connection for each circuit
21
SLIDE 22
Concerns for separate TCP connections
Individual TCP streams leak precise information about the size and rate of data to an adversary Tor already faces some scalability concerns regarding its clique topology Some versions of Windows suffer when opening many TCP sockets already Any modification must be backwards compatible with the existing Tor network
22
SLIDE 23
Our novel proposal: a TCP-over-DTLS tunnel
DTLS - a secure (cf. TLS) protocol for transporting datagrams (UDP sockets) TCP implementation in user-space is used to generate TCP/IP packets, which are sent over DTLS The other end injects the received packet into their user-level TCP stack, and reads from user-level sockets
23
SLIDE 24
How TCP-over-DTLS addresses our issues
Application Payload TCP IP IP DTLS Application Payload TLS UDP TORTP (b) TCP−over−DTLS Tor (a) TCP Tor
24
SLIDE 25
How TCP-over-DTLS addresses our issues
UDP operates in an unconnected mode, so it accepts packets from any destination Each node advertises a UDP socket that multiplexes data for all connections The sender is used to demultiplex the proper connection that is used to decrypt the DTLS payload Nodes that do not offer a UDP socket will use the existing transport, assuring backwards compatibility
25
SLIDE 26
Packet / Streams translation
cell_unpack Sockets TCP TCP Buffers cell_pack
Tor Interface
Processing Tor UDP Socket tx() rx() recv() send() Packets Streams DTLS decrypt DTLS encrypt TCP Re/Transmit Thread
User−level TCP
26
SLIDE 27
How TCP-over-DTLS addresses our issues
OR OR UDP Stream (over network)
Kernel UDP
RXed Packets
User TCP
Buffered/Waiting Readable m Readable
27
SLIDE 28
Experimental results from our implementation
28
SLIDE 29
Circuit latency comparison
Base Delay Delay under Load 200 400 600 800 1000 1200 1400 1600
Delay (ms) TCP-over-DTLS Tor TCP Tor
29
SLIDE 30
Future Work
30
SLIDE 31
Improved Memory Management socket 1’s output buffer: 1, 3, 11, 8 empty list: 2, 5, 9, 4 socket 1’s input buffer: 0, 6, 7 cell_t data: 10 Cell Pool 1 2 3 4 5 6 7 8 9 10 11
31
SLIDE 32
Back-propagation of Congestion Window
TCP cwnd=15 actual cwnd=10 data ack, cwnd=10 TCP cwnd=20 actual cwnd=10 data ack, cwnd=10 actual cwnd=10 TCP cwnd=10
32
SLIDE 33
Summary
We determined that TCP congestion control introduces latency into Tor’s datapath We determined that multiplexing circuits over TCP results in the unfair application of congestion control We proposed TCP-over-DTLS: a solution to address this issue that also addresses scalability issues and is backwards compatible with the existing Tor network We implemented our proposal and showed it successfully addressed cross-circuit interference.
33