Latest / CyberPodcast / How Tor Gets Deanonymized with Cell Counters | Paper Breakdown #1 | CyberPodcast

Transcript

How Tor Gets Deanonymized with Cell Counters | Paper Breakdown #1 | CyberPodcast · CyberPodcast

• 0:00
  Welcome to this explainer. Today we're diving straight into
• 0:03
  a totally fascinating vulnerability hidden deep inside
• 0:07
  the world's most famous anonymity network, Tor.
• 0:11
  O Look, if you wanted to hide your tracks on the Internet,
• 0:13
  you'd probably use Tor, right? But what if Tor isn't quite as
• 0:17
  invisible as you think? It turns out the way Tor handles
• 0:20
  pockets of data actually has a pretty massive structural quirk,
• 0:24
  and researchers have figured out exactly how to exploit it using
• 0:27
  something they call a cell Counter based attack, which
• 0:30
  basically shatters that invisible shield.
• 0:33
  We're going to break this digital heist down into five
• 0:36
  parts. The promise of Tor, anatomy of
• 0:38
  an attack, encoding the secret signal, surviving network chaos,
• 0:42
  and finally unmasking the users. Let's get right into it.
• 0:47
  Section 1 The Promise of Tour Before we can break the system,
• 0:51
  we really need to understand how it's actually supposed to
• 0:53
  protect you in the 1st place. OK, so picture Alice, our
• 0:57
  client, and Bob, the server in standard tour.
• 1:00
  Alice's data is chopped U into equal sized 512 byte packages
• 1:05
  called cells. These cells are wrapped in heavy
• 1:08
  layers of encryption, you know, like an onion, and then bounced
• 1:11
  through an entry, a middle, and an exit router.
• 1:14
  The whole point here is to completely separate her identity
• 1:17
  from Bob, but this brings us to an incredibly interesting
• 1:21
  concept called active watermarking.
• 1:24
  Instead of just passively watching traffic and crossing
• 1:26
  your fingers to find a pattern, an attacker actively forces a
• 1:30
  pattern right into the data stream.
• 1:32
  It's kind of like slipping a bright neon piece of Aer into a
• 1:35
  massive stack of standard mail and then just waiting at the
• 1:37
  destination to see exactly where that bright color OS out.
• 1:40
  Section 2 Anatomy of an Attack. So how do they actually pull off
• 1:45
  this covert heist? Well, it's a simple four step
• 1:48
  blueprint. First, they target traffic at a
• 1:50
  malicious exit router. Second, the encode a secret
• 1:53
  signal. Third, an accomplish records the
• 1:56
  incoming packets over at the entry router.
• 1:58
  And finally, they decode the signal to confirm the user's
• 2:00
  identity. And honestly, the craziest part
• 2:03
  about all of this? Because Tor relies entirely on
• 2:05
  volunteer nodes, setting up this trap is literally as easy as
• 2:09
  just donating a couple of computers to the network.
• 2:11
  The core vulnerability here lies in how these onion routers
• 2:14
  actually process the data. Cells arrive, and they just sit
• 2:17
  in a circuit queue. Then, when it's time to
• 2:20
  transmit, all those waiting cells are flushed into an output
• 2:23
  buffer and sent out into the network in one big batch.
• 2:27
  And that batching process? That is exactly what the
• 2:29
  attacker is going to manipulate. Section 3, encoding the secret
• 2:34
  signal. This is where the real magic
• 2:36
  happens. The attacker basically smuggles
• 2:39
  a binary signal right into the traffic stream by messing with
• 2:42
  those buffers. They flush exactly 3 cells to
• 2:45
  mean A1, and just one single cell to mean a 03 cells, 1 cell,
• 2:50
  three cells. It's essentially a digital Morse
• 2:52
  code of ones and zeros, perfectly hidden right there in
• 2:55
  the flow of traffic. Now, to keep honest nodes from
• 2:58
  noticing this trick, the attacker spaces these bits out
• 3:01
  randomly using a pseudo noise code.
• 3:04
  This is basically just a randomized timing sequence that
• 3:07
  only the attacker knows. So to literally anyone else
• 3:10
  monitoring the network, these manipulated bursts just look
• 3:13
  like the normal chaotic ebb and flow of Internet congestion.
• 3:17
  Section 4, Surviving network chaos.
• 3:20
  But what happens when this perfectly precise secret Morse
• 3:23
  code actually hits the wildly turbulent environment of the
• 3:27
  open web? Because, you know, network delay
• 3:29
  and congestion act like pure chaos.
• 3:32
  If an attacker sends three cells to represent A1A, super busy
• 3:35
  router in the middle might process one of those cells
• 3:38
  immediately, but delay the other two.
• 3:39
  Suddenly. Boom, the signal is distorted
• 3:42
  and the accomplice waiting over the entry node isn't seeing that
• 3:44
  clean 3 cell pattern anymore. To survive this chaos, the
• 3:48
  attackers built some incredibly robust recovery algorithms to
• 3:52
  stitch those distorted signals back together.
• 3:54
  So whether the original signal gets divided up into separate
• 3:57
  cells, merged with completely different signals, or just
• 3:59
  squashed into one massive packet, these mathematical
• 4:02
  algorithms reassemble all the broken pieces.
• 4:05
  This means the attacker can read the original binary code
• 4:07
  perfectly no matter what happens in transit.
• 4:09
  Section 5, unmasking the users. So just how successful is this
• 4:14
  attack at actually stripping away your anonymity?
• 4:17
  Well, unlike other attacks that need massive, massive amounts of
• 4:21
  data to work, this one is terrifyingly efficient.
• 4:24
  It only needs 10s of cells to confirm a connection.
• 4:28
  Think about that. That means even a super quick,
• 4:30
  fleeting anonymous web request can be fully unmasked and get
• 4:35
  this. If the delay interval between
• 4:37
  those signal bits is set properly, usually around 100
• 4:40
  milliseconds, the attacker's ability to decode the signal and
• 4:43
  confirm the connection approaches an astonishing 100
• 4:47
  percent, 100%. It is virtually guaranteed to
• 4:50
  confirm that Alice's without a doubt communicating with Bob.
• 4:53
  It's also incredibly precise operating with an almost non
• 4:57
  existent false positive rate of one in 100,000.
• 5:00
  So it is absolutely not guessing based on vague timings.
• 5:03
  It is definitively tracking a specifically engineered
• 5:07
  watermark straight across the network.
• 5:09
  The structural impact of this is just massive.
• 5:12
  By compromising nearly 10% of the onion routers on the
• 5:15
  network, an adversary could unmask over half of all
• 5:18
  communication sessions on tour. Since tour routing is random, if
• 5:22
  they control just one in 10 nodes, the mathematical
• 5:25
  probability of your traffic hitting their trap over time is
• 5:28
  actually staggering. I'll leave you with this final
• 5:31
  provocative thought. Tor relies on standardizing
• 5:35
  packet sizes into those 512 byte cells to hide your data.
• 5:39
  But as we've seen in this explainer, the way those exact
• 5:41
  packets are grouped can be directly weaponized against you.
• 5:45
  So if the fundamental architecture of the network
• 5:48
  itself can carry secret, traceable signals, is any
• 5:51
  anonymous network truly safe? Thanks for exploring this
• 5:54
  research with me today. Keep questioning the systems you
• 5:57
  trust and keep learning.