CS 3710

Introduction to Cybersecurity

Aaron Bloomfield (aaron@virginia.edu)
@github | ↑ |

Anonymity

Source

This slide set was based, with permission, on one created by Benjamin Kreuter while at the University of Virginia in 2012. He now works for Google, and you can see his work here and here.

Introduction

Why anonymity?

Whenever a big announcement is planned, the White House staff places a big order for pizza from a nearby pizza shop
Journalists begin publishing speculation about the big announcement hours before it is made
What went wrong?

(this is called the Pizza Meter, by the way)

Why anonymity?

Suppose you are running a human rights organization in a hostile nation
- The government is probably watching your communication, so you encrypt your email
A dissident in a nearby town sends you emails, encrypted using your public key
Suddenly the dissident is arrested and sent to a reeducation camp
What went wrong?

Why anonymity?

Encryption is great for preventing people from reading your messages

It does not prevent people from seeing who you correspond with, which in some cases is more important than the content of that correspondence

Receiving Messages Anonymously

Receiving a message anonymously is easier than sending a message anonymously
Intelligence agencies have been taking advantage of this for decades, by broadcasting instructions to covert agents using shortwave radio, which has a global range
We can duplicate this online:
- A message may be sent to a mailing list, posted to Usenet (e.g. in alt.anonymous.messages), uploaded to a blog or wiki, etc.

Broadcasting is not Enough

A medical doctor known for advocating medical marijuana use corresponds with his patients by broadcasting encrypted messages on Usenet and by having them do the same
The DEA raids the doctor’s office, taking his computer and identifying the doctor’s patients
What went wrong?

Sender Anonymity

Protecting the identity of a sender is more technically challenging

We will not address with situations where the use of encryption or of an anonymity system must be hidden.

Overview

Early systems - Penet
Cypherpunks remailers
Some theory
Tor
Attacking anonymity systems
Ethical questions

Should people be required to use their real name online?

In the early 90s, Johan Helsingius created the Penet Remailer as a response to suggestions that people should be required to use their legal names online
- References: 1 and 2

Should people be required to use their real name online?

To use Penet, send an email requesting a pseudonym
Mail sent to the pseudonym would be forwarded to the user’s real email address
A user could send an email to Penet, and have the message forwarded with their pseudonym in the “From:” field

Getting a Penet anon ID

From: daemon@anon.penet.fi
To: <user questing the ID>

You have sent a message using the anonymous contact 
service.
You have been allocated the code name an123456.
You can be reached anonymously using the address
an123456@anon.penet.fi.

If you want to use a nickname, please send a message to
nick@anon.penet.fi, with a Subject: field containing 
your nickname.

For instructions, send a message to help@anon.penet.fi.

(not a real anon ID, by the way)

The Demise of Penet

What could possibly go wrong?

The Demise of Penet

Penet maintained a database mapping pseudonyms to email addresses; a single point of failure!
Numerous requests for specific entries, and for the whole database, had been made while Penet was running -Penet was ultimately shut down after several legal attacks by the Church of Scientology, after secret Scientology material was posted to Usenet through Penet
See the Wikipedia article on the Penet remailer

After Penet

Lessons learned from Penet:

Remailer control messages should be encrypted, so that remailers can be chained. No single remailer should know both the sender and receiver of a message
Records should not be necessary for sending anonymous messages. To receive replies records need to be kept, but should allow for messages to be encrypted and forwarded through other remailers

After Penet

Lessons learned from Penet:

Those wishing to compromise remailers may be willing to go to great lengths, using legal or illegal means

Cypherpunks Remailers

After Penet, the next generation of remailers was built using PGP, a popular email encryption program

To use a Cypherpunks remailer, email messages encrypted, then sent to the remailer in the following format, encrypted using the remailer’s public key:

Cypherpunks Remailers

Cypherpunks remailers have a variety of directives that can be used to improve security:

Remailers can be instructed to re-encrypt part of a message using a symmetric cipher; this is used to setting up reply chains for pseudonymous remailers like Penet
Random padding can be added, or messages may be truncated below a certain line

Cypherpunks Remailers

Cypherpunks remailers have a variety of directives that can be used to improve security:

Messages can rendered invalid after a certain date, or after being remailed a certain number of times.
- This helps prevent replay attacks
The subject line can be encrypted
Random remailer hops can be added

Cypherpunks Remailers

Despite being somewhat ad-hoc, Cypherpunks remailers remain around in some form (see https://remailer.paranoici.org/)

Some remailers provide reliability statistics, by sending test messages through known remailers

Cypherpunks remailers can be used to send messages to email addresses or to Usenet newsgroups

Anonymity Theory

Cypherpunks remailers work well, but they provide only heuristic guarantees of security
- In particular, there is no proof that an adversary could not send specially crafted messages through a remailer to break the security model
David Chaum proposed a more rigorous formulation of security, which imposes stricter requirements than Cypherpunks remailers
- Chaum’s solution is implemented in more advanced remailers called Mixmaster remailers

Anonymity Theory

However, even these remailers do not provide strong security guarantees.
In cryptography research, there is a notion of provable security
- Showing that breaking a cryptosystem’s security is at least as hard as solving a computationally difficult problem, such as the discrete logarithm problem
Consdier ciphertext indistinguishability: “an adversary will be unable to distinguish pairs of ciphertexts based on the message they encrypt”

Anonymity Theory

It can be shown that if the ElGamal encryption system does not satisfy a given ciphertext indistinguishability, then the discrete logarithm problem is computationally easy
- ElGamal is based on, and similar in function to, the Diffie-Hellman key exchange
Since it is widely assumed that the discrete logarithm problem is computationally hard (for good reasons), we can conclude that ElGamal is secure against any chosen plaintext attack

Anonymity Theory

Stronger security guarantees can also be proved
Adaptive chosen ciphertext attacks allow the adversary to request decryptions of any ciphertext except the challenge, both before and after seeing the challenge
Cryptosystems that are secure against such attacks are known; Cramer-Shoup is an analog of ElGamal with this property

Anonymity Theory

We can apply this technique to the analysis of other security systems, including anonymity systems
We need a definition of anonymity that captures our notions about security

Anonymity Theory “Notions”

The adversary might be global - all the messages sent between nodes in the system might be observed
The adversary might be able to take control of some of the nodes, both before and after observing messages (adaptive corruption)
The adversary might send messages through the system, both before and after observing messages
The adversary might be able to choose what messages are sent by honest parties

Anonymity Theory

A rigorous definition of security for an anonymity system was given by Abe and Imai in 2003, which meets all of the above criteria
Camenisch and Mityagin in 2006 proposed a system which meets this definition and which can be securely composed with any other protocol
Following another rigorous definition, Danezis and Goldberg in 2008 presented Sphinx, a provably secure remailer that can be integrated with Mixmaster and Cypherpunks remailers
So we can have a secure remailer!

Anonymity Beyond Email

Remailers are great for high-latency communication like email
Can we get low-latency anonymity?
Can we browse the web anonymously?
In 1998, researchers at the US Naval Research Laboratory released a low-latency anonymity system call Tor, which uses onion routing

Onion Routing (Tor)

Tor

Tor stands for The Onion Router
Consdier a path to destination S through a network:
- Y (client) <-> A <-> B <-> C <-> S (server)
- A, B, and C are randomly chosen nodes in the Tor network
- Each node has a public key $pk_A$, $pk_S$, etc.
We have two functions we will use:
- $enc(pk,X)$: encrypt $X$ with public key $pk$
- $sendTo(dest,X)$: a request to send $X$ to destination $dest$

Tor Terminology

Given path: Y (client) <-> A <-> B <-> C <-> S (server)
A is the entry relay or a bridge
- These are generally well known and published
B and C are just relays
- Some are well known, others are kept secret
As for S:
- If the detination is outside of Tor, then it sends the data (via TLS) to the destination, and S is called the exit relay
- If the destination is inside of Tor, then S is the hidden service

How Tor works

Given path: Y (client) <-> A <-> B <-> C <-> S (server)
Given primitives: $enc(pk,X)$ & $sendTo(dest,X)$
To send $p$ to $S$, we construct:
- $m_1 = enc(pk_S,p)$
- $m_2 = sendTo(S,m_1)$
- $m_3 = enc(pk_C,m_2)$
- $m_4 = sendTo(C,m_3)$
- $m_5 = enc(pk_B,m_4)$
- $m_6 = sendTo(B,m_5)$
- $m_7 = enc(pk_A,m_6)$
$m_7$ is then sent by the client to entry relay $A$

The final message

$m = enc(pk_A,sendTo(B,enc(pk_B,$
$sendTo($$C,enc(pk_C,$
$sendTo(S,enc(pk_S,p)))))))$
Each successive message is encrypted, forming layers
- Layers like an onion, hence the name onion routing
Note that the final node ($S$) is the final node inside Tor
- Either the exit relay or the hidden service

Tor overview

Tor can be viewed as a chain of proxy servers, which is periodically changed to make attacks harder
Each Tor connection goes through three (or more) nodes: the entry node, middle node(s), and exit node
To connect to Tor, the client first downloads a list of all Tor nodes from several directory authorities
The nodes used for connections will be selected offline, so that no information about a client’s route will be leaked

Hidden Services

Tor also provides hidden services
- This allows a server to be run anonymously
Each hidden service selects several middle nodes to act as “guards” (or introduction points) which are connected through additional Tor nodes to the hidden service
- Clients can look up guards for a hidden service in a distributed hash table
- Thus, the client has to route to the introduction point, and the introduction point will route to the hidden service

The Silk Road

A Tor hidden serivce at silkroad7rn2puhj.onion
- Named after the historical trade route of the same name
It sold illegal drugs online
Run by the “Dread Pirate Roberts” (DPR), it was shut down in 2013
- Others quickly took its place

How was the Silk Road busted?

Two versions (from Wikipedia)
- “The FBI has claimed that the real IP address of the Silk Road server was found via data leaked directly from the site’s CAPTCHA”
- Security researchers believe that the PHP login page was manipulated to output its server variables, including the IP address
Ross Ulbricht, alleged to be DPR, was arrested in Oct 2, 2013
- The FBI siezed about $90 million in bitcoins
- He was sentenced to life in prison w/out parole

Tor stats

Tor is currently the most popular anonymity system in use, with hundreds of thousands of users
These metrics are as of November 2018 and are from https://metrics.torproject.org/:
- 653 entry relays (“bridges”)
- 6,398 (internal) relays
- About 2 million users
- 107,490 unique onion services
  - Some estimates indicate that about half are inactive

How to detect Tor

Assuming one can monitor a network link…
If TLS encrypted data is sent to a Tor entry node, then one is likely using Tor
Many gov’t’s (including the US) will pay “more” attention to you if it seems you are using Tor

Tor usage

Tor is often used to defeat national firewalls; an ongoing area of work is disguising a Tor connection as a typical Firefox TLS connection
A common pattern has emerged in such countries: Tor is typically blocked just before big announcements by the government, or immediately following news of rebellions or unrest in other countries

US gov’t policy on Tor

Some agencies support it: US state department, NSF, and Radio Free Asia
Some oppose it: NSA, FBI, CIA

Would you buy an Internet connected sex toy?

Apparently a lot of people do, but they keep getting hacked
- Wired article on this
Security idea: have the sex toy use Tor
This is a field of study called “onion dildonics”
- I promise that I did not make that name up…
While the application is somewhat humorous, the privacy implications involved are very real
Tor has been suggested as the answer to this

Tor: The Benefits

It allows for “true” anonymity online
It allows for much better privacy than otherwise is available
The gov’t cannot figure out what you are saying or doing
- (Although they can tell if you are using it)

Tor: The Drawbacks

This is how evil stuff is propogated online
- Child pornography, snuff films, torture films, rape videos, etc.
- The real stuff, not fake films done by actors for entertainment

Tor: Ethics

Considering the benefits and drawbacks, is it ethical to use Tor?

Tor: Ethics

Considering the benefits and drawbacks, is it ethical to use Tor?
If you are using for ethical means (privacy, for example), then yes
- Just because a technology can be used for illegal means does not automatically make that technology unethical
- If that were the case, then all of computers would be illegal

Attacking Anonymity Systems

Attacking Remailers

Global eavesdropping: watching all traffic through all remailers, and following one message
- This can be thwarted using random delays, random padding, and message reordering

Attacking Remailers

Replay attacks: sending the same message through a Remailer, and seeing what comes out
- This can be thwarted with maximum date headers, assuming PGP is non-malleable (which could be untrue), and does not work against Mixmaster remailers

Attacking Remailers

Attacking specific remailers: if the adversary has some remailers under his control, he can flood other remailers in an attempt to prevent them from being used

Attacking Remailers

Spamming attacks: If a pseudonym delivers messages to an email address, through a chain of remailers, the chain could be traced by an attacker who sends many “spam” messages to the pseudonym
- This can be thwarted by having messages delivered to a Usenet newsgroup or by limiting how many messages can be sent before the remailer chain is changed

Attacking Remailers

Mail loops: A pseudonym may be configured to have replies sent through several remailer chains; this could be used to create loops that exponentially increase the load on a remailer
- Typically, the amount of mail a pseudonym can receive in a given day before being disabled is deliberately limited

Attacking Tor

Global eavesdropping: Tor does not have random delays or padding, so a global eavesdropper can defeat Tor

Attacking Tor

Fingerprinting: a local eavesdropper can collect information about latency patterns for specific systems accessed through Tor, and check for such patterns in a target connection
- This is particularly bad in cases where some information about a person’s identity is leaked; for example, forum posts that use regional spellings or phrases

Attacking Anonymity Systems

Malicious exit nodes: if end-to-end encryption is not used, a malicious exit can sniff usernames, passwords, and other information

Ethical Questions

Should anonymity systems even be deployed?
- Criminals, terrorists, etc. may use anonymity systems to avoid detection or prosecution
- Anonymity systems have been used for harassing people and by child predators
- Human rights groups, political dissidents, law enforcement agencies, etc. also use anonymity systems
- In some cases, anonymity systems could save lives

Ethical Questions

Should attacks on anonymity systems be published?
- Law enforcement agencies may be given an advantage with unpublished attacks
- Such attacks are likely to be discovered and kept secret by hostile governments, criminal hackers, etc.