6.857: Computer and Network Security
Term Project Pitches
-
Captive portals (for WiFi and wired networks)
- Adam Belay abelay
- Ben Huan benhuan
- Joe Huan joehuan
- John Dong jdong
-
P2P networking capabilities for NaCl
Did some previous work in 6.893 on adding network access capabilities
to NaCl (Google's Native Client) applications. Previously looked at
server-client interactions, now wants to look at challenges unique to
building P2P networks out of the NaCl clients.
-
Privacy and security in social networks
- Christian Ternusternus
- Geoffrey Thomasgeofft
- Ian Ynda-Hummelianyh
- Jacky Changkyoki
End goal: design a social network with good privacy guarantees without
crippling functionality too much. There is some similarity beween
social networks and DHTs, so looking at some results in that area.
A scheme for distributed PKI certificates that would allow mutual
authentication without leaking user interactions. Understanding what
information is leaked accidentaly from a social network, like
everyone's sexual orientation and people's location from picture (PNG)
geolocation information.
-
First pre-image attacks on crypto hashes via hash chains
Looking at existing work on using hash chains to invert crypto hashes.
Hash chains take the idea of rainbow tables to the next level, by
reducing the amount of storage required. Investigating breaking A5/1
(GSM encryption) using hash chains.
-
Privacy on facebook
What are you sharing on Facebook? Wants to build a Facebook
application that acts as an advisor. The app would look at your
profile information that is available to 3rd-party apps and to
random users, and compare that to your desired privacy levels (that
you would indicate to the app). The output will be suggested changes
to your Facebook privacy settings.
-
Wireless network security
Wireshark (formerly known as ethereal) is awesome for sniffing wired
and wireless networks. Wants to use it to understand how much
information we leak right now, by connecting to shared wireless
networks. Also interested in cellphone wireless security (can talks be
intercepted by going near a cellphone? near a tower?) and Bluetooth
security.
-
TOR vs a government
Wants to understand the privacy offered by TOR to human rights
activists in China. Also interested in related TOR problems:
privacy issues stemming from the fact that DNS lookups aren't routed
through TOR, network abuse by BitTorrent users.
-
Shared accounts in online services
Invesigating groups and group accounts (accessed by multiple users) in
online services. Problems: limiting damage when a user's credentials
are leaked, limit cases changing group membership (2 users: no way
to vote on membership changes; many users: voting would take too
long).
-
Honey-client (honeypot) for P2P networks
Wants to implement a honey-client. The main issues are making the
honeypot hard to detect (so that malicious peers don't simply avoid
it) and understanding what kinds of data should the honeypot collect.
-
Security risks of rooting / jail-breaking mobile devices
Many users like to root their smart-phones, with an emphasis on
Android and iPhone devices. Rooting exposes features that are locked
down by carriers, such as tethering and innovative applications.
However, rooting also carries security risks, as it compromises the
mobile operating system's application isolation mechanisms. This
project will focus on the Android OS, and will look at the application
isolation isses. A particularly interesting issue is the SD card,
which is used for shared storage by applications, and doesn't have
any isolation mechanisms in place (uses a FAT32 filesystem).
-
Advanced CAPTCHAs
CAPTCHAs are used to distinguish between human users and computers,
hoping to reduce the abuse (like spam) on open systems. Currently,
CAPTCHAs rely on the hardness of the OCR problem. Sadly, automated
recognizers have made great strides in this area, which means that
CAPTCHAs are becoming increasingly difficult, and thus frustrating for
human users to solve. This project is looking at new methods for
CAPTCHAs, relying on the difficulty of object and feature recognition
in images (e.g. recognize a rotation of a 3-D object, or pictures of
a room from different angles).
-
Browser security
Interested in issues in the JavaScript security model.
-
E-commerce via mobile phones
Interested in using mobile phones as debit / credit cards, especially
in countries without a good infrastructure for processing plastic
card payments. Problems: authenticating a user via phone to the
financial institution, then the financial institution to the retail
chain; encrypting information on mobile devices; protecting against
malicious retailers.
-
Steganographic file systems
Steganography is
hiding information in plain sight
. This is the
only way to prevent against "firehose attacks" where the user is
physically detained by the attacker (e.g., hostile government) and
forced to reveal secrets. StegFS is a file-system that offers
plausible deniability. Interested in mitigating the space-inefficiency
of StegFS, and measuring its performance against established UNIX
filesystems.
-
Making the web of trust usable
PGP is a great PKI (public-key infrastructure) with a distributed
trust model. Unfortunately, current PGP software isn't very useful,
due to the difficulty of navigating the "web of trust" to find a
path to someone's key. This project aims to find a way to find
possible paths on-the-fly, as needed.
-
Survey of crypto hashes
- Daniel Firestonefstone
- Syed Razaraza
Proposing to conduct a survey on SHA-3 candidates. Interested in
learning about successful attacks on previous hash candidates (e.g.,
MD-5, SHA-1) and exploring their applicability to the new candidates
for SHA-3. Hoping to discover an interesting attack on one candidate.
-
Security in medical imagery
- Adian Dalcaadalca
- Katherine Kuankkuan
Interested in exploring the privacy implications of the increased
sharing of medical imagery (which contains a lot of personal
information) via elecronic means. Looking at related security
constraints, like making sure that images come from a certain
hospital, via the use of watermarking. Expecting to apply the results
to the emerging problem of securing human genomes. Will take into
account the legal framework for this scenario (DIACOM, HIPAA). If
time permits, will look into PAX (picture archiving system) and try
to uncover any issues with its access control scheme.
-
Searchable encryption
Aiming to find a good trade-off between security and efficiency on the
searchable encryption problem. The problem is being able to search
an encrypted body of text for keywords, without learning too much
information about the text.
-
Security in P2P networks
Interested in two aspects of P2P security: the search index and the
routing tables. The search index allows users to look up shared files
by name, and find out who is offering them. It can be attacked by
inserting bogus IPs / IDs for popular entries, in order to frustrate
P2P users who would spend a lot of time waiting for connection
time-outs. P2P systems create overlay networks with their own routing
tables. A malicious client can poison the routing tables of its
neigbors. Bogus routing entries can lead to DoS by congesting parts
of the network. Interested in developing some attacks, and proposing
fixes for them.
-
Detecting trojans in compilers
- Nicholas Zehenderzehender
A compromised compiler can insert trojans into the binary it outputs.
This is problematic, because the security issue cannot be detected
by looking at the source code of the binaries. The currently proposed
solution is distributed double-compiling, which requires a trusted
compiler whose output is deterministic and matches perfectly with the
output of the compiler under test. This project aims to relax these
restrictions, either by improving the current approach, or by
proposing an entirely new approach.
-
Privacy of TOR
Wants to understand the privacy offered by TOR (The Onion Router) in
the presence of an adversary with almost complete knowledge (e.g.,
the Chinese government).
-
Security for finger-print readers
Finger-print readers are an increasingly popular method of biometric
authentication. This project aims to explore the security of systems
that include finger-print readers. EAL certifications for these
readers impose constraints on the reading process, but are silent on
methods and security for storing the finger-print (e.g., as an image
or as a wavelet). The certifications are also incomplete with respect
to how finger-print information should be processed. Last but not
least, in some cases, the finger-print information stored in the
reader might be more valuable than the rest of the data secured by the
reader.
-
Analysis of VoIP security
- Heymian Nayakarturd
- Nina Guoarturd
- Rajeev Wongarturd
A survey of the current state of security in VoIP (voice over IP).
Looking at the issue of integrating hardware, extra features such as
caller ID and voicemail, and maintaining good performance. The survey
will cover the currently used protocols: SRTP (relies on AES and
HMAC), ZRTP (session key negotiation using Diffie-Hellman), and Mikey
(proposed alternative to ZRTP offering more methods for key exchange,
and a more complex system for negotiating session keys and security
parameters). Will pay particular attention to how ZRTP and Mikey
prevent man-in-the-middle attacks.
-
Prevent phishing by determining visually similar websites
SSL is a great piece of infrastructure for securing the Web, but it
doesn't match regular users' cognitive model. This project aims to
discover a method for determining whether two different websites look
very similar, and thus one of them should be suspected of being a
phishing attempt.
-
Direct mail phishing schemes
Direct mail has some advantages over e-mail for scammers: spamming is
not illegal, it is much harder to track, and it's more trusted. This
project will experiment with two scams via direct mail, to see their
response rate and assess people's vulnerability to phishing via direct
mail. The first experiment will try to trick recipients into visiting
a website to activate their ATM cards, and the second experiment will
trick people into paying for stock price predictions, by sending
exhaustive predictions to a lot of people, and then selecting the ones
who received correct predictions. Problem: figuring out a framework to
keep the experiments legal.
-
Access control system for FOAF + SSL
FOAF authenticates users using RDF stores that contain published
friendship relationships, and the FOAF (friend-of-a-friend)
relationship. It seems similar to PGP's web of trust. This project
will explore the security of a system using FOAF and SSL, working
towards the author's end goal of building a distributed social
network.
-
IM security and privacy
Exploring the security issues in using today's IM networks. Both the
client binaries and protocols are closed, which exposes users to
potentially malicious clients (root kits, key loggers, etc) and
security threats (there is no guarantee that a screen name is
associated with the same person that it was associated with yesterday,
no guarantees for the security of the data transfer).
-
P2P privacy
Aims to design a P2P publishing system that gives privacy to the
publisher. This is essential for freedom of speech, etc. The main idea
is breaking up the information to be shared into multiple pieces
which must be xor'ed together according to a recipe. Users would
also be able to combine the pieces to obtain work in the public domain
(e.g., linux distributions), to offer plausible deniability. In the
end, this would reduce the problem of privately distributing large
chunks of data to privately distributing small recipes, which can be
solved by TOR.
-
File servers for the generic end-user
Generic = computer-illiterate. Right now, setting up a file server
involves building a server environment, obtaining a static IP, and
setting up port forwarding on a home router. Will go around these
issues by introduing a third-party Web server that intermediates
transfers and potentially stores the file contents. Will achieve
security by designing a minimalistic protocol restricted to GET and
PUT operations, client-server operation, and limited to certain
directories on the user's drive.
-
Security in Prophet
Prophet is a distributed database system. The current implementation
assumes that the parties involved in a sync operation fully trust
each other. This project aims to design a sync operation where the
parties only trust in each other partially. Security will be achieved
by limiting the scope of the sync operation, and using PGP for
authentication, secure data transfer, and data signing.
-
Anonymous file sharing
Explore why anonymous file-sharing hasn't taken off, and propose a
fix. Will focus on college networks, which have high-bandwidth
low-latency connections between all the users. Will survey the many
projects out there, with a special interest on DC (Direct-Connect).
