Here are some ideas if you don't know what to write your thesis or similar work about or just want to perform some interesting research. I wish I had time to work on them; unfortunately, I have time only for a small fraction.
An audio+video conferencing multiplatform application providing end-to-end encryption
Code injector (proxy that injects your code to executable files transferred over insecure link), some projects with similar aim:
evilgrade,
mitmproxy How to make basic chemicals like nitrates or antibiotics in amateur conditions when an apocalypse (e.g. nuclear war) occurs; can contain own experiments
Cheap extraction of important ensymes and chemicals for genetic engineering (thermostable DNA-Polymerase, EcoRI, dNTPs) at home without buying them online
Exporting steganographic bits (least significant bits or noise) from WAVs and RAW photos as a block device. Linux nbd seems pretty simple.
Finish dm-steg.
Silicon engineering: An Open Source Trusted Platform Module
Portable multiplatform encryption of a flash drive. Current solutions (TrueCrypt) require admin/root access to install and setup. However, for example an encrypted WebDAV service can be run from unprivileged userspace.
Build a HFCC vest. Commercial implementations cost $10k. Could it be build for one tenth of that price?
rtl-sdr showed us that crappy SDR can cost $10. However, much better A/D converters are sold for similar price in China. Can a better SDR be build for an affordable price?
Write decoder and cracker for MOTOTRBO (it's 40-bit ARCFOUR), build open-source sdr-based client for GSM and Tetra. Reverse-engineer and crack Tetrapol encryption.
Investigate possibilities of radio transmitting using graphic card. Create a gnuradio sink for transmitting arbitrary signals. –
FIXED in
osmo-fl2k Create an implementation of GNSS satellites (NAVSTAR, GALILEO, GLONASS) that runs on customer-grade FPGA and SDR. There already is
a GPS one!
Implement some nice app-level isolation framework into Linux (like Qubes
OS)
Fork Firefox. Fix all that NSA “features” and nasty bugs.
Create sdr-based implementation of TAMARA or VERA passive sensor. Is it possible to fingerprint individual transmitters like rumors about ASMKS say?
Implement some nice
GUI for Bitcoin Contracts and Ethereum, so even non-experts can use it.
Implement cryptoanarchy procedures described
here (Czech only, sorry)
Build a decentralized, censorship-resistant and surveillance-free distributed web search engine. YaCy could be a good starting point
A decent MITM suite for SSHv2 would be cool. Exploiting SSH agent would be even cooler.
OpenStreetMap + SRTM + OpenGL + Oculus Rift = ultimate map explorer.
Create an uncluttered and usable Mapnik theme or just finally write one working renderer for OSM without nasty bugs.
Create Arduino-like toolchain, documentation and examples for beginners for much faster and cheaper ARM LPC processors.
Create a decent Linux worm and some working kernel rootkit. Can binary portability be achieved?
Sonography should be a piece of cake with currently available signal processing hardware. However, such devices still cost $5k+. Fix it.
Even the original probe with piezo elements is cheaper than complete sonograph - an alternative electronics could be built for it.
speaking of which, we still don't have usable ECG and EEG
A $50 30MHz+ data capture card, for example on LPC4370, is a missing instrument.
PDA/smartphone that runs standard Linux, i.e., no Android crippled system. Currently, you can pick from N900 with outdated hardware and Neo900 that costs 1k€+ and will be ready the next year, maybe.
End discussions “which bike is faster” with software that eats data from GPS, power meter and possibly anemometer and computes friction coefficients and air resistance from it.
Bitcoin demonstrated that Sybil attacks can be mitigated, at least to some extent, without a centralized authority. DDoS attacks are a growing concern on current internets. Can an inherently DDoS-safe network be built?
Checking manufactured chips for backdoors is really hard. Is it possible to implement a softcore CPU on reconfigurable hardware that could be trusted even if the reconfigurable blocks are backdoored?
postfwd is no longer developed, eats 300
MB RAM and is pain to configure. Create something better.
I suck at drawing/art/illustrations. Create a graphic editor that interpolates poorly drawn sketches into photorealistic/cool drawings. For example using
pix2pix or
artistic style transfer or
vid2vid.
-