Free access to information is a matter of course for us in Germany - but in countries such as China, Russia or Iran it is a utopia. What contribution can research make to an uncensored Internet? An interview with computer scientist Felix Lange about his doctoral thesis.
Your doctoral thesis deals with the topic of censorship, i.e. the censorship of certain information on the internet. What inspired you to do research in this area - and what is your vision?
Here in Germany, most websites are relatively freely accessible. In some countries it's not so easy, for example in China: Wikipedia is censored there, so access to this website is not possible. This severely restricts the availability of information. Another common form of censorship is the censorship of political topics. During the Russian war of aggression against Ukraine, for example, Russia censored many Ukrainian news portals.
Overall, I would like to create as much clarity as possible about how censorship takes place in different countries and find new innovative ways to circumvent this censorship. In this way, I want to make my contribution to an Internet that is as free as possible and to the availability of information.
What research questions are you working on?
My research focuses on what is censored in which countries, i.e. which websites are censored and how they are censored. When users access a particular site, various “messages” are sent. These messages are used to establish the initial connection between the user and the server and to transmit the website data. These are analyzed by censors. Censors are devices or middle boxes that intercept and read the communication somewhere. In my doctoral thesis, I am investigating how censorship can be circumvented: How can you modify the messages so that the censor no longer says: “I have to censor this”.
The important thing is that the server must still be able to accept or read the messages despite the modification. Because if you were to send just any “meaningless” message to bypass the censorship, the message would not make sense to the server either. And that would be of no use to the users.
How do the censors know what to censor?
To explain it with an analogy from everyday life: Imagine you go to a library and want to read certain books, but every time you take a book off the shelf, the librarian comes and reads the title of the book first. Based on the title, he then decides: “Okay, you may read this book” or “You may not read this book”.
With Internet censorship, the title of the book corresponds to the available plain text information that Zensor can read without restriction. In the case of encrypted communication via the Transport Layer Security (TLS) protocol, for example, this corresponds to the so-called Server Name Indication. This provides information about which server a user wants to access. Censors use this information from the plain text to be able to classify “I have to censor this or not”.
So the title of a book in a library would be the same as the Server Name Indication of a website?
Correct. It always depends on what information is available to the censor. If websites are encrypted with the TLS protocol, it is as if the librarian can only ever check the title of the book - i.e. the Server Name Indication - but not the content of the book. If a website is not encrypted and could only be accessed via HTTP, the censor could also take a closer look at the content of the website. In reality, however, this does not happen so often, as most websites are encrypted.
So in order to circumvent censorship, you have to manipulate the title - or the information on which page you want to access?
Exactly. When I access Wikipedia, for example, the Server Name Indication also says “Wikipedia”. This is plain text information that is not encrypted, even if the websites are encrypted using the TLS protocol. The censors therefore use the plain text information to identify what is being accessed. To circumvent censorship, the plain text information in this Server Name Indication could be changed. For example, you could modify the bytes that represent Wikipedia by adding a zero byte in the middle. In other words, you change individual data fields so that the censor no longer knows which server the user wants to access.
To use the analogy from the library: By putting a sticker on the title of the book, for example, you can prevent the librarian from deciding: “You're not allowed to read this book.”
What fascinates you about this field of research?
Overall, censorship is a very topical issue that is developing rapidly. Current technologies, both on the part of censors and on the part of censorship circumvention, are becoming more and more sophisticated. It is a constant cat-and-mouse game between the censors and current research: the censors are constantly trying to develop further in order to better assess and analyze news. Researchers are trying to counter this and find more ways around it.
The subject area is extremely fast-moving: if you look at a censorship machine today and analyze it, it can be out of date in a week. This is challenging, but also extremely exciting.
What innovative approach are you pursuing in your doctorate?
I try to combine individual modifications of messages in order to find even more sophisticated ways of circumventing them. The tools we are developing in the KoTeBi project also help me with this. Brief background: The aim of the KoTeBI project is to check TLS implementations for vulnerabilities fully automatically. To this end, we are researching methods for the automatic detection of security vulnerabilities in the project, which primarily result from the combination of individual tests. As part of my dissertation, I also use the family of TLS attacker tools and draw inspiration from these combinations. I combine individual modifications in order to circumvent censorship as effectively as possible.
What trends do you see in the area of censorship?
At the moment, everything is still in its infancy and censorship can still be circumvented relatively easily. One of the reasons for this is that the censors have to keep the time required for analysis relatively low. After all, there is a huge volume of messages that need to be analyzed. This must be done as quickly as possible so that there is no delay in accessing the site. This is why censorship can currently still be easily circumvented. There is a lot of research into circumvention tactics and many new papers are published every year presenting these new tactics.
A current trending topic, for example, is machine learning: research is currently preparing to find circumvention tactics that are robust against machine learning. The aim would be for the modified messages to be as similar as possible to normal messages so that a machine learning algorithm cannot distinguish between them.
Thank you very much for this outlook and the interesting discussion.
Further information:
Professorship of System Security under the direction of Prof. Dr.-Ing. Juraj Somorovsky at the University of Paderborn: https://cs.uni-paderborn.de/syssec
Project “Combinatorial testing of TLS libraries at all levels” (KoTeBi): https://www.kotebi.de/de/
