Inside the Quiet Revolution – How Derandomization Theory in Coding Is Rewriting the Rules of Communication

The idea of eliminating randomness from a system that relies on it has an almost poetic quality. Mahdi Cheraghchi’s 2011 EPFL doctoral thesis, a 206-page document that has subtly grown to be one of the more cited works at the nexus of computational complexity and coding theory, has that peculiar, lovely tension. It will never be read by most people. However, the concepts it contains influence how contemporary communication systems are developed, evaluated, and trusted.

Derandomization theory poses a seemingly straightforward question at its core. It turns out that randomness is very helpful in computer science; it can be used to create error-correcting codes, create quick algorithms, and prove things that mathematicians would otherwise be unable to. However, randomness is costly as well. Sensitive information is difficult to create, verify, and even more difficult to trust in a system. Thus, scientists work for years to achieve the same outcomes with much less of it or without it. It’s similar to a chef attempting to make a dish that tastes better by using half the ingredients.

Three areas are the focus of Cheraghchi’s contribution, and each one feels oddly concrete for something so abstract. The first is the wiretap channel model, in which a portion of your transmission is being eavesdropped on. Imagine an intruder tapping into a fiber-optic line that is stretched across the ocean floor in the dark. No matter where the intruder listens, the math here creates protocols that leak as little as possible. Reading it gives the impression that the work is preparing for issues that we haven’t yet encountered but most likely will.

Group testing, the second area, seems almost unremarkable until you consider how frequently it occurs in real life. The classic example dates back to World War II, when the U.S. Army combined blood samples to more effectively test soldiers for syphilis. The same reasoning reappeared during COVID-19 testing in regions such as Wuhan and parts of India, where laboratories aggregated samples to manage excessive demand. Randomness condensers are used in Cheraghchi’s work to create schemes that function even in situations where test results are unreliable, which is, of course, frequently the case in the field.

Building error-correcting codes that reach the theoretical capacity of communication channels is the third strand. Everything from your phone call to a satellite transmission is subtly supported by this type of work. It doesn’t garner media attention. In popular writing, it is rarely adequately explained. However, the digital world we inhabit on a daily basis would collapse in an afternoon without it.

The way the thesis fits into a longer discussion is noteworthy. Around the same time, researchers like Ronen Shaltiel and Avi Wigderson were investigating extractors and pseudorandom generators. Papers from more recent times, such as Eric Ruzomberka’s 2025 work on derandomizing codes for adversarial wiretap channels, demonstrate that the field is still active, still debating, and still discovering new avenues for investigation. Given that quantum computing forces everyone to reconsider what “secure” really means, it’s possible that we’re just starting to realize how profoundly derandomization will influence future cryptographic systems.

As this field develops, it is difficult to ignore how frequently theoretical work such as Cheraghchi’s becomes infrastructure decades later. The name won’t be familiar to many people creating networks in the future. After subtly shedding its randomness along the way, the math will just be there, performing its function.