At system boot up several system daemons utilize the Linux kernel's accumulated randomness via /dev/urandom to generate keys or initialize a pseudorandom number generator (PRNG). The randomness is typically being accumulated by mixing several entropy sources to a pool. The Linux kernel however may operate on various hardware with different specifications, or even in a pure virtual environment without any physical hardware. For that it is desirable to know whether the randomness gathered in the first boot is sufficient for key generation or other cryptographic purposes. The goal of this thesis is to study the Linux kernel boot process on a system, which is a constrained system in terms of hardware. The objectives are:

• Analyse real-world cases of hardware-constrained systems running Linux kernel to be able to determine suitable environment for data collection.
• Study the possibilities of gathering the input of Linux kernel PRNG and modify the kernel source code to enable data collection.
• Collect the data.
• Identify the sources of the entropy and examine the randomness contribution of these sources.
• Evaluate collected dataset using statistical methods for computing entropy contained in the data.
• Interpret the results in context of impact on the quality of the random data provided by Linux Kernel.
• Propose possible improvements and optimizations that would enhance the quality of randomness gathered in the boot process.