14:00
17:00

Peer sampling services are fundamental building blocks of large-scale distributed systems. They provide each node with a continuously updated and nearly uniform view of the system’s population.
Implementations based on gossip protocols are particularly well-suited to this task, thanks to their scalability, self-healing properties, and robustness against node dynamics. 
However, in open and adversarial environments such as public blockchain networks, Byzantine nodes can systematically skew view exchanges, corrupting the sampling distribution and enabling attacks such as eclipse attacks, denial-of-service attacks, or manipulation of consensus protocols.
This thesis addresses the problem of Byzantine-fault-tolerant peer sampling and focuses on proposing scalable and memory-efficient solutions.
 

Amphi LaBRI