Analytical Modeling of Denial-of-Service Attacks in Blockchain Shardings

Sharding in blockchains divides nodes into smaller groups which allow for partial transaction processing, relaying and storage. Accordingly, multiple blockchains (shards) will run in parallel, resulting in an increase in total throughput. Nonetheless, with shards comprised of less nodes, they will be more prone to security issues. In this paper, we propose a denial-of-service (DoS) attack where attackers are nodes that are eligible to participate in the process of intra-shard consensus. They then overflow a specific shard by proposing biased blocks containing cross-shard transactions where their receipts are all aimed at that shard. In particular, we utilize a queueing-theoretic model to evaluate different conditions in which the proposed DoS attack can successfully be performed or on the other hand avoided. We show that this attack can be performed irrespective of means to prevent shard takeover attack, e.g., randomization. Accordingly, for different loads in the system and attackers capabilities, we derive the maximum number of shards under which this attack can be averted, so that the system can perform safely.