Statically Defend Network Consumption Against Acker Failure Vulnerability in Storm

Wenjun Qian, Qingni Shen, Yizhe Yang, Yahui Yang, Zhonghai Wu

Abstract: Storm has been a popular distributed real-time computation system for stream data processing, which currently provides an acker mechanism to enable all topologies to be processed reliably. In this paper, via the source code analysis, we point out that the acker failure and message retransmission result in the consumption of network resources. Even worse, adversary conducts a malicious topology to consume over unconstrained network resources, which seriously affects the average processing time of topology for normal users. Aiming at defending the vulnerability, we design an offline static detection against acker failure in Storm, mainly including the code decompile, the function call relationship and the judgement rules in offline module. Meanwhile, we validate the protection scheme in Storm 0.10.0 cluster, and experimental results show that our mentioned judgement rules can achieve well precision.