Defeating Denial-of-Service Attacks in a Self-Managing N-Variant System

被引:3
作者
Jones, Jessica [1 ]
Hiser, Jason D. [2 ]
Davidson, Jack W. [2 ]
Forrest, Stephanie [1 ]
机构
[1] Arizona State Univ, Biodesign Ctr Biocomputat Secur & Soc, Tempe, AZ 85287 USA
[2] Univ Virginia, Dept Comp Sci, Charlottesville, VA 22904 USA
来源
2019 IEEE/ACM 14TH INTERNATIONAL SYMPOSIUM ON SOFTWARE ENGINEERING FOR ADAPTIVE AND SELF-MANAGING SYSTEMS (SEAMS 2019) | 2019年
关键词
adaptive systems; security; software systems; GENERATION; EVOLUTION; SECURITY;
D O I
10.1109/SEAMS.2019.00024
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
N-variant systems protect software from attack by executing multiple variants of a single program in parallel, checking regularly that they are behaving consistently. The variants are designed to behave identically during normal operation and differently during an attack. When different behavior (divergence) is detected, N-variant systems self-heal by either rolling back to a safe state or restarting. Unfortunately, an attacker can create a denial-of-service (DoS) attack from a diverging input by using it to force an N-variant system into an endless diverge/restart cycle. This paper describes a defense, CRISPR-Inspired Program Resiliency (Crispy), that automatically protects N-variant systems from such DoS attacks. Crispy mitigates DoS attacks against N-variant systems using an automatic signature generation technique modeled on CRISPR/Cas, the bacterial adaptive immune system. Experiments on two webservers using exploits developed by an independent Red Team showed Crispy protected against 87.5% of DoS attacks with zero false positives. Overhead was minimal and varied according to the number of signatures maintained, which can be tailored to the threat model and performance requirements.
引用
收藏
页码:126 / 138
页数:13
相关论文
共 73 条
[1]  
Alcoy P., 2018, TECH REP
[2]  
Ali J., 2017, NEW DDOS LANDSCAPE
[3]  
[Anonymous], 1997, NEW SEC WORKSH
[4]  
[Anonymous], 2019, CVE20199590
[5]  
[Anonymous], 2018, CVE20188011
[6]  
[Anonymous], 2018, CVE20181303
[7]  
[Anonymous], 2003, Proceedings of The 10th ACM Conference on Computer and Communications Security
[8]  
Antonatos S., 2005, P 2005 ACM WORKSHOP, P30, DOI DOI 10.1145/1103626.1103633
[9]  
Barrantes G., 2003, P 10 ACM C COMP COMM
[10]  
Berger ED, 2006, ACM SIGPLAN NOTICES, V41, P158, DOI 10.1145/1133981.1134000