Achieving network resiliency using sound theoretical and practical methods

Noureddine, Mohammad

Permalink

https://hdl.handle.net/2142/108512 Copy

Description

Title

Achieving network resiliency using sound theoretical and practical methods

Author(s)

Noureddine, Mohammad

Issue Date

2020-07-17

Director of Research (if dissertation) or Advisor (if thesis)

Sanders, William H

Doctoral Committee Chair(s)

Sanders, William H

Committee Member(s)

• Nahrstedt, Klara
• Basar, Tamer
• Caesar, Mathhew
• Fawaz, Kassem

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• network security
• network resiliency
• game theory
• data-plane programs
• denial of service attacks

Abstract

Computer networks have revolutionized the life of every citizen in our modern intercon- nected society. The impact of networked systems spans every aspect of our lives, from financial transactions to healthcare and critical services, making these systems an attractive target for malicious entities that aim to make financial or political profit. Specifically, the past decade has witnessed an astounding increase in the number and complexity of sophisti- cated and targeted attacks, known as advanced persistent threats (APT). Those attacks led to a paradigm shift in the security and reliability communities’ perspective on system design; researchers and government agencies accepted the inevitability of incidents and malicious attacks, and marshaled their efforts into the design of resilient systems. Rather than focusing solely on preventing failures and attacks, resilient systems are able to maintain an acceptable level of operation in the presence of such incidents, and then recover gracefully into normal operation. Alongside prevention, resilient system design focuses on incident detection as well as timely response. Unfortunately, the resiliency efforts of research and industry experts have been hindered by an apparent schism between theory and practice, which allows attackers to maintain the upper hand advantage. This lack of compatibility between the theory and practice of system design is attributed to the following challenges. First, theoreticians often make impractical and unjustifiable assumptions that allow for mathematical tractability while sacrificing accuracy. Second, the security and reliability communities often lack clear definitions of success criteria when comparing different system models and designs. Third, system designers often make implicit or unstated assumptions to favor practicality and ease of design. Finally, resilient systems are tested in private and isolated environments where validation and reproducibility of the results are not publicly accessible. In this thesis, we set about showing that the proper synergy between theoretical anal- ysis and practical design can enhance the resiliency of networked systems. We illustrate the benefits of this synergy by presenting resiliency approaches that target the inter- and intra-networking levels. At the inter-networking level, we present CPuzzle as a means to protect the transport control protocol (TCP) connection establishment channel from state- exhaustion distributed denial of service attacks (DDoS). CPuzzle leverages client puzzles to limit the rate at which misbehaving users can establish TCP connections. We modeled the problem of determining the puzzle difficulty as a Stackleberg game and solve for the equilibrium strategy that balances the users’ utilizes against CPuzzle’s resilience capabilities. Furthermore, to handle volumetric DDoS attacks, we extend CPuzzle and implement Midgard, a cooperative approach that involves end-users in the process of tolerating and neutralizing DDoS attacks. Midgard is a middlebox that resides at the edge of an Internet service provider’s network and uses client puzzles at the IP level to allocate bandwidth to its users. At the intra-networking level, we present sShield, a game-theoretic network response engine that manipulates a network’s connectivity in response to an attacker who is moving laterally to compromise a high-value asset. To implement such decision making algorithms, we leverage the recent advances in software-defined networking (SDN) to collect logs and security alerts about the network and implement response actions. However, the programma- bility offered by SDN comes with an increased chance for design-time bugs that can have drastic consequences on the reliability and security of a networked system. We therefore introduce BiFrost, an open-source tool that aims to verify safety and security proper- ties about data-plane programs. BiFrost translates data-plane programs into functionally equivalent sequential circuits, and then uses well-established hardware reduction, abstrac- tion, and verification techniques to establish correctness proofs about data-plane programs. By focusing on those four key efforts, CPuzzle, Midgard, sShield, and BiFrost, we believe that this work illustrates the benefits that the synergy between theory and practice can bring into the world of resilient system design. This thesis is an attempt to pave the way for further cooperation and coordination between theoreticians and practitioners, in the hope of designing resilient networked systems.

Graduation Semester

2020-08

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/108512

Copyright and License Information

2020 Mohammad A. Noureddine

Owning Collections