ARO: Advanced Security Games For Cyber-Physical Systems
Abstract
Advanced sophisticated cyber attacks is a major concern for the nation s infrastructure systems and the information technology systems in corporations. These attacks, often classified under the name, Advanced Persistent Threat, (APT), are launched by highly motivated attackers with abundant resources, and are persistent in compromising a system as long as the expected payoff is high. These attacks lead to significant degradation of our technological advantage and could inflict massive damage to our nation s infrastructure and its security. They are extremely difficult to combat because they are inherently adaptive, exhibiting dynamic behavior in response to defense actions. Developing defense mechanisms against these APT attacks is critical to the safety and security of our nation, its technological resources, and its secrets. Many of today s cyber-physical systems (CPS) are organized in a networked and layered structure; each layer carries out a specific set of functions, which may depend on functions or impact functionality implemented at other layers. Such dependence is beneficial to the normal operation of the system, but can also be utilized by an attacker to harm the system. Moreover, a myopic defense action focusing on a specific attack observed at one level may lead to inefficiencies or vulnerabilities at other levels and trigger new attacks. Therefore, an efficient defense strategy against APT attacks must take the interdependencies among the components in a networked system into account. For a networked system with multiple interdependent levels and that are possibly operated by multiple entities with self-interest, it becomes even more critical to design defense mechanisms that are aligned to their incentives, for both the defenders and the attackers. The overall goal of the proposed research is to study game theoretical models to understand the incentives and fundamental tradeoffs involved in defending/attacking multi-level networked systems, and to design efficient defense strategy accordingly. Although game theory has been extensively applied to cybersecurity and network security, traditional models are mainly static with complete information, and largely ignore the risk of a system being attacked at multiple levels that have inherent dependencies. We first propose a two-player dynamic game with imperfect/incomplete information to capture the persistency and adaptivity of the players. Our game model hinges on the interdependence structure of a networked system, which determines both the action spaces of the players and the information structure of the game. Building upon the dynamic game model, we will address the major challenges in defending against APT attacks by carrying out the following three tasks. Defense against the Unknown: To defend against advanced attackers with unknown or uncertain behavior, we propose to develop adaptive defense strategies that can achieve a guaranteed payoff. By utilizing learning frameworks, we will further investigate the impact of stealth behavior and bounded rationality in the context of security games. Multi-level Attack/Defense: In this thrust, we propose to investigate the impact of multi-level dependencies and design coordinated defense strategies accordingly. We propose to identify both the challenges and the opportunities imposed by the dependencies, and propose strategies that can minimize the impact of the attacks as a whole. Multi-player Security Games: A large CPS often faces attacks of different types in terms of their dynamics and objectives. We propose to extend the game models to allow multiple independent attackers with diverse behavioral patterns. We will further consider the setting when a large system is managed by multiple entities (defenders) with self-interest, hence has to be protected through joint investment.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 15, 2018
- Source ID
- W911NF1510277
Entities
People
- Prasant Mohapatra
Organizations
- Army Contracting Command
- United States Army
- University of California, Davis