Achieving cyber resiliency against lateral movement through detection and response

Cyber resiliency;Trustworthiness;Integrity checking;Anomaly detection;Lateral movement;Response;Detection;Recovery;Response engine;Adaptive response;Game theory;Control theory

Fawaz, Ahmed Mohamad

关键词: Cyber resiliency;    Trustworthiness;    Integrity checking;    Anomaly detection;    Lateral movement;    Response;    Detection;    Recovery;    Response engine;    Adaptive response;    Game theory;    Control theory;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/98194/FAWAZ-DISSERTATION-2017.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Systems and attacks are becoming more complex, and classical cyber security methods are failing to protect and secure those systems. We believe that systems must be built to be resilient to attacks. Cyber resilience is a dynamic protection strategy that aims to stop cyber attacks while maintaining an acceptable level of service. The strategy monitors a system to detect cyber incidents, and dynamically changes the state of the system to learn about the incidents, contain an attack, and recover. Thus, instead of being perfectly protected, a cyber-resilient system survives a cyber incident by containing the attack and recovering while maintaining service.Cyber resiliency has the potential to secure the modern systems that control our critical infrastructure. However, several practical and theoretical challenges hinder the development of cyber-resilient architectures. In particular, an architecture needs to support and make use of a large amount of monitoring; the problem is especially serious for a large network in which hosts send low-level information for fusion. The problem is not only computational; the semantics of the data also creates a challenge. In combining information from multiple sources and across multiple abstractions, we need to realize that the sources are describing different events in the system which are occurring at varying time scales.Moreover, the system is dependent on the integrity of the monitoring data when estimating the state of the system. The estimated state is used to detect malicious activities and to drive responses. The integrity of the monitoring data is critical to making “correct” decisions that are not influenced by the attacker. In addition, choosing an appropriate response to specific attacks requires knowledge of the at- tackers’ behavior, i.e., an attacker model. If the attacker model is wrong, then the responses selected by the mechanism will be ineffective. Finally, the response mechanisms need to be proven effective in maintaining the resilience of the system. Proving such properties is particularly challenging when the systems are highly complex.In this dissertation, we propose a resiliency architecture that uses a model of the system to deploy monitors, estimates the state of the system using monitor data, and selects responses to contain and recover from attacks while maintaining service. Then we describe our design for the essential components of the said resiliency architecture for a multitude of systems including operating systems, hosts, and enterprise net- works, to address lateral movement attacks. Specifically, we have built components that address monitor design, fusion of monitoring data, and response. Our pieces address the challenges that face cyber-resilient architectures.We set out to provide resilience against lateral movement. Lateral movement is a step taken by an attacker to shift his or her position from an initial compromised host into a target host with high value. First, we designed a host-level monitor Kobra that generates different estimations of the state of a host. Kobra combines the various aspects of application behavior into multiple views: (1) a discrete time signal used for anomaly detection, and (2) a host-level process communication graph to correlate events that happen in a network. We use the host correlations to generate chains of network events that correspond to suspicious lateral movement behavior. We use a novel fusion framework that enables us to fuse monitoring events for different sources over a hierarchy. Finally, we respond to lateral movement by changing the topology and healing rates in the network. The changes are enacted by a feedback controller to slow down and stop the spread of the attack.Since our cyber resiliency architecture depends on the integrity of the monitoring data, we propose PowerAlert, an out-of-box integrity checker, to establish the “trustworthiness” of a machine. PowerAlert is resilient to attacker evasion and adaptation. It uses the current drawn by the CPU, measured using an external probe, to confirm that the machine executed the check as expected. To prevent an attacker from evading PowerAlert, we use an optimal initiation strategy, and to resist adaptation, we use randomly generated integrity-checking programs. We pick the optimal initiation strategy by modeling the problem of low-cost integrity checking when an attacker is attempting to evade detection as a continuous-time game called Tireless. The optimal strategy is the Nash equilibrium that optimizes the defender’s cost of checking and utility of detection against an adaptive attacker.