Adaptive Mitigation of Multi-Virus Propagation: A Passivity-Based Approach

Phillip Lee, Andrew Clark, Basel Alomair, Linda Bushnell, Radha Poovendran

Abstract

Traditional malware defense strategies often rely on precise knowledge of propagation rates, which are frequently unknown or variable in real-world networks. This paper proposes a control-theoretic framework for mitigating the spread of multiple malware strains simultaneously. By leveraging the property of passivity in Susceptible-Infected-Susceptible (SIS) models, we develop adaptive patching and traffic filtering rules that guarantee virus elimination without prior knowledge of infection parameters.

Introduction

Securing large-scale networks against malware involves a fundamental trade-off between security and operational efficiency. Aggressive patching and exhaustive traffic filtering can halt an epidemic but may lead to significant downtime and latency. Conversely, insufficient defense allows malware to persist and spread. This challenge is compounded by the presence of multiple malware strains that may compete for hosts or co-exist within them.

Existing mitigation models often assume that the propagation rates of these viruses are known. However, in practice, these rates are stochastic and difficult to estimate. This research addresses the need for "just right" defense strategies—those that are strong enough to eradicate infections while minimizing performance costs—using an adaptive, passivity-based approach.

Modeling Multi-Virus Dynamics

To analyze the spread of multiple viruses, we utilize a mean-field approximation of the network state. Rather than tracking every individual node, we model the probability of infection across the network. A node can exist in several states: susceptible (empty set), infected by a single strain, or co-infected by multiple strains.

State transition diagram for multi-virus propagation.

As shown in Figure 1, the transitions between these states represent the dynamics of infection and recovery. In a multi-virus environment, strains can be competing (where one virus displaces another) or co-existing (where multiple viruses inhabit the same host). The mathematical framework accounts for these interactions, providing a comprehensive view of the epidemic's evolution.

Passivity-Based Control Framework

The core of the proposed solution is the application of Passivity-Based Control (PBC). We demonstrate that the SIS propagation model can be characterized as an "Output Feedback Passive" (OFP) system. In control theory, a passive system is one that does not generate internal energy; in this context, it implies that the malware's growth can be suppressed by a dissipative force—the defense mechanism.

Feedback control loop for adaptive malware mitigation.

Figure 2 illustrates the integrated control architecture. The system consists of three primary components:

  1. Propagation Dynamics: The underlying model of how viruses spread through the network.
  2. Filtering: An adaptive mechanism that adjusts the probability of packet inspection ($q$) based on detected traffic.
  3. Patching: An adaptive mechanism that increases the patching rate ($\beta$) in response to detected infections.

Adaptive Defense Strategies

The framework derives two main adaptive rules that do not require knowledge of propagation rates:

  • Adaptive Patching: The patching intensity increases automatically whenever the infection level rises, ensuring that the recovery rate eventually overcomes the infection rate.
  • Adaptive Filtering: Traffic inspection rates are dynamically adjusted based on the volume of malicious traffic detected, optimizing the balance between security and network latency.

When propagation rates are known, the authors provide a convex optimization formulation to determine the minimum-cost patching strategy that still guarantees total virus removal.

Key Insights and Results

Numerical simulations on Erdos-Renyi random graphs (100 nodes, $p=0.2$) validated the theoretical findings. Key insights include:

  • The Passivity Index: The "passivity index" of the network dynamics identifies the critical threshold for defense. If the patching rate exceeds this index, the malware is guaranteed to be eradicated.
  • The Independence Assumption: The common assumption that a node's state is independent of its neighbors is shown to be conservative. Defenses designed under this assumption are more than sufficient for real-world networks where dependencies exist.
  • Strain Interaction: Competing viruses are generally easier to manage than co-existing ones, as the competition between strains naturally aids in reducing the prevalence of any single virus.

Conclusion

This work provides a robust mathematical foundation for adaptive malware mitigation. By treating propagation as a passive dynamical system, network administrators can deploy defense strategies that automatically scale to the threat level, ensuring network security even against unknown or evolving malware strains. Future research will explore decentralized control laws and extensions to SIR/SIRS models to accommodate a wider range of biological and digital epidemics.

Summary updated

Abstract

Malware propagation poses a growing threat to networked systems such as computer networks and cyber-physical systems. Current approaches to defending against malware propagation are based on patching or filtering susceptible nodes at a fixed rate. When the propagation dynamics are unknown or uncertain, however, the static rate that is chosen may be either insufficient to remove all viruses or too high, incurring additional performance cost. In this paper, we formulate adaptive strategies for mitigating multiple malware epidemics when the propagation rate is unknown, using patching and filtering-based defense mechanisms. In order to identify conditions for ensuring that all viruses are asymptotically removed, we show that the malware propagation, patching, and filtering processes can be modeled as coupled passive dynamical systems. We prove that the patching rate required to remove all viruses is bounded above by the passivity index of the coupled system, and formulate the problem of selecting the minimum-cost mitigation strategy. Our results are evaluated through numerical study.
BibTeX
@article{lee2018adaptive,
  author = {Phillip Lee and Andrew Clark and Basel Alomair and Linda Bushnell and Radha Poovendran},
  title = {{Adaptive Mitigation of Multi-Virus Propagation: A Passivity-Based Approach}},
  journal = {IEEE Trans. Control. Netw. Syst.},
  year = {2018},
  url = {https://openreview.net/forum?id=w395bg-LtFyl},
}
Figure