Abstract
Due to physical isolation as well as use of proprietary hardware and protocols, traditional real-time systems (RTS) were considered to be invulnerable to security breaches and external attacks. This assumption is being challenged by recent attacks that highlight vulnerabilities in RTS. Besides, a straightforward integration of security mechanisms might compromise the safety and predictability guarantees of such systems. In this paper, we focus on integrating security mechanisms into RTS (especially legacy RTS) and define a metric to measure the effectiveness of such integration. We combine opportunistic execution with hierarchical scheduling to maintain compatibility with legacy systems while still providing flexibility. The proposed approach is shown to increase the security posture of RTS without impacting their temporal (and hence, safety) constraints.
Introduction
Modern infrastructure, from power grids to unmanned aerial vehicles (UAVs), relies on "hard" real-time systems (RTS). Historically, these systems were deemed secure due to physical isolation. However, increasing connectivity has exposed them to sophisticated cyber-attacks. Integrating security into these systems is non-trivial; a single delayed calculation in a flight controller or power relay can lead to catastrophic failure. This paper proposes a framework to "retrofit" security monitoring into legacy RTS without compromising their strict timing deadlines.
The Challenge: Security vs. Schedulability
The primary obstacle in securing legacy RTS is the "security vs. schedulability" tension. Engineers are often reluctant to add intrusion detection systems (IDS) because they might consume CPU cycles needed for safety-critical tasks. If security checks run too frequently, the system may miss deadlines; if they run too infrequently, attackers may go undetected. This research addresses how to find the mathematical "sweet spot" for security integration in systems where the original schedule is fixed.
Proposed Methodology: Opportunistic Execution
The authors introduce a two-pronged approach to solve the integration problem:
1. The Security Server Abstraction
Instead of allowing security tasks to preempt critical operations, they are encapsulated within a "Security Server" defined by a capacity ($Q$) and a replenishment period ($P$). This server operates at the lowest priority, ensuring it only utilizes "slack time"—the idle CPU cycles left over by high-priority real-time tasks.
2. Optimization via Geometric Programming (GP)
Determining the optimal execution frequency for various security tasks is a complex, non-linear problem. The researchers transformed this into a Geometric Programming problem. By proving that scheduling constraints can be expressed as posynomials, they enabled the use of efficient convex optimization techniques to find globally optimal monitoring periods.
The framework adapts the periods of specific security tasks—such as kernel integrity checks or file system scans—to maximize a "security posture" metric while remaining within the CPU's available capacity.
Experimental Evaluation
The framework was evaluated using both synthetic workloads and a real-world implementation.
- Synthetic Workloads: The UUniFast algorithm generated thousands of random task sets to test performance across varying CPU utilization levels.
- Real-World Application: The framework was implemented on a Linux-based RTOS (RTAI-patched) using Tripwire, an open-source IDS. The researchers simulated attacks to verify if the optimized security tasks could detect compromises in real-time.
Key Results
- Computational Efficiency: The GP-based optimization solved complex scheduling problems in seconds, compared to hours for exhaustive search methods.
- Near-Optimal Monitoring: In synthetic tests, the algorithm consistently achieved monitoring periods within 20% of the ideal desired frequency, even under high CPU load.
- Effective Attack Detection: In the Tripwire experiment, the adaptive approach detected simulated attacks significantly faster than non-adaptive baselines, demonstrating a tangible improvement in defense.
Conclusion and Future Work
This research demonstrates that retrofitting security into legacy RTS is feasible through opportunistic execution. By treating security as a flexible workload that fills the gaps in a fixed schedule, engineers can enhance system resilience without sacrificing safety. Future work will explore "high-alert" modes for faster response to suspected breaches and handling dependencies between different security tasks.
Summary updated