Security issues in mission-critical real-time systems (e.g., command and control systems) are becoming increasingly important as there are growing needs for satisfying information assurance in these systems. In such systems, it is important to guarantee real-time deadlines along with the security requirements (e.g., confidentiality, integrity, and availability) of the applications. Traditionally, resource management in real-time systems has focused on meeting deadlines along with satisfying fault-tolerance and/or resource constraints. Such an approach is inadequate to accommodate security requirements into resource management algorithms. Based on the imprecise computation paradigm, a task can have several Quality of Service (QoS) levels, higher QoS result incurs higher computational cost. Similarly, achieving a higher level of confidentially requires stronger encryption, which incurs higher computational cost. Therefore, there exists a tradeoff between schedulability of the tasks on the one hand, and the accuracy (QoS) and security of the results produced on the other hand. This tradeoff must be carefully accounted in the resource management algorithms. In this context, this dissertation makes the following contributions: (i) formulation of scheduling problems accounting both deadline and security requirements of workloads in real-time systems, (ii) development of novel task allocation and scheduling algorithms for such workloads, (iii) and evaluation of the results through simulation studies and a limited test evaluations in one case. In particular, the following are the three key contributions.

Firstly, the problem of scheduling a set of non-preemptable real-time tasks with security and QoS requirements with the goal of maximizing integrated QoS and security of the system is addressed. This problem is formulated as MILP, and then its complexity is proved to be NP-hard. An online efficient heuristic algorithm is developed as the problem is NP-hard. Simulation studies for a wide range of workload scenarios showed that the proposed algorithm outperforms a set of baseline algorithms. Further, the proposed algorithm's performance is close to the optimal solution in a specific special case of the problem.

Secondly, a static assignment and scheduling of a set of dependent real-time tasks, modeled as Directed Acyclic Graph (DAG), with security and QoS requirements in heterogeneous real-time system with the objective of maximizing Total Quality Value (TQV) of the system is studied. This problem is formulated as MINLP. Since this problem is NP-hard, a heuristic algorithm to maximize TQV while satisfying the security constraint of the system is developed. The proposed algorithm was evaluated through extensive simulation studies and compared to a set of baseline algorithms for variations of synthetic workloads. The proposed algorithm outperforms the baseline algorithms in all the simulated conditions for fully-connected and shared bus network topologies.

Finally, the problem of dynamic assignment and scheduling of a set of dependent tasks with QoS and security requirements in heterogeneous distributed system to maximize the system TQV is addressed. Two heuristic algorithms to maximize TQV of the system are proposed because the problem is NP-hard. The proposed algorithms were evaluated by extensive simulation studies and by a test experiment in InfoSpher platform. The proposed algorithms outperform the baseline algorithms in most of the simulated conditions for fully-connected and shared bus network topologies.