Small and medium-sized enterprises (SMEs) are increasingly targeted by cyberattacks yet face persistent problems in countering them such as limited security budgets, small Information Technology (IT) teams, and diverse network setups. SMEs rely more on basic security controls that lack advanced and continuous detection measures. Timely intrusion detection is crucial for SMEs, but most existing intrusion detection systems (IDS) are engineered for large organizations, as they require costly hardware, high maintenance, and expertise, conditions that SMEs cannot meet in most cases. Unlike existing hybrid systems, this study introduces a modular hybrid IDS designed for SME operational constraints. The system employs a dual detection layer, network-based and host-based, enhanced with external threat intelligence, machine learning and risk scoring. Its novelty lies in SME-specific custom detection rules, adaptive alerting mechanisms, and lightweight architecture. The system was validated in a real SME across 21 controlled attack scenarios (seven attack types, three repetitions each), with detection accuracy, false positive rate, response time, CPU utilization, and rule maintenance effort measured. The system achieved 95.2% detection accuracy, a 4.1% false positive rate, and an average response time of 2.3 seconds, while maintaining CPU and memory usage below 65% and 77%, respectively. Weekly rule updates require less than 25 minutes. By delivering enterprise-grade detection performance in a cost-efficient design, this work bridges the gap between IDS research and SME realities. The findings establish a foundation for future SME-focused security architectures, including automated rule tuning and expanded threat intelligence integration.

Improvement of the flow-based load balancing model for multipath routing in the telecommunications networks with non-uniform topology was proposed through transition to complex optimality criterion of the obtained solutions. The proposed criterion is of the linear-quadratic nature, which made it possible to ensure a more detailed account of each network link utilization separately and substantiated the choice of routes with the maximum capacity and minimum number of hops during the implementation of the multipath routing strategy. As shown by the comparative analysis, the results of use of the improved model allowed achieving the order of routing and load balancing over the set of the calculated paths, whereby the average packet end-to-end delay was improved for 8-23% in comparison with the most effective single path and multipath solutions.

In this paper the improvement of the flow-based load balancing model for multipath routing in telecommunications networks with nonuniform topology were proposed through transition to complex optimality criterion of the solutions obtained. The proposed criterion is of the linear-quadratic character, which allowed a more detailed account of the each network link utilization separately and substantiated choice of routes with the maximum capacity and minimum number of hops during the implementation of multipath routing strategy. As shown by the comparative analysis, the results of use of an improved model allowed achieving the order of routing and load balancing over the set of paths calculated, whereby the average packet end-to-end delay improved for 8-23% in comparison with the most effective single path and multipath solutions.

The balanced queue management scheme on the interfaces of routers of telecommunication networks has been developed. The novelty of the proposed scheme lies in the application of the two-level method of calculation. At the first level, the problem of allocation and optimal aggregation of packet flows by queues formed on the router interface is solved. At the second level, the problem of allocating and balancing the bandwidth of the interface, taking into account the classification of flows and queues, is solved.

The paper proposes a mathematical model of queue management on routers interfaces of telecommunication networks. The novelty of this model is the approach to a consistent and coordinate solution of such interface tasks as Congestion Management, Resource Allocation and Congestion Avoidance. These tasks were solved in the course of the optimization problem of quadratic programming.