OMNI-Vector: Versatile Vectorization for Cross-Task, Multi-Modal AI Pipelines

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Conventional Intrusion Detection Systems (IDS) operate under the premise that advanced threats will eventually exhibit recognizable behavioural signatures. However, this assumption fails in contemporary adversarial landscapes where threats evolve more rapidly than the defences that aim to contain them. To address this, we propose OMNI-Vector—a simulation-oriented framework crafted to investigate the adaptive behaviour of AI-powered, quantum-resilient malware agents under controlled conditions. Central to OMNI-Vector is the SV-Phantom agent, an intelligent adversary capable of learning stealth tactics through vector modulation and reinforcement learning. Unlike traditional static malware, SV-Phantom continuously evolves its system call patterns based on entropy variations, detection delays, and stealth efficiency metrics. The OMNI-Vector framework is built on three foundational components: the Adaptive Quantum Stealth Engine (AQSE) for intelligent behaviour filtering, the Multi-Agent Vector Network (MAVN) for inter-agent coordination, and a reinforcement learning loop based on Proximal Policy Optimization (PPO) to train adaptive evasion techniques. In controlled Qiskit-powered sandbox trials (5–10 runs), SV-Phantom demonstrated superior performance in both evasion and mission success rates when compared to standard PPO-only and DQN-based agents. It is important to note that all experiments were conducted in simulated environments, and the research is strictly intended for academic and ethical study purposes.