Anomaly paper

I propose a nonlinear dynamical system for structured string anomaly detection that combines logarithmic compression, oscillatory perturbation, permutation mixing, and interaction-based coupling. The system evolves feature representations of strings through iterative updates and exhibits controllable regimes of convergence, stability, and bounded chaos. We demonstrate that this approach improves detection of structurally irregular patterns compared to baseline methods such as Bag-of-Characters and TF-IDF in simulated experiments