SpectralGuard: Detecting Memory Collapse Attacks in State Space Models
About
State Space Models (SSMs) such as Mamba achieve linear-time sequence processing through input-dependent recurrence, but this mechanism introduces a critical safety vulnerability. We show that the spectral radius rho(A-bar) of the discretized transition operator governs effective memory horizon: when an adversary drives rho toward zero through gradient-based Hidden State Poisoning, memory collapses from millions of tokens to mere dozens, silently destroying reasoning capacity without triggering output-level alarms. We prove an Evasion Existence Theorem showing that for any output-only defense, adversarial inputs exist that simultaneously induce spectral collapse and evade detection, then introduce SpectralGuard, a real-time monitor that tracks spectral stability across all model layers. SpectralGuard achieves F1=0.961 against non-adaptive attackers and retains F1=0.842 under the strongest adaptive setting, with sub-15ms per-token latency. Causal interventions and cross-architecture transfer to hybrid SSM-Attention systems confirm that spectral monitoring provides a principled, deployable safety layer for recurrent foundation models.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Adversarial Attack Detection | Balanced 500-sample non-adaptive | Precision94 | 6 | |
| Adversarial Detection | Wikipedia HiSPA word-shuffle benign adversarial N=500 per model (held-out) | F1 Score65 | 3 | |
| Adversarial Attack Detection | Adaptive Threshold Evasion | Precision94.1 | 2 | |
| Adversarial Attack Detection | Adaptive Multi-layer-aware Split | Precision88.9 | 1 | |
| Adversarial Attack Detection | Cross-architecture Transfer (Split) | F1 Score0.891 | 1 |