This work presents Language Model Security Modules (LMSM), a security framework that adapts the separation behind Linux Security Modules (LSM) to LLM serving and gives advances in interpretability and model-internal analysis a common path to runtime enforcement.
Abstract
Large language models (LLMs) are increasingly deployed with layered defenses, yet malicious prompts can still bypass them. Interpretability methods can expose model-internal signals along the generation path that could inform enforcement, but these signals are not security controls by themselves. Deployments that adapt them for safety typically couple each signal to its own calibration, policy logic, and intervention code, so each new artifact creates integration work instead of strengthening a shared defense. We present Language Model Security Modules (LMSM), a security framework that adapts the separation behind Linux Security Modules (LSM) to LLM serving. In LMSM, a selected security backend exposes calibrated evidence, a versioned policy evaluates active rules over trusted per-request context, and a separate gate authorizes buffered output release. This design separates mediation correctness from policy effectiveness, and it allows backend, rule, or schedule changes without rebuilding request handling or enforcement. Our prototype shows the separation working in practice: with Hugging Face Transformers and continuously batched vLLM, the same substrate hosts artifact-backed sparse autoencoder (SAE) and transcoder deployments and task-fitted dense probes, preserves request-specific decisions under scheduler churn, and selectively enforces and composes multiple rules per request. On Qwen3-4B, LMSM-Checkpoint reduces HarmBench attack success rate from 39.20% to 3.32%, with XSTest false refusals rising from 2.40% to 4.40%, while retaining 98.14% of the throughput of a matched serving path that performs no monitoring work at 32 active sequences. LMSM gives advances in interpretability and model-internal analysis a common path to runtime enforcement.
Large Language Model (LLM)-based agents automate complex tasks by integrating external tools such as web browsers, e-mail clients, file readers, and APIs, but this same integration exposes them to indirect prompt injection (IPI) attacks, in which malicious instructions hidden in tool content hijack the agent. A central but often overlooked question is how defending against such attacks affects the LLM and its own task performance and computational efficiency. In this study, we design a comprehensive testbed and a layered defense, Spotlight-Guard, that combines spotlighting-based input isolation, an LLM detection-and-quarantine pipeline, and instruction integrity based on a Hash-based Message Authentication Code (HMAC) into a single framework, and we evaluate it jointly along two axes: security and LLM performance. Experiments on locally hosted 7B-class open-weight models (Qwen-2.5-7B, Mistral-7B, and DeepSeek-Coder) use Attack Success Rate (ASR) for security and benign-task success rate together with confusion-matrix-based metrics (precision, recall, and F1) for task performance, all with bootstrap 95% confidence intervals. Across a stratified, fixed-seed benchmark of 250 adversarial and 250 benign cases per configuration, the full system reduces the ASR from 36.0% to 17.2% while preserving a 97.2% benign-task success rate and raising the detection F1 from 0.749 to 0.892, demonstrating that strong protection need not degrade the model’s task performance. A component ablation isolates each layer’s contribution, an adaptive-attack evaluation confirms a low ASR (6.7%) under attacks crafted to target the pipeline, and an analysis of computational cost (model invocations per request) quantifies the efficiency overhead, characterizing the security–performance trade-off of layered defenses on open-weight LLMs.
Doygun Demirol, Murat Aydoğan· Applied Sciences· 0 citations
This paper proposes SecureMCP, a policy-enforced framework that integrates Role-Based Access Control with an MCP server to establish multi-layer defense for LLM-generated SQL execution, and evaluates filter performance—false positive rate (FPR) and false negative rate (FNR))—separately from LLM generation quality.
Wonbae Kim, Hee-Kyong Yoo, Nammee Moon· Applied Sciences· 0 citations
Large Language Models (LLMs) are increasingly integrated into software development workflows, yet their ability to autonomously generate secure authentication code remains uncertain. This paper evaluates the security architecture of authentication systems generated by five prominent AI coding assistants through a bi-modal assessment framework combining static code analysis and dynamic penetration testing, mapped to NIST SP 800-63B guidelines. The study examines model behavior across four prompting strategies Basic, Secure, NIST-Based, and Reprompting to reflect varying levels of developer guidance. Empirical results demonstrate that code generated from functional or generically secure prompts consistently omits critical protections, particularly concerning brute-force resistance, session management, and robust password handling. While providing explicit, single-shot NIST context significantly improves compliance, the findings reveal that this remains structurally inadequate. Instead, iterative Reprompting: forcing models into a contextual self-auditing loop is strictly required to achieve a comprehensive, defense-in-depth security architecture. Ultimately, this study proves that current AI coding assistants do not produce secure-by-default applications, dictating that enterprise deployments must transition from single-shot prompt engineering to continuous, standards-driven verification pipelines.
Ishpuneet Singh, Shreyas Mahajan, Gurjot Singh et al.· 0 citations
D-RELLM is presented, a defensive reverse-engineering framework for black-box security assessment of deployed LLM applications that treats the deployed application as a socio-technical system whose risk depends on instruction hierarchy, retrieval trust, authorization, tool agency, output handling, monitoring, and operational controls.
Bhavesh B. Prajapati, Bhavya Shah· International journal of com...· 0 citations
Runtime security in DevSecOps pipelines remains a significant challenge because widely used tools such as Bandit, Hadolint, and Trivy detect vulnerabilities during development but do not enforce security at execution time. This paper evaluates a three-layer defense-in-depth model combining DevSecOps tools, Linux Discretionary Access Control (DAC), and SELinux Mandatory Access Control (MAC). Experiments were conducted on Flask and Django applications running on Red Hat Enterprise Linux 9.6 with SELinux in enforcing mode. Results show that DevSecOps tools reported warnings but provided no runtime protection, DAC failed under permissive configurations such as chmod 777, and SELinux MAC consistently blocked unauthorized file-write attempts, generating 10,324 AVC denials for Flask and 2,065 for Django. SELinux introduced approximately 50% throughput overhead in Flask, representing a measurable but acceptable trade-off for strong runtime enforcement.
Harsh Dhruv· Annual International Compute...· 0 citations
LogSanitizer is proposed, a family of input sanitization defenses operating at two levels: a pre-prompt log-transformation pipeline that disrupts trigger patterns in the structured log representation, and a post-tokenizer perturbation strategy that corrupts trigger-bearing token configurations before they reach the model.
Leszek Wronski, Bogdan Ksiezopolski· Proceedings of the 23rd Inte...· 0 citations