v299: Proceedings of CAMLIS 2025

2026-06-04 · Source: Proceedings of Machine Learning Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, quick

Summary

The 2025 Conference on Applied Machine Learning for Information Security (PMLR 299), held October 22-24, 2025, at Sands Capital, Arlington VA, presents 14 papers edited by Edward Raff and Ethan M. Rudd. This volume covers a broad spectrum of topics at the intersection of machine learning and cybersecurity. Key research includes adversarial ML attacks on financial reporting, evaluating Visual Language Model (VLM) alignment training using Text2VLM, and a self-sustained attack detection framework for enterprise security. Other contributions address red teaming AI red teaming, developing generalizable cyber defense agents, causal reinforcement learning for cyber anomaly detection, and automatic restoration of poisoned ML examples via PD-AutoR. Additionally, papers explore backdoors in whitebox LLMs (ShadowLogic), AI agents simulating scam calls (ScamAgents), frameworks for rapid LLM attack protection, explainable detection of AWS role chaining attacks (RoleSentry), efficient continual learning for malware analysis (MADAR), and GraphRAG-inspired agentic cloud infrastructure (RIG-RAG).

Key takeaway

For AI Security Engineers and ML practitioners developing secure systems, this conference volume underscores the critical need to integrate advanced adversarial ML defenses and robust detection frameworks. You should prioritize evaluating LLM and VLM vulnerabilities, implementing solutions like ShadowLogic for backdoor detection, and exploring causal reinforcement learning for anomaly detection to proactively counter sophisticated cyber threats.

Key insights

The 2025 AMLIS conference highlights diverse advancements and threats at the intersection of machine learning and information security.

Principles

• ML security requires continuous adaptation.
• Adversarial robustness is multi-faceted.
• Explainability enhances attack detection.

Method

The conference presents methods like Maximum Violated Multi-Objective Attack, Text2VLM for VLM alignment evaluation, PD-AutoR for poisoned example restoration, and MADAR for continual malware analysis.

In practice

• Implement ShadowLogic to detect LLM backdoors.
• Use RoleSentry for AWS role chaining attack detection.
• Apply RIG-RAG for agentic cloud infrastructure.

Topics

• Applied Machine Learning
• Information Security
• Adversarial ML
• Cyber Anomaly Detection
• Large Language Models
• Visual Language Models
• Cloud Security

Code references

• Advai-Ltd/public-advai-Text2VLM-python

Best for: Research Scientist, CTO, VP of Engineering/Data, AI Scientist, AI Security Engineer, Machine Learning Engineer

Related on AIssential

• SECUREVENT: Hybrid AI/ML Security Monitoring for Distributed Event-Based Systems — Hybrid AI/ML security monitoring is crucial for dynamic distributed event-based systems where static controls fail.
• Pushing the Frontier of Black-Box LVLM Attacks via Fine-Grained Detail Targeting — M-Attack-V2 improves black-box LVLM attacks by denoising gradients and refining local alignment for better transferability.
• AI cybersecurity is not proof of work — AI cybersecurity success depends on model intelligence and understanding, not merely "proof of work" computational resources.
• MLDAS: Machine Learning Dynamic Algorithm Selection for Software-Defined Networking Security — MLDAS dynamically selects optimal ML algorithms for SDN security based on real-time network traffic.
• What if your security system could think like a fraudster before they act? — AI systems provide real-time, adaptive fraud detection by learning normal behavior and flagging anomalies.
• While Everyone Watches Glasswing, Attackers Are Walking Through Your Front Door. — Most major cyberattacks exploit common, known vulnerabilities, not zero-days, with AI accelerating their scale.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Proceedings of Machine Learning Research.