Research map / Adversarial robustness and security / Attacks on text and graphs

The Web Doesn't Sit Still: Adversarial Self-Evolving Attacks on Search Agents

NeurIPS 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Adversarial prompts for language models

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes adversarial self-evolving attacks that automatically evolve injection strategies against LLM search agents instead of relying on static human-crafted templates.

Problem

LLM-based search agents are vulnerable to adversarial attacks such as malicious information injection and stealthy context manipulation, but existing attacks rely on static, human-crafted templates that fail to adapt as the web changes.

Approach

Attacks that self-evolve over time rather than using fixed templates, adapting injected content to the evolving web and search agent behavior (details not available in the abstract).

Closest papers by meaning

• Adversarial Search Engine Optimization for Large Language Models (ICLR 2025 · similarity 0.87)
• REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective (ICML 2025 · similarity 0.86)
• ReliabilityRAG: Effective and Provably Robust Defense for RAG-based Web-Search (NeurIPS 2025 · similarity 0.86)
• SEAS: Self-Evolving Adversarial Safety Optimization for Large Language Models (AAAI 2025 · similarity 0.86)
• Dissecting Adversarial Robustness of Multimodal LM Agents (ICLR 2025 · similarity 0.85)
• AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents (NeurIPS 2024 · similarity 0.85)