92 matches found
memory-poison-lab
Memory Poisoning Lab Your agent remembered it. Should it have? A local test range for agent memory poisoning: persistent memory as an attack surface. Feed an agent's memory controlled untrusted writes, watch a planted memory change what the agent does, then try to stop it - and prove the repair...
memform-code
MemForm evaluation release This repository accompanies the paper MemForm: Benchmarking Memory Formation Poisoning in LLM Agents. MemForm is a deterministic benchmark of persistent-memory poisoning in LLM agents. It asks a causal question that ordinary jailbreak or prompt-injection suites do not...
Accept Changes, Reject Attacks: Robust Memory Management for LLM Agents
Since large language model LLM agents rely on long-term memory, what they store now shapes what they do later. A memory manager decides how each input changes this memory and which records the agent acts on. It can fail in two ways: keeping a fact that has since changed outdated memory, or storin...
GRAD: Graph-Structured Antibody Defense for LLM Agents
Memory poisoning and decomposition attacks distribute malicious evidence across the operations of large language model agents. Existing request-level defenses can miss such attacks because they assess individual inputs without connecting earlier information to later actions. Moreover, detectors...
memory-poisoning-axis
memory-poisoning-axis Deterministic memory-poisoning / prompt-injection measurement — anchored to CVE-2026-24301 "CoSnitch" Microsoft Copilot Personal, CVSS 3.1 8.8, patched 18 Aug 2026. CoSnitch chained: i automatic prompt execution ?q= + undocumented ?autorun=1, ii data exfiltration via...
When Malicious Instructions Persist: Persistent Memory Poisoning Attack on Harness-Based Agents
Harness design has transformed the development of LLM-based agents by integrating memory, tool use, and runtime control. However, this design also introduces security and privacy risks because malicious instructions from external sources may be written into persistent memory and persist across...
MemSentry: A Framework for Detecting Persistent Memory Poisoning in Agentic AI
Agentic AI systems with persistent memory introduce a distinct attack surface known as memory poisoning, in which adversarially crafted content is stored in long-term memory and subsequently influences future agent behavior. Such attacks can suppress security alerts, facilitate privilege...
CAPTURE: Disentangling Preference Drift from Memory Poisoning in Personalized LLM Agents
Personalized language agents use persistent memory to adapt to users over time, but the same mechanism creates an attack surface. When new information conflicts with stored preferences, an agent must distinguish genuine preference drift from temporary context shifts, ambiguity, or adversarial...
AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory
A new class of prompt injection is spreading across commercial websites. It requires no malware, no stolen credentials, and no zero-day exploit. It abuses a standard feature built into almost every major AI assistant: pre-filled deep links. We observed production websites embedding hidden prompt...
Adversarial Attacks in Multi-Agent LLM Pipelines: Unveiling Structural Vulnerabilities in Agentic AI Architectures
Multi-agent LLM pipelines orchestrate multiple specialized language model agents into structured workflows where intermediate outputs are passed across agents to solve complex tasks. This design introduces a security gap absent in single-agent settings: once an agent accepts adversarial content, ...
MemSecBench: Tracking Agent Memory Poisoning from Persistence to Consequence and Repair
Memory systems allow agents to retain and reuse information from past interactions, but they can also let malicious content persist. A malicious instruction crafted by an attacker may be stored in long-term memory, recalled much later, and quietly shape a real action. Recent benchmarks increasing...
CVE-2026-64182
The CVE-2026-64182 vulnerability concerns the Linux kernel memory subsystem: memblk_nr_poison_inc() and memblk_nr_poison_sub() look up a memory block via find_memory_block_by_id(), which returns a reference to the memory block device. Both helpers use the returned memory block without dropping th...
Serena: Unauthenticated Flask dashboard on fixed port enables DNS rebinding → memory poisoning → RCE
SummarySerena's built-in web dashboard exposes an unauthenticated Flask API on a fixed, predictable port TCP 24282, hardcoded as 0x5EDA in constants.py. The server has no authentication, no CSRF protection, and no Host header validation. A DNS rebinding attack allows a malicious webpage to reach...
GHSA-37H2-6P4F-MP3Q Serena: Unauthenticated Flask dashboard on fixed port enables DNS rebinding → memory poisoning → RCE
Summary Serena's built-in web dashboard exposes an unauthenticated Flask API on a fixed, predictable port TCP 24282, hardcoded as 0x5EDA in constants.py. The server has no authentication, no CSRF protection, and no Host header validation. A DNS rebinding attack allows a malicious webpage to reach...
CVE-2026-49471 Serena: Unauthenticated Flask dashboard on fixed port enables DNS rebinding → memory poisoning → RCE
Serena is a powerful MCP toolkit for coding that provides semantic retrieval and editing capabilities. Prior to v1.5.2, Serena's built-in web dashboard exposes an unauthenticated Flask API on a fixed, predictable port, with no authentication, no CSRF protection, and no Host header validation. A D...
CVE-2026-49471
Serena exposes an unauthenticated Flask API on a fixed port prior to v1.5.2, with no authentication, CSRF protection, or Host header validation. A DNS rebinding attack can reach this API from any browser and write arbitrary content to the agent’s persistent memory store, which the agent may read ...
CVE-2026-49471 Serena: Unauthenticated Flask dashboard on fixed port enables DNS rebinding → memory poisoning → RCE
Serena is a powerful MCP toolkit for coding that provides semantic retrieval and editing capabilities. Prior to v1.5.2, Serena's built-in web dashboard exposes an unauthenticated Flask API on a fixed, predictable port, with no authentication, no CSRF protection, and no Host header validation. A D...
Forensic Trajectory Signatures for Agent Memory Poisoning Detection
We discover a behavioral invariant in LLM agents under persistent memory poisoning: in architectures where routing information is retrieved through observable memory-tool invocations, successful attacks require calling memoryrecallfact before emailsendemail, a transition that non-exfiltrating...
MemVenom: Triggered Poisoning of Multimodal Memories in Web Agents
External memory has become a core component of modern web agents, enabling long-horizon reasoning through the retrieval of past experiences. However, this paradigm introduces a critical vulnerability: malicious content injected into memory can be persistently recalled and repeatedly influence age...
Updating the taxonomy of failure modes in agentic AI systems: What a year of red teaming taught us
In this article 1. Why the Taxonomy Needed Updating 2. Seven new failure modes 3. Operational findings: What red teaming showed 4. New mitigations 5. What to do this quarter When the Microsoft AI Red Team published the Taxonomy of Failure Modes in Agentic AI Systems in April 2025, the goal was a...