CVE-2026-43629

llama.cpp builds b4882 through b9058 contain a heap buffer overflow vulnerability in the KV cache state restore path where the statereaddata function computes write size without overflow checking, allowing attackers with write access to the slotsavepath directory to corrupt heap memory. Attackers...

CVE-2026-43629

The CVE-2026-43629 entry concerns llama.cpp builds b4882–b9058 with a heap buffer overflow in the KV cache state restore path. The root cause is state_read_data() computing the write size without overflow checks, enabling an attacker with write access to the slot_save_path directory to craft mali...

PYSEC-2026-3480 LMCache: 16-bit multimodal hash collision can poison KV cache entries

A flaw has been found in LMCache up to 0.4.6. This affects the function hexhashtoint16 of the file lmcache/integration/vllm/utils.py of the component KV Cache Handler. Executing a manipulation can lead to use of weak hash. The attack needs to be launched locally. The attack requires a high level ...

GHSA-3HH9-752G-5G22 LMCache: 16-bit multimodal hash collision can poison KV cache entries

A flaw has been found in LMCache up to 0.4.6. This affects the function hexhashtoint16 of the file lmcache/integration/vllm/utils.py of the component KV Cache Handler. Executing a manipulation can lead to use of weak hash. The attack needs to be launched locally. The attack requires a high level ...

PT-2026-46251

Name of the Vulnerable Software and Affected Versions LMCache versions prior to 0.4.7 Description A flaw in the KV Cache Handler component affects the hex hash to int16 function within the lmcache/integration/vllm/utils.py file. A local attacker can perform a manipulation that leads to the use of...

kv-cache-side-channel-poc

KV Cache Side-Channel: Cross-Tenant Timing Oracle Proof of co...

CacheTrap: Injecting Trojans in LLMs without Leaving Any Traces in Inputs or Weights

Adversarial weight perturbation has emerged as a concerning threat to LLMs that either use training privileges or system-level access to inject adversarial corruption in model weights. With the emergence of innovative defensive solutions that place system- and algorithm-level checks and correctio...

GHSA-7XCV-9J6C-2FMC Modular Max Serve has Unsafe Deserialization vulnerability

Unsafe Deserialization vulnerability in Modular Max Serve before 25.6, specifically when the "--experimental-enable-kvcache-agent" feature is used allowing attackers to execute arbitrary code...

Whose Narrative Is It Anyway? A KV Cache Manipulation Attack

The Key ValueKV cache is an important component for efficient inference in autoregressive Large Language Models LLMs, but its role as a representation of the model's internal state makes it a potential target for integrity attacks. This paper introduces "History Swapping," a novel block-level...

Shadow in the Cache: Unveiling and Mitigating Privacy Risks of KV-Cache in LLM Inference

The Key-Value KV cache, which stores intermediate attention computations Key and Value pairs to avoid redundant calculations, is a fundamental mechanism for accelerating Large Language Model LLM inference. However, this efficiency optimization introduces significant yet underexplored privacy risk...

Selective KV-Cache Sharing to Mitigate Timing Side-Channels in LLM Inference

Global KV-cache sharing has emerged as a key optimization for accelerating large language model LLM inference. However, it exposes a new class of timing side-channel attacks, enabling adversaries to infer sensitive user inputs via shared cache entries. Existing defenses, such as per-user isolatio...

CachePrune: Neural-Based Attribution Defense against Indirect Prompt Injection Attacks

Large Language Models LLMs are identified as being susceptible to indirect prompt injection attack, where the model undesirably deviates from user-provided instructions by executing tasks injected in the prompt context. This vulnerability stems from LLMs' inability to distinguish between data and...