CVE-2026-103634

Published Oct 22, 2026

Last updated 17 days ago

C++

Overview

Description
Out-of-bounds read and write in the Count-Min sketch deserialization of Apache DataSketches C++ (repo: datasketches-cpp). count_min_sketch::deserialize() did not include the preamble in its input size check, so a truncated sketch could cause a read of up to 16 bytes past the end of the input. In addition, the table size was computed from the serialized number of buckets and number of hash functions in 32-bit arithmetic. A crafted sketch could make it wrap to zero, so that the sketch deserialized with an empty table, and later updates and estimate queries read and wrote outside the heap allocation. This can corrupt heap memory, causing a crash and potentially enabling further exploitation. This issue affects Apache DataSketches C++: from 4.1.0 before 5.3.0. Only applications that deserialize Count-Min sketches from untrusted sources are affected. Users are recommended to upgrade to version 5.3.0, which fixes this issue.
Source
security@apache.org
NVD status
Awaiting Analysis

Weaknesses

security@apache.org
CWE-190

Social media

Hype score
Not currently trending

Related CVEs

  1. A flaw was found in the libstdc++ component of GCC. An attacker who can provide an excessively large element count to a boolean vector (std::vector<bool>) constructor can trigger an integer wrap during internal memory calculations. This issue leaves the container in an inconsistent state, leading to out-of-bounds memory access and resulting in an application crash or a Denial of Service (DoS) during subsequent operations.•CVE-2026-114691
  2. A flaw was found in GCC libstdc++. When an application extracts data into a bitset from a wide-character stream, internal stack memory allocation fails to properly account for character size and alignment. An attacker capable of supplying crafted input to an affected application could trigger a stack-based buffer overflow leading denial of service (DoS).•CVE-2026-115455
  3. llama.cpp b8841 before b11865 contains an out-of-bounds write vulnerability in the ggml-rpc RDMA transport rdma_recv() function that allows unauthenticated peers to overflow the rpc-server stack. Attackers advertising RDMA capabilities in the HELLO message can send up to 256 KiB into a 1-byte stack variable in rpc_serve_client(), crashing the server or potentially executing code.•CVE-2026-118375
  4. llama.cpp b5899 before b10261 contains a heap out-of-bounds write vulnerability in llm_tokenizer_plamo2::build that allows attackers to corrupt heap memory via crafted GGUF models. Attackers can supply BYTE-typed token text like <0x-1> whose negative parsed value writes token ids before the bytes_ buffer during model loading, causing crashes or potential code execution.•CVE-2026-116816
  5. A security flaw has been discovered in ggml-org llama.cpp up to 0.7.0. Affected by this vulnerability is an unknown functionality of the file src/llama-sampler.cpp of the component Mirostat Sampler. The manipulation results in reachable assertion. It is possible to launch the attack remotely. The project was informed of the problem early through a pull request but has not reacted yet.•CVE-2026-116467

References

Sources include official advisories and independent security research.