| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| When a broker decodes a compressed record batch, the per-record decoder reads a declared body size from the decompressed stream and allocates a buffer of that size before validating it against the bytes available or any configured limit. The size is attacker-controlled, and the batch CRC does not constrain it, so a small (around 90-byte), CRC-valid compressed produce request can declare a body size of up to about 2 GiB and force the broker to attempt that allocation, raising OutOfMemoryError. The record size limit (max.message.bytes), the request size limit, and client/broker quotas bound only the compressed wire size, not the declared decompressed size.
Impact:
An authenticated principal with Write on a topic can crash broker request-handler threads (reached when a batch's codec differs from the topic's compression.type and records are re-encoded) and any consumer that fetches the record, denying service. On brokers started with -XX:+ExitOnOutOfMemoryError, a common hardening, a single crafted request terminates the broker JVM. More seriously, a crafted record is accepted at ingest and written to disk; when the log-cleaner later reads it on a compacted topic, the OutOfMemoryError (an Error, not an Exception) terminates the cleaner thread, which is not restarted. With the default of a single cleaner thread, log compaction then stops for the entire broker, including the internal __consumer_offsets and __transaction_state topics, and the condition persists across restarts until the offending record is removed. The impact is availability only; no data is disclosed or altered.
Mitigation:
Upgrade to Apache Kafka 4.2.2, 4.3.2, or 4.4.0, and set max.decompressed.message.bytes (at the broker and/or topic level) to a bounded value appropriate for your workload. Upgrading alone is not sufficient: for backward compatibility this configuration defaults to an effectively unlimited value (about 2 GiB), so a broker remains exposed until the limit is set. Choose a value at or slightly above the largest legitimate decompressed record your applications produce, and comfortably below the broker heap divided by the number of request-handler and cleaner threads. Until you can upgrade and configure the limit, restrict Write access to trusted producers; quotas and max.message.bytes do not mitigate this issue.
Resolution:
From Apache Kafka 4.2.2, 4.3.2, and 4.4.0: the broker validates a record's declared decompressed body size against max.decompressed.message.bytes before allocating, on the produce-validation, log-compaction, and timestamp-lookup paths, and rejects oversized or negative sizes with InvalidRecordException. Operators must set this configuration to a bounded value to be protected, as its default is effectively unlimited. |
| Several Kafka deserialization paths read a size or count field from input data and use it directly to allocate an array, collection, or byte buffer, with no check against the bytes actually available. A small crafted input can therefore declare a very large size and force a multi-gigabyte allocation. The affected paths are, in Kafka Streams, ProcessorRecordContext (header count), ChangedDeserializer (varint value length), CombinedKeySchema (foreign-key length, which could also produce a negative size), ProcessorMetadata (key length), and AssignmentInfo group-assignment decoding; and, in the clients library, ListDeserializer (used by ListSerde).
Impact:
An attacker can trigger the oversized allocation and raise an OutOfMemoryError that crashes the affected consumer or Kafka Streams thread; the broker and the rest of the cluster are unaffected. The reachable surface depends on the path: writing a crafted record to a Streams application's internal changelog or repartition topics, or committing crafted offset metadata to its consumer group (Streams internal paths); writing a crafted record to any topic that a consumer reads with ListSerde (ListDeserializer); or, as a member of a Streams consumer group, returning crafted assignment data when elected leader (AssignmentInfo). Because a crafted record remains in its topic, the error can recur on restart. The record size limit (max.message.bytes) does not help, because the allocation is driven by a size field rather than by the input's length.
Mitigation:
Upgrade to Apache Kafka 4.2.2, 4.3.2, or 4.4.0. Where authorization is enabled, restrict write access to each application's internal topics and consumer group to its own principal, and restrict who can produce to topics that are consumed with ListSerde.
Resolution:
From Apache Kafka 4.2.2, 4.3.2, and 4.4.0: these deserialization paths validate the declared size against the remaining input before allocating and reject input that fails the check. |
| Apache Kafka's code-generated protocol readers pre-size an array's collection from the element count declared by the sender. That count is checked only against the bytes remaining in the request buffer, which assumes about one byte per element; because most elements need several bytes, a request can declare far more elements than the buffer holds and still pass. For a keyed array the reader then allocates an ImplicitLinkedHashCollection of capacity (2 * count +1), so a request of a few tens of megabytes can force a heap allocation several times larger before any element is read. The pattern is emitted into every generated reader, so it affects all Kafka API types with an array field.
Impact:
Request parsing happens before any authorization check. On a listener that does not require authentication (for example PLAINTEXT), an unauthenticated attacker can send a few concurrent requests that each declare a very large=
array length and allocate many times their wire size in broker heap, crashing the broker with an OutOfMemoryError. The reporter demonstrated three concurrent ~90 MB ProduceRequest (v3) connections exhausting a 1 GB heap (~8x amplification). On listeners that require authentication, the attacker needs valid credentials but no specific ACL, since parsing precedes authorization..
Mitigation:
Upgrade to Apache Kafka 4.2.2, 4.3.2, or 4.4.0. Until then, lowering socket.request.max.bytes and restricting network access to broker listeners reduces, but does not remove, exposure.
Resolution:
From Apache Kafka 4.2.2, 4.3.2, and 4.4.0: generated readers cap the initial array pre-allocation at MessageUtil.MAX_PREALLOCATED_ARRAY_CAPACITY (1000) and grow it on demand, and reject a declared array length above MessageUtil.MAX_ARRAY_LENGTH (1,000,000). |
| PostgreSQL Anonymizer's range-based dummy functions dont constraint the range size. This vulnerability can be exploited to exhaust the memory of a backend and OOM-kill it (when ovecommit is on), which results in a cluster-wide crash recovery. The problem is resolved in PostgreSQL Anonymizer 3.3.0 and later versions. |
| PostgreSQL Anonymizer's `anon.image_blur()` function doesn't check the declared dimension of the provided image, allowing an attacker to provide a small image with enormous declared dimensions. This vulnerability can be exploited to exhaust the memory of a backend and OOM-kill it (when ovecommit is on), which results in a cluster-wide crash recovery. The problem is resolved in PostgreSQL Anonymizer 3.3.0 and later versions. |
| Pillow is a Python imaging library. From 8.1.0 until 13.0.0, Pillow can allocate an APNG disposal background at the full canvas size before applying the decompression-bomb check. A crafted APNG with a small first frame, a very large declared canvas, and background disposal or first-frame OP_PREVIOUS reaches PngImageFile._seek while Image.open() positions frame zero. The full-canvas allocation occurs inside Image.open() without requiring pixel loading or frame iteration. This can exhaust memory and deny service. This issue is fixed in version 13.0.0. |
| In Progress® Telerik® Document Processing Libraries prior to version 2026.3.1027, uncontrolled recursion, infinite loop, inefficient algorithmic complexity and resource allocation vulnerabilities in RadPdfProcessing and AI document summarization and question answering allow an unauthenticated attacker to terminate or hang the host process via a crafted PDF document or document text submitted for summarization or question answering. |
| Memory allocation with excessive size value in the HSS/LMS signature code (HssPublicKeyParameters, HssSignature) in Legion of the Bouncy Castle Inc. bc-csharp before 2.7.0 allows a remote unauthenticated attacker who can supply both an HSS public key and a signature to cause a denial of service through memory exhaustion via a public key encoding with an excessive level count, because the level count L read when parsing an HSS public key was not checked against the RFC 8554 maximum of 8, and signature parsing then allocated an array of L - 1 entries before reading any further signature data. A single verification can commit up to about 17 GB of memory or fail with an OutOfMemoryException. |
| Memory allocation with excessive size value in the OpenPGP signature and user attribute subpacket parsers (SignatureSubpacketsParser.ReadPacket, UserAttributeSubpacketsParser.ReadPacket) in Legion of the Bouncy Castle Inc. bc-csharp before 2.7.0 allows a remote, unauthenticated attacker who can supply a crafted OpenPGP public key, certificate or signature to cause a denial of service (OutOfMemoryException or memory exhaustion in the parsing process) via a subpacket header using the five-octet length form, because the declared length was used to size the subpacket buffer with no upper bound and without being compared with the size of the enclosing subpacket area or packet, so a few bytes of input could demand an allocation of up to about 2 GB before any subpacket data was read. |
| The asynchronous Smile parser (NonBlockingByteArrayParser) uses a raw-binary length prefix taken from untrusted input to allocate a byte[] before the body is read, with no upper bound and no overflow check. On the raw-binary marker 0xFD, _decodeVInt() computes (value << 6) + (i & 0x3F) without the (value >>> 25) != 0 overflow guard present in the blocking SmileParser._readUnsignedVInt, and _startRawBinary then allocates the declared length immediately.
A Smile feed of roughly ten bytes can therefore force an eager allocation of approximately 2 GB, producing an OutOfMemoryError on reactive or asynchronous Smile endpoints. A VInt crafted to overflow int yields a negative length and a NegativeArraySizeException, a RuntimeException that is not a JacksonException and so escapes the parser contract. Either outcome crashes the request or processing thread.
This is the same class of issue as CVE-2020-28491 in the CBOR parser of this library, which was fixed with a LONGEST_NON_CHUNKED_BINARY guard; the asynchronous Smile parser never received the equivalent guard. No privileges or user interaction beyond the ability to submit data for parsing are required. |
| Memory allocation with excessive size value in the DTLS handshake reassembly (DtlsReliableHandshake, DtlsReassembler) in Legion of the Bouncy Castle Inc. bc-csharp before 2.7.0 allows a remote unauthenticated DTLS peer to cause a denial of service through memory exhaustion via crafted handshake message fragments, because the reassembly buffer for each incoming handshake message was allocated at the 24-bit length declared in the fragment header, without the check against the peer's maximum handshake message size that TLS already applied. A fragment carrying no payload can force an allocation of almost 16 MB, for each of up to 16 pending messages per handshake, before the handshake is authenticated. DTLS servers and DTLS clients are both affected; TLS is not. |
| zip is a zip library for Rust that supports reading and writing simple ZIP files. From 1.3.0 until 9.0.0, ZipArchive::extract in src/read/mod.rs calls Vec::with_capacity for a symbolic link target using the attacker-controlled ZIP64 uncompressed_size before reading entry data. A small archive can request an impossible capacity, causing the infallible allocation path to abort the process before the entry is validated. The failure cannot be contained with catch_unwind or a worker thread, allowing repeated denial of service in archive-processing applications. This issue is fixed in version 9.0.0. |
| Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. Prior to 1.22.0, _unpack_2bit() and _unpack_4bit() resize unpacked arrays according to attacker-controlled TensorProto.dims without verifying that raw_data or int32_data encodes the declared number of elements. Materializing an untrusted UINT2, INT2, UINT4, INT4, or FLOAT4E2M1 tensor through onnx.numpy_helper.to_array() or onnx.reference.ReferenceEvaluator can therefore expand a small packed payload into a disproportionately large allocation and exhaust memory. This issue is fixed in version 1.22.0. |
| zip is a zip library for Rust that supports reading and writing simple ZIP files. From 5.1.0 until 9.0.0, applications built with the legacy-zip feature can reserve output capacity from an attacker-controlled ZIP64 uncompressed size before validating the compressed stream. The ShrinkDecoder path in src/legacy/shrink.rs can panic with a capacity overflow during ZipArchive::extract, and the Reduce and Implode decoders contain the same unbounded preallocation pattern. An unhandled panic can terminate an archive-processing service. This issue is fixed in version 9.0.0. |
| Mattermost versions 11.9.x <= 11.9.1, 11.8.x <= 11.8.5, 11.7.x <= 11.7.10, 11.10.x <= 11.10.1 Mattermost fail to apply allocation limits to the byte-string decode paths in its WebSocket msgpack decoder, which allows an authenticated remote attacker to cause excessive server memory allocation via a crafted binary WebSocket frame declaring a Bin32 payload length of up to 4,294,967,295 bytes.. Mattermost Advisory ID: MMSA-2026-00768 |
| **UNSUPPORTED WHEN ASSIGNED** Unbounded allocation in XML parsing in Apache Vysper; all versions, may lead to an out of memory condition.
This issue affects all versions of Apache Vysper. Apache Vysper is an XMPP (Jabber) server library.
As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative. |
| HTTP::Server::PSGI versions before 1.0056 for Perl allow a single request to crash the server by declaring a large Content-Length.
The request body is read with sysread() calls sized by the remaining Content-Length. Perl grows the read buffer to the requested length before any data arrives, so the server tries to allocate as much memory as the client declares.
A remote attacker can send one request with a large Content-Length and a few body bytes, and perl exits with an out-of-memory error before the application is called. HTTP::Server::PSGI runs as a single process, so the server stays down until restarted. |
| A vulnerability was identified in smg-project smg 1.7.0/1.8.0/1.9.0/1.10.0/1.10.1. Affected is the function jpeg_turbo::decode_jpeg_rgb of the file crates/multimodal/src/jpeg_turbo.rs of the component llm-multimodal. The manipulation leads to uncontrolled memory allocation. The attack can be initiated remotely. The exploit is publicly available and might be used. Upgrading to version 1.11.0 is able to address this issue. The identifier of the patch is 8f51b57e219a914089e345b704c3c8de37a3bb8e. It is advisable to upgrade the affected component. |
| Web::Simple versions before 0.035 for Perl allocate a buffer sized by the Content-Length header for urlencoded body parameters.
Web::Dispatch::ParamParser::get_unpacked_body_from reads an application/x-www-form-urlencoded body using a single read call, with no cap and the return value ignored. Perl's read grows the buffer to the requested length before reading, so the declared Content-Length determines the allocation.
An attacker can trigger this through any route whose dispatch specification has a % body-parameter match, such as sub (POST + %name=), sending a Content-Length far larger than the body actually sent. No authentication is involved. |
| ImageSharp is a 2D graphics library. From 1.0.0-beta0001 until 4.1.2, ICC CLUT parsing calculates allocation sizes from attacker-declared channel and grid dimensions before confirming that the profile contains the declared values. IccDataReader.ReadClutF32 can request a large float array, and earlier public IccProfile.Entries parsing paths can allocate a large jagged representation, from a short truncated profile. In version 4, automatic image conversion reaches the parser when DecoderOptions.ColorProfileHandling is Convert; the default Preserve mode avoids that conversion path. The demonstrated impact is memory pressure and input-validation failure, not unhandled process termination. This issue is fixed in version 4.1.2. |
