Recherche « compression »

rthalley dnspython

dnspython is a DNS toolkit for Python. Prior to 2.9.0, dns.name.from_wire_parser follows attacker-controlled DNS compression pointer chains without a hop limit. A malicious DNS wire-format message can force excessive pointer traversal during name decompression, causing high CPU use, application degradation, or denial of service in applications that process untrusted DNS messages. This issue is fixed in version 2.9.0.

body-parser

body-parser is body parsing middleware for Node.js and Express. In versions before 1.20.9 and in versions from 2.0.0 before 2.4.0, a single request can crash any application using body-parser with default options. When a request carries a Content-Encoding such as gzip, body-parser pipes it into a decompression stream and, after reading the body, leaves that stream with no error listener. A payload of one valid compressed frame followed by trailing bytes decodes and the application responds normally, but the trailing bytes then reach the already-finished decoder, which emits an error that is re-thrown as an uncaught exception and terminates the process. No matching route is required because body parsing runs before routing, so any endpoint that parses a body is affected. The issue is fixed in body-parser 2.4.0 and 1.20.9, and users should upgrade to one of those versions or later. As a workaround, disable request-body inflation (set the body-parser inflate option to false) so encoded requests are rejected before any decompression stream is created.

handnot2 esaml · arekinath esaml · dropbox esaml

Improper Handling of Highly Compressed Data (Data Amplification) in esaml_binding in dropbox esaml allows an unauthenticated attacker to exhaust server memory and crash the Erlang VM via a crafted compressed SAMLResponse. esaml_binding:decode_response/2 base64-decodes the attacker-supplied SAMLResponse, inflates it with zlib:unzip/1 with no bound on the output, then converts the result to a charlist with binary_to_list/1 before parsing. This runs on the SAML consume (ACS) and single-logout endpoints, which accept unauthenticated requests by design, and it runs before any signature check. A small deflate stream reaches a compression ratio near 1000 to 1, so a request body of tens of kilobytes inflates to hundreds of megabytes and then to several gigabytes of list heap, so a few requests exhaust memory and take the node down. This issue affects esaml: from 0.1.0 onward.

python-pillow Pillow

Pillow is a Python imaging library. Prior to 13.0.0 on Windows, src/PIL/WmfImagePlugin.py passes unchecked 32-bit WMF or EMF dimensions to the native drawwmf handler in src/display.c. The handler sizes a DIB allocation from 16-bit-truncated bcWidth and bcHeight fields but copies pixel data using the original full dimensions. A crafted wide, short metafile can pass the pixel-count decompression-bomb check and make the Windows-only renderer read beyond the heap buffer, potentially disclosing memory or crashing the process. This issue is fixed in version 13.0.0.

python-pillow Pillow

Pillow is a Python imaging library. Prior to 13.0.0, SpiderImageFile.seek() in src/PIL/SpiderImagePlugin.py reparses a later SPIDER stack frame and replaces self._size without rerunning Image._decompression_bomb_check(). A tiny file can use a safe first frame to pass Image.open() and then declare very large dimensions in a later mode F frame. Loading that frame allocates approximately four bytes per attacker-declared pixel and can consume multiple gigabytes of memory. This issue is fixed in version 13.0.0.

Apache Software Foundation Apache CouchDB

An unlimited gzip decompression routine in chttpd in Apache CouchDB 3.5.2 and below on all platforms allows any remote client to cause a denial of service via memory exhaustion. Users are recommended to upgrade to version 3.5.3, which fixes this issue.

python-pillow Pillow

Pillow is a Python imaging library. From 4.1.0 until 13.0.0, Pillow's MSP RLE decoder can expand a small file into an unbounded in-memory buffer. The decoder writes run-length output to BytesIO without limiting expansion to the declared image size. An attacker can supply a crafted MSP image to an application that loads its pixels. This can exhaust memory and deny service even when the declared pixel count passes the decompression-bomb check. This issue is fixed in version 13.0.0.

python-pillow Pillow

Pillow is a Python imaging library. From 9.2.0 until 13.0.0, Pillow can retain a rejected GIF frame size after a decompression-bomb exception. In GifImageFile._seek, a crafted GIF can leave rejected dimensions stored after Image.DecompressionBombError; if an application catches that exception and continues processing the same image object, later seek or load operations allocate according to those dimensions and can exhaust memory, denying service. This issue is fixed in version 13.0.0.

python-pillow Pillow

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.

undici · Nodejs undici

undici is an HTTP client library for Node.js. In versions 6.5.0 through 6.29.0, 7.0.0 through 7.30.0, and 8.0.0 through 8.12.1, its EventSource client can be driven to exhaust process memory by a malicious or compromised Server-Sent Events peer. By sending millions of tiny completed data fields with no blank event separator, or a single unfinished line split across many chunks, the parser retains a growing amount of per-field and per-chunk state that the byte-based maxEventSize and maxResponseSize limits do not bound, and because maxResponseSize is enforced on the compressed body before decompression, a small compressed response can expand into enough retained memory to terminate the process. An application only needs to maintain an EventSource connection to an untrusted peer, which controls only the response, and availability is the only impact. The issue is fixed in undici 6.29.1, 7.30.1, and 8.12.2, which bound the retained parser state independently of the byte limits, and users should upgrade to one of those versions or later. There is no workaround.

sgl-project sglang

SGLang through 0.5.21 contains a denial of service vulnerability that disables Pillow decompression bomb protection process-wide by setting Image.MAX_IMAGE_PIXELS to None in NanoNemotronVLImageProcessor. Remote attackers can submit small, highly compressed images with huge dimensions to the generic image loader, exhausting worker memory during decoding.

sgl-project sglang

SGLang through 0.5.21 contains a decompression amplification vulnerability in load_audio that allows remote attackers to exhaust memory by submitting highly compressed audio decoded without duration or sample limits. Attackers can supply small FLAC-encoded silence files through audio_url or audio_data inputs that expand into massive PCM buffers, exhausting worker memory and CPU for denial of service.

Apache Software Foundation Apache CXF · Apache cxf

CompressionUtils.inflate() decompressed attacker-controlled DEFLATE data with no output-size cap. A small (~KB) crafted payload could expand to gigabytes on the heap. Reachable via JWE decryption when zip=DEF (e.g. JoseSessionTokenProvider with RSA-OAEP key wrap) and via SAML redirect/POST binding token inflation — in both cases decompression happens before/independent of trust validation. Fix: Added a configurable maximum inflated-size cap (default 10 MiB, org.apache.cxf.compression-max-inflated-size system property) to CompressionUtils.inflate(); aborts with DataFormatException once exceeded. Users are recommended to upgrade to versions 4.2.4 or 4.1.9 or 3.6.13, which fix this issue.

Progressive Robot Ltd hMailServer

Inefficient algorithmic complexity in the JSON reader of Progressive Robot hMailServer allows a remote unauthenticated attacker to make the mail services unavailable. Reading a JSON object kept the first of each duplicated member name by searching the members already read, so an object of N distinct member names cost O(N^2): 1 MB took 8.9 seconds and 4 MB took 300 seconds against the affected code. A remote attacker reaches the reader without an account by mailing a crafted TLS-RPT report (up to 16 MB after decompression) to a hosted domain's published report mailbox, which is read on a delivery thread; a few such reports hold every delivery thread (ten by default), so the server delivers no mail, local or outbound, for over an hour per report. A signed-in account reaches the same 16 MB body through the webmail's own REST routes, holding the REST API's own worker threads. Where no report mailbox is configured, the unauthenticated REST sign-in routes that read a JSON body are capped at 64 KB and are not affected.

AsyncHttpClient async-http-client

The AsyncHttpClient (AHC) library allows Java applications to easily execute HTTP requests and asynchronously process HTTP responses. From 2.2.0 until 3.0.14, WebSocket permessage-deflate decompression is unbounded when compression is enabled. The inbound pipeline aggregates compressed frames before WebSocketClientCompressionHandler inflates them, so webSocketMaxFrameSize and webSocketMaxBufferSize do not bound decompressed output. A malicious WebSocket peer can send a small compressed message that expands to a very large Netty buffer and exhausts JVM heap. This issue is fixed in version 3.0.14.

yawkat lz4-java

yawkat LZ4 Java provides LZ4 compression for Java. From 1.7.0 until 1.11.4, net.jpountz.util.Native.load() uses File.createTempFile to create an exclusive temporary .lck file but derives the native-library path by removing the suffix, then FileOutputStream opens that predictable path without exclusive creation, allowing another local user with access to the same shared temporary directory to create or replace the library file before System.load() uses it. Successful exploitation depends on shared-directory permissions, host protections, and winning the race, and can execute native code as the victim; hardened systems may instead cause library loading to fail and fall back to Java implementations. Configurations using a system library, a private java.io.tmpdir, or Java-only implementations are not affected. This issue is fixed in version 1.11.4.

SixLabors ImageSharp

ImageSharp is a 2D graphics library. From 3.0.0 until 4.1.1, decoding a strip TIFF using CCITT Group 3 or Modified Huffman compression can pass attacker-expanded runs to BitWriterUtils.WriteBits without first checking the current row width. T4TiffCompression.WritePixelRun can accumulate oversized makeup-code runs, and ModifiedHuffmanTiffCompression.Decompress validates the width only after writing. The unchecked writes can overflow the strip buffer, corrupt heap memory, and terminate the process. This strip-path vulnerability is distinct from the tiled decompressor-width mismatch. This issue is fixed in version 4.1.1.

SixLabors ImageSharp

ImageSharp is a 2D graphics library. From 2.1.0 until 4.1.2, the TIFF CCITT Group 4 encoder allocates Width times rowsPerStrip bytes even though T6BitCompressor.CompressStrip can emit encoded row data and two 12-bit end-of-facsimile-block codes beyond that capacity. TiffCcittCompressor.WriteCode performs unchecked writes, and a decode-and-re-encode flow can inherit TiffCompression.CcittGroup4Fax and one-bit metadata from attacker-supplied input. The resulting out-of-bounds writes can corrupt memory and terminate the process. This issue is fixed in version 4.1.2.

SixLabors ImageSharp

ImageSharp is a 2D graphics library. From 2.0.0 until 4.1.2, the TIFF CCITT Group 3 encoder allocates an undersized compressed-data buffer for narrow 1-bit images. TiffCcittCompressor.Initialize does not reserve enough space for the row data and T4 end-of-line codes, and T4BitCompressor.CompressStrip reaches unchecked writes when TiffCompression.CcittGroup3Fax is selected directly or inherited from decoded TIFF metadata. An attacker-controlled encode or decode-and-re-encode flow can write beyond the logical output span, corrupt process memory, and terminate the process. This issue is fixed in version 4.1.2.

tornadoweb tornado

Tornado versions before 6.5.9 contain an unbounded memory accumulation vulnerability in CurlAsyncHTTPClient that allows remote attackers to cause denial of service by sending a compressed response. Attackers can send a gzip-encoded decompression bomb that accumulates in memory without size limits, causing the application process to be killed by out-of-memory conditions.

py-pdf pypdf · Pypdf project pypdf

pypdf is a free and open-source pure-python PDF library. Prior to 6.18.1, a crafted PDF containing a partially malformed /FlateDecode stream with padded data can force pypdf/filters.py to use inefficient byte-by-byte decompression while the earlier recovery counter fails to advance for bytes that successfully decode, causing long runtimes and application unavailability. This is a residual issue after the malformed FlateDecode recovery fix. This issue is fixed in version 6.18.1.

TeamViewer Full Client · TeamViewer Host

TeamViewer Full Client and Host for Linux and macOS prior version 15.82 contain a heap-based buffer overflow vulnerability in the processing of .tvs session recording files. A size mismatch during decompression of recorded session data can result in out-of-bounds heap writes. By convincing a user to open a specially crafted session recording through the "Play or convert recorded session…" feature, an attacker may achieve arbitrary code execution with the privileges of the current user

nicotine-plus

Nicotine+ is a graphical client for the Soulseek peer-to-peer network. Prior to version 3.3.11, a modified remote client can send zlib-compressed peer messages containing a decompression bomb, exhausting available memory of the recipient's operating system. This issue has been patched in version 3.3.11.

klauspost compress

compress provides various compression algorithms. Prior to version 1.18.7, a signed integer overflow vulnerability in s2.NewDict() allows an attacker to bypass repeat index validation by supplying a dictionary with a uvarint-encoded repeat value exceeding MaxInt64. When Dict.Encode() is subsequently called, the overflowed negative repeat value causes an out-of-bounds memory access via unsafe.Pointer arithmetic, crashing the process with SIGSEGV. This issue has been patched in version 1.18.7.

rabbitmq-server · Broadcom rabbitmq server

RabbitMQ is a messaging and streaming broker. Prior to versions 3.13.15, 4.0.20, 4.1.11, 4.2.6, and 4.3.0, The cowboy WebSocket options at line 117 set compress => true, enabling RFC 7692 permessage-deflate negotiation. The handler does not set max_frame_size, so cowboy's default of infinity applies. cowlib's cow_ws:parse_payload/9 calls zlib:inflate/2 on the compressed payload with no output-size limit. An attacker can negotiate permessage-deflate during the WebSocket upgrade and send a frame containing a zlib bomb (e.g. 50 KB → 5 GB). Decompression occurs in the connection process before websocket_handle/2 ever sees the MQTT bytes. An unauthenticated attacker can crash a RabbitMQ node running the Web-MQTT plugin by sending a single highly-compressed WebSocket frame (a few KB on the wire) that inflates to gigabytes in memory. The cowboy WebSocket handler decompresses the entire frame before the MQTT CONNECT packet is processed, so no credentials are required. Preconditions include rabbitmq_web_mqtt plugin enabled (not default, but common for browser clients) Network reachability to port 15675/15676 No authentication required. This issue is fixed in versions 3.13.15, 4.0.20, 4.1.11, 4.2.6, and 4.3.0.

netdata

Netdata is an open source observability tool. Prior to 2.11.0, Netdata's unauthenticated WebSocket server negotiates permessage-deflate before authentication, and src/web/websocket/websocket-compression.c allows websocket_client_decompress_message() to grow decompressed output toward WS_MAX_DECOMPRESSED_SIZE without enforcing a compressed-to-decompressed ratio. Small highly compressed frames can therefore cause large server-side allocations, and repeated concurrent connections can exhaust memory and terminate monitoring. This vulnerability is fixed in 2.11.0.

vectordotdev vector

Vector is a high-performance observability data pipeline. From 0.15.0 until 0.57.0, the logstash source feeds each decompressed frame back into its decoder without limiting nested compression depth. An unauthenticated remote peer that can reach the default 0.0.0.0:5044 listener can send many nested compressed frames, causing recursive decoding that exhausts the worker thread stack and aborts the process. The same nested construction amplifies decompressed input, and process termination can halt log ingestion for every tenant on a shared pipeline. This issue is fixed in version 0.57.0.

kap-sh zapros

Zapros, a Python HTTP client, prior to version 0.14.0 is vulnerable to denial of service via memory exhaustion. The issue affects all callers who streamed compressed responses relying on the chunk size — explicit (`iter_bytes(chunk_size=...)`) or the default — to bound memory. The decoder ignored that bound, so a chunk could be far larger than requested and a single compressed response could overflow memory. Version 0.14.0 contains a patch. Some workarounds are available. Read the still-compressed body with `Response.iter_raw()` / `Response.async_iter_raw()`, which bypass the built-in decoders, and decompress it yourself with an explicit output-size bound (e.g. `zlib`'s `max_length`), aborting once a configured limit is exceeded. Where feasible, send `Accept-Encoding: identity` to disable response compression so bodies are not decompressed client-side. Avoid decoding response bodies from untrusted servers.

Apache Software Foundation Apache MINA

The CompressionFilter class uses ZLib to deflate and inflate data sent and received. When we inflate incoming data, the filter does not control the resulting size, and create a buffer no matter what. Some compressed data may have a compression ration greater than 1 thousand, leading to an exhaustion of the application memory, as we don't control the deflated size. The fix adds such a control by allowing the application developer to provide a fixed size limit, which when reached throws an exception. It also allows the user to provide a compression ratio that should not be exceeded, protected the application from small inflated files that inflate in gigantic files, but with a grace limit for the resulting size (1Mb) to avoid false positive (like a very small file inflating with a high ratio, but resulting with a acceptable size, like a few thousands bytes) For application using this feature, it is highly recommended to create the CompressionFilter and to pass the maximum limit as a forth constructor parameter, maxDecompressedSize: public CompressionFilter(final boolean compressInbound, final boolean compressOutbound, final int compressionLevel, final int maxDecompressedSize)Optionally one can also provide a maxDecompressRatio fifth parameter, and a decompressRatioMinSize sixth parameter to allow small inflated files with a high compression ratio to still be accepted. Here are the additional constructor: public CompressionFilter(final boolean compressInbound, final boolean compressOutbound, final int compressionLevel, final int maxDecompressedSize, final long maxDecompressRatio, final long decompressRatioMinSize) Also note that a fluent API has been added to spare the users the pain to call a constructor with that many parameters:  CompressionFilter compressionFilter = new CompressionFilter()     .setCompressionLevel(Zlib.COMPRESSION_MAX)   .setMaxDecompressedSize(1_000_000)   .setMaxDecompressRatio(100).   .setDecompressRatioMinSize(100_000);  Applications using Apache MINA are advised to upgrade and configure their CompressionFilter instance.

OISF suricata

Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. From 8.0.0 until 8.0.6, the HTTP/1 parser limits decompression work per transaction but does not limit how many small brotli compression bombs a single flow can submit. With response-body-decompress-layer-limit enabled, repeated compressed responses make the decompression paths in rust/htp perform expensive work for every transaction, degrading packet processing and potentially causing loss of monitoring visibility or denial of service. This issue is fixed in version 8.0.6.

c-ares

c-ares is an asynchronous resolver library. Prior to 1.34.7, ares_dns_name_parse() enforces backward DNS compression pointers but does not bound the total pointer hops or assembled name length. A malicious DNS server can send a response containing a long descending pointer chain and many resource records whose NAME or RDATA fields refer to the chain, causing repeated decompression work that grows quadratically with message size. A single crafted response can stall the single-threaded c-ares event loop and deny DNS resolution, without causing memory corruption or information disclosure. This issue is fixed in version 1.34.7.

cthackers adm-zip

adm-zip is a JavaScript library for creating and extracting ZIP archives in Node.js. Prior to 0.6.1, getData() in zipEntry.js trusts an entry's central-directory uncompressed size and allocates output memory before validating that value against the actual compressed data and decompression result. A small crafted ZIP can declare a multi-gigabyte uncompressed size, causing Buffer.alloc and decompression handling to commit excessive resident memory before CRC validation reports an error. Applications that read entries from untrusted archives can therefore be terminated by the operating system or suffer service-wide memory exhaustion. This issue is fixed in version 0.6.1.

pomerium

Pomerium is an identity and context-aware access proxy. Prior to 0.32.8, decodeQueryStringV2 in pkg/hpke/url.go performs zstd decompression of attacker-controlled data without an output-memory limit when DecryptURLValues processes HPKE V2 values for Stateless.Callback in internal/authenticateflow/stateless.go. In hosted or stateless authentication deployments, an unauthenticated attacker can obtain the receiver key from /.well-known/pomerium/hpke-public-key, provide a matching attacker-controlled sender key, and send a compressed payload to /.pomerium/callback that expands before validateSenderPublicKey rejects the sender. This can allocate hundreds of megabytes per request, exhaust proxy memory, crash or degrade the process, and block access to applications protected by the deployment. Stateful deployments are not affected because the stateful callback verifies its HMAC signature before decryption and decompression. This issue is fixed in version 0.32.8.

AsyncHttpClient async-http-client

The AsyncHttpClient (AHC) library allows Java applications to easily execute HTTP requests and asynchronously process HTTP responses. From 2.0.0 until 2.16.1 and 3.0.12, automatic response decompression on the HTTP/1.1 path uses ChannelManager.newHttpContentDecompressor() to install Http1ContentDecompressor without a cumulative output-size limit. A hostile or compromised server, or an attacker who can alter a response in transit, can send a small gzip, deflate, or snappy response that expands across chunks until the client exhausts its heap and raises OutOfMemoryError; brotli and zstd are also affected when their optional codecs are present. In versions 3.0.8 through 3.0.10, the HTTP/2 decompressor is also unbounded, so switching protocols does not mitigate the issue on those releases. A limit applied to each decode call is insufficient because the response can be delivered as many small chunks, so the fixed implementation tracks total decompressed bytes for the whole response. This issue is fixed in versions 2.16.1 and 3.0.12.

Apache Software Foundation Apache NiFi · Apache nifi

Apache NiFi 2.11.0 disabled support for gzip-encoded HTTP requests for the application REST API and rejected requests that included the standard Content-Encoding header indicating gzip encoding. The framework enforcement filter did not check multiple instances of the Content-Encoding header and did not reject non-standard identifiers for gzip encoding, allowing a malicious client to send crafted requests that could consume excessive amounts of memory. Upgrading to Apache NiFi 2.12.0 is the recommended mitigation, which disables decompression of gzip-encoded HTTP requests regardless of header number or encoding identifiers.

coredns

CoreDNS is a DNS server written in Go. Prior to 1.14.7, the DNS-over-HTTPS, DNS-over-HTTP/3, DNS-over-QUIC, and DNS-over-gRPC request paths in plugin/pkg/doh/doh.go, core/dnsserver/server_quic.go, and core/dnsserver/server_grpc.go call dns.Msg.Unpack on attacker-controlled DNS section counts before dns.DefaultMsgAcceptFunc validates the fixed header. An unauthenticated client can use DNS name compression and excessive section counts to amplify allocation before the plugin chain, so plugin-level rate limiting cannot prevent concurrent requests from exhausting memory and terminating CoreDNS. The ordinary UDP and TCP listeners are not affected because they validate the header first. This issue is fixed in version 1.14.7.

Red Hat build of Keycloak 26.4 · Red Hat build of Keycloak 26.4.16 · Red Hat build of Keycloak 26.6 · Red Hat build of Keycloak 26.6.7

A flaw was found in the SAML Redirect Binding implementation of Keycloak, an open-source identity and access management solution. The issue occurs because the custom DEFLATE compression and decompression helpers fail to release native zlib memory after use. An unauthenticated attacker can exploit this by sending repeated malformed SAML requests, leading to native memory exhaustion and a denial of service.

Linux

In the Linux kernel, the following vulnerability has been resolved: drm/xe: Don't hand out the flat CCS storage as usable VRAM get_flat_ccs_offset() reads the base of the flat CCS storage from the hardware, scales it by the number of enabled L3 nodes, and rounds the result up to 128K. Everything below that offset is then handed to the VRAM allocator as usable memory. Rounding a limit that means "usable memory ends here" upwards publishes whatever lies between the real base and the rounded one as free memory, and that memory belongs to the compression hardware. The scaled value has no reason to be 128K aligned, and on a Battlemage G21 with 16 GiB it is not: flat CCS base: raw 0x3fafff800, rounded 0x3fb000000 so the last 2 KiB of page 0x3fafff000 is CCS storage, in the allocator's pool. Whatever is allocated there gets that tail overwritten by the compression hardware, which needs no page-table entry, no buffer object and no GPU submission to do it, and does it before userspace exists. On this machine a Mesa VM's level-3 page table landed on that page on every cold boot. It lost the entry covering the compositor's batch-buffer heap, so the compositor's first submission faulted fetching its batch and gdm restarted it forever: a black screen on an otherwise working machine. Restarting gdm cleared it because the next VM's page tables were allocated somewhere else. Round down instead, to the page size the allocator works in. On this machine that excludes exactly one page. Reading the reserved page afterwards shows what had been writing it: [369] 0xcccc000000000000 [371] 0xcc77000000000000 [373] 0xcccc000000000000 [375] 0xcc77000000000000 compression metadata, two bytes per sixteen, sitting where the driver used to hand out memory. The assertion that should have caught this compares the offset against GSMBASE - ccs_size for equality. That value is 128K aligned, so it agrees with the rounded-up offset precisely when the base is not aligned - the check cannot fail in the case it exists to catch, and is compiled out unless CONFIG_DRM_XE_DEBUG is set. Replace it with one that can fail: CCS storage must not run into GSM. [ And this was a debug session from hell, enormously helped by an AI doing much of the grunt-work. I'd like to call it my tireless helper, but the AI several times stated flat out that this was impossible and unsolvable and that we should just write a report about it. I suspect those things have been trained by people who may not be quite as stubborn as I am. But while the AI was ready to give up several times, it did keep adding debug code and analyzing it faithfully when I pushed. So credit where credit is due and I let the AI write the commit message above. This is basically a one-liner fixing a bogus "round_up()" to a "round_down()", but there were 24 patches adding more and more debug information to this, and 18 kernel boot to finally narrow it down to this. - Linus ]

cthackers adm-zip

adm-zip versions 0.5.14 through 0.6.0 fail to apply zlib decompression output limits when ZIP entries declare zero uncompressed size. Attackers can craft malicious ZIP archives with highly compressible entries declaring zero size to exhaust memory and cause denial of service.

http4k

http4k is a functional toolkit for Kotlin HTTP applications. Prior to 4.51.0.0, 5.42.0.0, and 6.49.0.0, ServerFilters.GZip, RequestFilters.GunZip, and the underlying Gzip request-body decompression functions impose no limit on decompressed size. An unauthenticated client can send a small gzip-encoded request body that expands to gigabytes, exhausting the JVM heap and denying service to other clients. The fix uses SizeLimitedInputStream to enforce a default 10 MiB limit, causes ServerFilters.GZip and RequestFilters.GunZip to return 413 Request Entity Too Large, and causes other decompression paths to throw SizeLimitExceededException. This issue is fixed in versions 4.51.0.0, 5.42.0.0, and 6.49.0.0.

compression

compression is a Node.js and Express compression middleware. In versions before 1.8.2, when a client aborts the connection while a compressed response is still being sent, the zlib stream created to compress that response is never destroyed, so each aborted compressed response leaks its native zlib memory. A remote unauthenticated attacker can repeatedly open requests and disconnect early, exhausting the available memory and crashing the server. All applications using compression are affected. The issue is fixed in compression 1.8.2, and users should upgrade to 1.8.2 or later.

OISF suricata

Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. Prior to versions 7.0.16 and 8.0.5, Suricata's HTTP/2 decompression path could grow the decompressed response-body buffer without an effective upper bound. A crafted HTTP/2 DATA payload using a high compression ratio, such as gzip, deflate, or brotli compressed data, could cause Suricata to allocate excessive memory while decompressing the payload. Versions 7.0.16 and 8.0.5 contain a fix. As a workaround, disable HTTP2.

luben zstd-jni

zstd-jni versions before 1.5.7-14 fail to validate offset and length parameters in the ZstdDictCompress constructor, allowing out-of-bounds memory reads. Attackers can supply untrusted offset or length values to read native heap memory into the compression dictionary, typically causing JVM crashes.

Amazon ion-java

Improper handling of highly compressed data in Amazon ion-java before 1.12.1 might allow remote attackers to cause a denial of service via a crafted compressed Ion document that expands to an arbitrarily large size upon decompression due to insufficient coverage of the GZIP auto-decompression opt-out introduced for CVE-2026-75936. To remediate this issue, users should upgrade to version 1.12.1.

WPMU DEV Smush Image Compression and Optimization

Unauthenticated Denial of Service Attack in Smush Image Compression and Optimization <= 4.2.0 versions.

Linux

In the Linux kernel, the following vulnerability has been resolved: btrfs: lzo: reject compressed segment that overflows the compressed input lzo_decompress_bio() validates each on-disk segment length seg_len only against the workspace cbuf size, not against the compressed input size (compressed_len, the total folio bytes of the bio). A crafted extent can carry a segment whose seg_len passes the cbuf check but runs past the end of the bio, so copy_compressed_segment() walks off the last folio: get_current_folio() then returns the NULL folio from bio_next_folio(), and with CONFIG_BTRFS_ASSERT disabled (default) folio_size(NULL) faults. BUG: KASAN: null-ptr-deref in lzo_decompress_bio (fs/btrfs/lzo.c:383) Read of size 8 at addr 0000000000000000 by task kworker/u8:1/29 Workqueue: btrfs-endio simple_end_io_work kasan_report (mm/kasan/report.c:590) lzo_decompress_bio (fs/btrfs/lzo.c:383) end_bbio_compressed_read (fs/btrfs/compression.c:1065) btrfs_bio_end_io (fs/btrfs/bio.c:135) btrfs_check_read_bio (fs/btrfs/bio.c:180 fs/btrfs/bio.c:285) simple_end_io_work process_one_work worker_thread Reject any segment whose payload would extend beyond compressed_len before copying it, treating it as corruption like the other on-disk validation failures in this function.

Linux

In the Linux kernel, the following vulnerability has been resolved: ntfs3: fix out-of-bounds read in decompress_lznt decompress_lznt() does not validate array index bounds before accessing the decompression table. A corrupted NTFS3 image with invalid compressed data can trigger an out-of-bounds read. Add index bounds checking to prevent the OOB access.

exceljs

exceljs through 4.4.0 decompresses all entries from supplied xlsx archives into memory without limits on entry size, total size, or compression ratio. Attackers can upload highly compressed workbooks that expand to gigabytes in memory, exhausting available resources and causing denial of service.

strukturag libheif

libheif is a HEIF and AVIF file format decoder and encoder. From 1.19.0 until 1.23.1, a crafted uncompressed HEIF image using generic zlib unci full-item compression can crash an application that decodes an advertised tile with heif_image_handle_decode_image_tile(). In libheif/codecs/uncompressed/unc_decoder.cc, unc_decoder::fetch_tile_data() computes a large tile offset and unc_decoder::get_compressed_image_data_uncompressed() validates it with range_start_offset plus range_size. For the last advertised tile (4095, 4095), the addition can wrap to zero, bypass the bounds check, and pass an invalid source pointer and a one-terabyte length to memcpy. The observed result is an out-of-bounds read and process crash; opening the file alone does not trigger the issue because tile decoding is required. This issue is fixed in version 1.23.1.

Amazon Ion Java

Improper handling of highly compressed data in the GZIP auto-decompression handler in Amazon ion-java before 1.12.0 might allow remote actors to cause a denial of service via a crafted compressed Ion document that expands to an arbitrarily large size upon decompression. To remediate this issue, users should upgrade to version 1.12.0 and configure withGzipDecompressionEnabled(false) and/or set an explicit withMaximumBufferSize() when parsing untrusted input.

pandora-analysis pandora

Pandora contains a denial-of-service vulnerability in its handling of DAA (Direct Access Archive) files. When extracting the internal ISO image from a DAA archive, compressed chunks were decompressed using zlib.decompress() without enforcing a limit on the resulting uncompressed data. An attacker able to submit a crafted DAA file containing highly compressed data could cause Pandora to decompress a relatively small input into a very large amount of data in memory. Because the decompressed chunks are accumulated to construct the internal ISO image, this could result in excessive memory consumption and potentially CPU exhaustion, causing the extraction worker to become unresponsive, terminate, or affect the availability of the Pandora service. The patch introduces bounded decompression using decompressobj().decompress() with max_extracted_filesize, verifies the cumulative size of decompressed chunks, and raises a dedicated ZipBomb exception when the configured limit is exceeded. Pandora then aborts extraction and reports the file as too large.

sigstore rekor

Rekor is a software supply chain transparency log. Starting in version 0.3.0 and prior to version 1.5.2, the `Package.Unmarshal()` function in `pkg/types/alpine/apk.go` decompresses the signature and control gzip members of an APK file into in-memory buffers without bounding the total decompressed size. The existing `max_apk_metadata_size` check (default 1MB) is only applied to individual tar entry header sizes after decompression completes, so it does not prevent a decompression bomb from consuming unbounded heap memory. An attacker can craft a gzip stream that compresses at a ~1000:1 ratio (e.g., 2MB compressed zeros → 2GB decompressed). When submitted as spec.package.content in an Alpine `ProposedEntry`, the server decompresses the full payload into memory during request processing, triggering a fatal Go runtime out-of-memory error or OS OOM-kill that cannot be caught by the server's recover() middleware. This is reachable via two unauthenticated endpoints, `POST /api/v1/log/entries (createLogEntry)` and `POST /api/v1/log/entries/retrieve (searchLogQuery)`. Both invoke `V001Entry.Canonicalize()` → `fetchExternalEntities()` → `apk.Unmarshal(packageData)`, which performs the unbounded decompression. Version 1.5.2 patches the issue. There is no effective workaround. Setting `max_request_body_size` reduces but does not eliminate exposure due to the ~1000:1 compression ratio (a 1MB body limit still allows ~1GB heap allocation). Setting `max_apk_metadata_size` has no effect on this vulnerability since the check is applied after decompression.

Fujitsu Research OneCompression

Fujitsu Research's OneCompression library before 1.2.1 contains an unsafe deserialization vulnerability that allows attackers to execute arbitrary code by supplying a crafted model.pt checkpoint file, as QuantizedModelLoader.load_quantized_model_pt() unconditionally calls torch.load with weights_only=False, invoking Python's pickle machinery during deserialization. Attackers can embed malicious __reduce__ methods in a crafted model checkpoint to execute arbitrary Python code, including system commands, when the library loads the file from a caller-selected model directory.

Linux

In the Linux kernel, the following vulnerability has been resolved: btrfs: don't propagate EXTENT_FLAG_LOGGING to split extent maps When btrfs_drop_extent_map_range() splits an extent map, the new split maps inherit the original map's flags through a local 'flags' variable. Commit f86f7a75e2fb ("btrfs: use the flags of an extent map to identify the compression type") changed the EXTENT_FLAG_LOGGING clearing to operate on em->flags instead of that local 'flags' copy, so a split of an extent map that is currently being logged wrongly inherits EXTENT_FLAG_LOGGING. The flag is then never cleared on the split, and when it is freed while still on the inode's modified_extents list (for example by the extent map shrinker) it trips the WARN_ON(!list_empty(&em->list)) in btrfs_free_extent_map() and leads to a use-after-free. Clear EXTENT_FLAG_LOGGING from the local 'flags' copy used for the splits and only clear EXTENT_FLAG_PINNED from em->flags, restoring the behaviour prior to f86f7a75e2fb.

timescaledb · Timescale timescaledb

TimescaleDB through 2.29.1, fixed in commit 517c13e, contains an out-of-bounds read vulnerability that allows authenticated attackers to cause query-result integrity failures or backend crashes by supplying a crafted Simple8b selector-11 value, which is stored in the signed int16 Arrow dictionary-index type and bypasses index validation checks in bulk text dictionary decompression. Attackers with direct DML access to a non-frozen physical compressed hypertable relation can trigger an out-of-bounds read before the base of the live offsets array through the VectorAgg single-text hashing strategy, resulting in incorrect aggregation output, backend SIGSEGV, or PostgreSQL crash recovery depending on build configuration.

timescaledb · Timescale timescaledb

TimescaleDB through 2.29.1, fixed in commit 517c13e, contains an out-of-bounds read in the Dictionary compression reverse row iterator (tsl/src/compression/algorithms/dictionary.c). The forward path validates the decoded index; the reverse path uses an assertion compiled out of release builds, leaving the 64-bit Simple8b index unvalidated and the read offset attacker-controlled. Attackers with DML access to a physical compressed relation can store a crafted datum and run a reverse-order scan. With a pass-by-value column type the out-of-bounds Datum is returned to the client as a normal column value, disclosing backend memory including the shared buffer pool, which SQL access control does not cover.

Apache Software Foundation Apache NiFi · Apache nifi

Apache NiFi 1.5.0 through 2.10.0 support gzip-encoded HTTP requests for the application REST API using a Jersey encoding filter. The framework enforced a configurable maximum request size on the compressed payload rather than the decompressed output, allowing a malicious client to send crafted requests that could consume excessive amounts of memory. Upgrading to Apache NiFi 2.11.0 is the recommended mitigation, which relocates response compression to Jetty Server and disables decompression of gzip-encoded HTTP requests.

netty · Netty netty

Netty is an asynchronous, event-driven network application framework. Prior to versions 4.1.136.Final and 4.2.16.Final, the `Bzip2Decoder` handler in Netty's compression codec pipeline is vulnerable to a denial-of-service attack through a malformed bzip2 stream that permanently captures the event-loop thread in an infinite loop. The vulnerability exists in the run-length encoding (RLE) state machine within [`Bzip2BlockDecompressor.read()`]. This issue has been fixed in versions 4.1.136.Final and 4.2.16.Final.

7-Zip · 7-zip 7-zip

7-Zip XZ Decompression Heap-based Buffer Overflow Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of 7-Zip. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the processing of XZ chunked data. Crafted XZ-compressed data can trigger an overflow of a heap-based buffer. An attacker can leverage this vulnerability to execute code in the context of the current process. Was ZDI-CAN-30169.

Linux

In the Linux kernel, the following vulnerability has been resolved: crypto: qat - remove unused character device and IOCTLs The QAT driver exposes a character device (qat_adf_ctl) with IOCTLs for device configuration, start, stop, status query and enumeration. These IOCTLs are not part of any public uAPI header and have no known in-tree or out-of-tree users. Device lifecycle is already managed via sysfs. The ioctl interface also increases the attack surface and is the subject of a number of bug reports. Remove the character device, the IOCTL definitions, and the related data structures (adf_dev_status_info, adf_user_cfg_key_val, adf_user_cfg_section, adf_user_cfg_ctl_data). Drop the now-unused adf_cfg_user.h header and strip adf_ctl_drv.c down to the minimal module_init/module_exit hooks for workqueue, AER, and crypto/compression algorithm registration. Clean up leftover dead code that was only reachable from the removed IOCTL paths: adf_cfg_del_all(), adf_devmgr_verify_id(), adf_devmgr_get_num_dev(), adf_devmgr_get_dev_by_id(), adf_get_vf_real_id() and the unused ADF_CFG macros. Additionally, drop the entry associated to QAT IOCTLs in ioctl-number.rst.

Linux · Linux linux kernel

In the Linux kernel, the following vulnerability has been resolved: 6lowpan: fix NHC entry use-after-free on error path lowpan_nhc_do_uncompression() looks up an NHC descriptor while holding lowpan_nhc_lock. If the descriptor has no uncompress callback, the error path drops the lock before printing nhc->name. lowpan_nhc_del() removes descriptors under the same lock and then relies on synchronize_net() before the owning module can be unloaded. That only waits for net RX RCU readers. lowpan_header_decompress() is also exported and can be reached from callers that are not necessarily covered by the net core RX critical section, for example the Bluetooth 6LoWPAN L2CAP receive path. This leaves a race where one task drops lowpan_nhc_lock in the error path, another task unregisters and frees the matching descriptor after synchronize_net() returns, and the first task then dereferences nhc->name for the warning. With the post-unlock window widened, KASAN reports: BUG: KASAN: slab-use-after-free in lowpan_nhc_do_uncompression+0x1f4/0x220 Read of size 8 lowpan_nhc_do_uncompression lowpan_header_decompress Fix this by printing the warning before dropping lowpan_nhc_lock, so the descriptor name is read while unregister is still excluded. The malformed packet is still rejected with -ENOTSUPP.

Linux · Linux linux kernel

In the Linux kernel, the following vulnerability has been resolved: net: af_key: initialize alg_key_len for IPComp states pfkey_msg2xfrm_state() handles the IPComp (SADB_X_SATYPE_IPCOMP) case by allocating x->calg and copying only the algorithm name: x->calg = kmalloc_obj(*x->calg); if (!x->calg) { err = -ENOMEM; goto out; } strcpy(x->calg->alg_name, a->name); x->props.calgo = sa->sadb_sa_encrypt; Unlike the authentication (x->aalg) and encryption (x->ealg) branches of the same function, the compression branch never initializes calg->alg_key_len. IPComp carries no key and the allocation only reserves sizeof(struct xfrm_algo) (i.e. no room for a key), so the field is left containing uninitialized slab data. calg->alg_key_len is later used as a length by xfrm_algo_clone() when an IPComp state is cloned during XFRM_MSG_MIGRATE: xfrm_state_migrate() xfrm_state_clone_and_setup() x->calg = xfrm_algo_clone(orig->calg); kmemdup(orig, xfrm_alg_len(orig)); where xfrm_alg_len() returns sizeof(*alg) + (alg_key_len + 7) / 8. With a non-zero garbage alg_key_len, kmemdup() reads past the end of the 68-byte calg object. Adding an IPComp SA via PF_KEY and then migrating it triggers (net-next, KASAN, init_on_alloc=0): BUG: KASAN: slab-out-of-bounds in kmemdup_noprof+0x44/0x60 Read of size 4164 at addr ff11000025a74980 by task diag2/9287 CPU: 3 UID: 0 PID: 9287 Comm: diag2 7.1.0-rc6-g903db046d557 #1 Call Trace: <TASK> dump_stack_lvl+0x10e/0x1f0 print_report+0xf7/0x600 kasan_report+0xe4/0x120 kasan_check_range+0x105/0x1b0 __asan_memcpy+0x23/0x60 kmemdup_noprof+0x44/0x60 xfrm_state_migrate+0x70a/0x1da0 xfrm_migrate+0x753/0x18a0 xfrm_do_migrate+0xb47/0xf10 xfrm_user_rcv_msg+0x411/0xb50 netlink_rcv_skb+0x158/0x420 xfrm_netlink_rcv+0x71/0x90 netlink_unicast+0x584/0x850 netlink_sendmsg+0x8b0/0xdc0 ____sys_sendmsg+0x9f7/0xb90 ___sys_sendmsg+0x134/0x1d0 __sys_sendmsg+0x16d/0x220 do_syscall_64+0x116/0x7d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f </TASK> Allocated by task 9287: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 pfkey_add+0x2652/0x2ea0 pfkey_process+0x6d0/0x830 pfkey_sendmsg+0x42c/0x850 __sys_sendto+0x461/0x4b0 __x64_sys_sendto+0xe0/0x1c0 do_syscall_64+0x116/0x7d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f The buggy address belongs to the object at ff11000025a74980 which belongs to the cache kmalloc-96 of size 96 The buggy address is located 0 bytes inside of allocated 68-byte region [ff11000025a74980, ff11000025a749c4) Depending on the uninitialized value the same field can instead request an oversized kmemdup() allocation and make the migration clone fail. The XFRM netlink path is not affected: verify_one_alg() rejects an XFRMA_ALG_COMP attribute shorter than xfrm_alg_len(), so a calg added via XFRM_MSG_NEWSA is always self-consistent. Initialize calg->alg_key_len to 0, matching the aalg/ealg branches.

netty · Netty netty

Netty is a network application framework for development of protocol servers and clients. In versions 4.2.0.Final through 4.2.15.Final and 4.1.0.Final through 4.1.135.Final, a remote unauthenticated peer can leak one direct `ByteBuf` per HTTP/2 `DATA` frame in applications that enable HTTP/2 content decompression via `DelegatingDecompressorFrameListener`. When a `DATA` frame is processed for a stream whose decompressor has already been closed, `Http2Decompressor.decompress(...)` calls `decompressor.writeInbound(data.retain())` and does not release the retained buffer on the error path, eventually exhausting direct memory and crashing the JVM. This issue is fixed in versions 4.1.136.Final and 4.2.16.Final.

netty · Netty netty

Netty is a network application framework for development of protocol servers and clients. Prior to 4.1.136.Final and 4.2.16.Final, Netty SPDY header decoding continues inflating zlib-compressed header blocks after the raw header parser has exceeded `maxHeaderSize` and marked the frame truncated in `SpdyFrameCodec`, allowing a remote peer to send a small compressed `HEADERS` block that expands into much larger raw header data and causes compression-amplified CPU and allocation churn. This issue is fixed in versions 4.1.136.Final and 4.2.16.Final.

NLNETLABS Net::DNS

Net::DNS versions through 1.55 for Perl allow Denial of Service via deep DNS compression pointer chains. Net::DNS::DomainName::decode follows RFC 1035 compression pointers by recursing into itself with no depth limit. It is possible to construct a name which saturates the call stack (at least with larger TCP responses), leading to a potential Denial of Service. The guard `$link < $offset` prevents forward and circular chains, but still allows arbitrarily long backward chains. The per-offset cache (`$cache`) is populated at the start of each call and short-circuits only re-traverses of the same offset - the initial descent through a fresh chain still recurses at full depth. A crafted packet can chain two-byte compression pointers so that each one points two bytes earlier than the previous, producing a chain length of `offset / 2`. For the 14-bit pointer field (max offset 16383) this gives up to ~8191 recursive frames. For a TCP DNS message the limit is the 16-bit length field (~32767 frames). Perl's default C stack handles only a few thousand frames; beyond that the process receives SIGSEGV or similar, which is a denial-of-service for any application parsing untrusted DNS data. The vulnerability is triggered by `Net::DNS::Packet->new(\$wire)` i.e. any point where the library decodes a DNS message from the network.

Linux · Linux linux kernel

In the Linux kernel, the following vulnerability has been resolved: erofs: fix managed cache race for unaligned extents After unaligned compressed extents were introduced, the following race could occur: [Thread 1] [Thread 2] (z_erofs_fill_bio_vec) <handle a Z_EROFS_PREALLOCATED_FOLIO folio> ... filemap_add_folio (1) (z_erofs_bind_cache) <the same folio is found..> .. .. folio_attach_private (2) filemap_add_folio (3) again Since (1) is executed but (2) hasn't been executed yet, it's possible that another thread finds the same managed folio in z_erofs_bind_cache() for a different pcluster and calls filemap_add_folio() again since folio->private is still Z_EROFS_PREALLOCATED_FOLIO. Fix this by explicitly clearing folio->private before making the folio visible in the managed cache so that another pcluster can simply wait on the locked managed folio as what we did for other shared cases [1]. This only impacts unaligned data compression (`-E48bit` with zstd, for example). [1] Commit 9e2f9d34dd12 ("erofs: handle overlapped pclusters out of crafted images properly") was originally introduced to handle crafted overlapped extents, but it addresses unaligned extents as well.

h2o · H2o h2o

h2o is an HTTP server with support for HTTP/1.x, HTTP/2 and HTTP/3. Prior to commit 9265bdd, there is an HTTP/2 state amplification issue that combines HPACK decompression amplification with Slowloris-style stream stalling. Amplified decoded header state can be retained by stalled HTTP/2 streams, and depending on the configuration, additional limits are needed to bound decoded header state and prevent attack. This issue has been fixed by commit 9265bdd.

centrifugal centrifugo

Centrifugo is an open-source scalable real-time messaging server. Prior to 6.8.4, Centrifugo unidirectional WebSocket transport with uni_websocket.compression enabled enforced uni_websocket.message_size_limit against compressed wire-frame length in internal/websocket/conn.go advanceFrame, but ReadMessage used io.ReadAll after decompression without an output cap, allowing unauthenticated requests to /connection/uni_websocket to trigger large memory and CPU consumption. This issue is fixed in version 6.8.4.

tornadoweb tornado

Tornado is a Python web framework and asynchronous networking library. Prior to 6.5.6, Tornado gzip decompression routines processed limited-size chunks but did not enforce an overall limit on accumulated decompressed chunks, allowing a malicious server accessed by SimpleAsyncHTTPClient or an HTTPServer configured with decompress_request=True to consume effectively unlimited memory. This issue is fixed in version 6.5.6.

Red Hat Enterprise Linux 10 · Red Hat Enterprise Linux 10.0 Extended Update Support · Red Hat Enterprise Linux 8.4 Advanced Mission Critical Update Support · Red Hat Enterprise Linux 8.4 Extended Update Support Long-Life Add-On

A flaw was found in libsoup's WebSocket implementation when using the permessage-deflate extension. The extension's decompression loop (inflate()) processes data in chunks without enforcing an upper boundary limit on the output buffer size. While libsoup limits the incoming compressed frame size via max_incoming_payload_size, it fails to track or limit memory allocation during decompression. A separate check for decompressed size (max_total_message_size) exists but executes only after inflation is complete, and it is entirely disabled by default for client connections. A remote, unauthenticated attacker can exploit this by sending a small, highly compressed payload (a decompression bomb), causing unbounded memory allocation that triggers an Out-of-Memory (OOM) crash and a Denial of Service (DoS).

Microsoft Windows 10 Version 1607 · Microsoft Windows 10 Version 1809 · Microsoft Windows 10 Version 21H2 · Microsoft Windows 10 Version 22H2

Improper access control in Windows Audio Compression Manager (ACM) allows an authorized attacker to elevate privileges locally.