CVEs of 2026.
Every CVE published in 2026 that we hold, ranked by exploit odds.
-
CVE-2026-101184
Permission control vulnerability in the Bluetooth module. Impact: Successful exploitation of this vulnerability may affect availability.
-
CVE-2026-101182
Race condition vulnerability in the location service module. Impact: Successful exploitation of this vulnerability may affect availability.
-
CVE-2026-55360
In secure_transport_usecase_handshake of secure_transport.c, there is a possible way to bypass secure transport channel encryption due to a logic error. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.
-
CVE-2026-92606
Time-of-check time-of-use (TOCTOU) race condition vulnerability in TÜBİTAK BİLGEM Software Technologies Research Institute Pardus MyComputer allows Leveraging Time-of-Check and Time-of-Use (TOCTOU) Race Conditions. This issue affects Pardus MyComputer: from 0.8.0 before 0.8.1.
-
CVE-2026-56916
In _vqueue_get_avail_buf_locked of trusty/kernel/lib/trusty/vqueue.c, there is a possible Double-read / TOCTOU due to a race condition. This could lead to local escalation of privilege with execution privileges needed. User interaction is for exploitation.
-
CVE-2026-116196
Rclone before v1.75.2 contains a credential exposure vulnerability in the b2 backend that leaks the live account Authorization token and SSE-C keys on HTTP redirects from --b2-download-url hosts. Attackers can observe a same-host HTTPS-to-HTTP downgrade redirect to capture the cleartext B2 token, or receive cross-host redirects disclosing SSE-C customer encryption keys.
-
CVE-2026-56898
In aes256_cmac_sw of aes.rs, there is a possible way to forge message authentication codes over multi-block messages due to improperly used crypto. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.
-
CVE-2026-114024
This issue was addressed with improved checks. This issue is fixed in iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, visionOS 27.2, watchOS 27.2. An app may be able to access sensitive user data.
-
CVE-2026-112603
A logic issue was addressed with improved checks. This issue is fixed in Xcode 27.2. An app may be able to gain root privileges.
-
CVE-2026-112662
A race condition was addressed with additional validation. This issue is fixed in iOS 27.2 and iPadOS 27.2. An app may be able to access sensitive user data.
-
CVE-2026-113953
An integer overflow was addressed with improved input validation. This issue is fixed in iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, macOS Sequoia 15.8.2, macOS Tahoe 26.7.2, tvOS 27.2, visionOS 27.2. Processing a maliciously crafted USD file may lead to unexpected app termination.
-
CVE-2026-113955
A cross-origin issue was addressed with improved tracking of security origins. This issue is fixed in iOS 26.7.2 and iPadOS 26.7.2. Processing maliciously crafted web content may disclose sensitive user information.
-
CVE-2026-134169
Qwik City before 1.20.3 (@builder.io/qwik-city) and @qwik.dev/router before 2.0.0-rc.4 contain an improperly controlled object modification vulnerability in formToObj that allows unauthenticated attackers to overwrite built-in function properties. Attackers can send a form POST with nested field names traversing inherited members of string values, breaking Object.prototype.hasOwnProperty.call process-wide until server restart.
-
CVE-2026-107942
Improper Content Security Policy (CSP) nonce implementation in HCL BigFix MCM allows a remote attacker to reduce the effectiveness of the Content Security Policy protection mechanism via an orphaned nonce value in the style-src directive of the CSP header without a corresponding nonce attribute on any <style> or <script> element.
-
CVE-2026-121714
RabbitMQ is a messaging and streaming broker. In the affected revision identified by the advisory, the AMQP 1.0 server_mode parser in deps/amqp10_common/src/amqp10_binary_parser.erl stops at the first body section and preserves the remaining bytes in bare_and_footer without validating message-section order. An authenticated publisher with write permission can place a properties section carrying a forged user-id after the body, causing rabbit_access_control:check_user_id/2 to see no user-id while permissive downstream AMQP 1.0 decoders or AMQP 0-9-1 conversion through classic and quorum queues expose the forged identity. Applications that rely on RabbitMQ-validated user-id metadata can therefore accept attacker-controlled publisher attribution. Strict consumers that reject out-of-order sections are not vulnerable to identity spoofing, and the issue does not bypass broker routing or destination authorization. This issue is fixed in version 4.3.5.
-
CVE-2026-94597
NukeViet CMS is a multi Content Management System. Prior to 4.5.12 and 4.6.02, an attacker who can reach an image-upload flow can upload a valid image with appended PHP short-tag code and a double-extension filename. NukeViet\Files\Upload::save_file() relied on verify_image(), which missed the <? short-open tag, while filename handling preserved embedded dots, validation inspected only the final extension, and an .htaccess rule blocked PHP extensions only at the end of the URI. On Apache mod_php systems configured with short_open_tag and a PHP extension handler, the stored polyglot can execute as PHP and run operating-system commands as the web-server user; PHP-FPM, nginx, and end-anchored handler configurations are not exposed to the execution step. This issue is fixed in versions 4.5.12 and 4.6.02.
-
CVE-2026-125273
No replay or command-integrity protection. Commands lack sequence numbers, nonces, or session tokens; captured commands can be replayed, and the last command persists across sessions.
-
CVE-2026-124281
The affected products use an unprotected firmware-update channel, which may allow for unauthorized firmware modification.
-
CVE-2026-95825
net_if_ipv4_addr_rm() and net_if_ipv6_addr_rm() in subsys/net/ip/net_if.c unconditionally called net_if_addr_unref() on the target address. When an earlier removal could not complete because another subsystem still held a reference (the TCP stack takes one on the local address of every connection, in tcp_conn_new() and on connect()), the address is left in place with is_added cleared and is_used still set. A second removal of such an address decremented the reference count again, consuming a reference owned by that other holder. When it was the last one, the address slot was released, its NET_EVENT_IPV*_ADDR_DEL event fired, and the slot became reusable while the holder still believed it owned a reference. Every address lookup helper matches on is_used and not on is_added, so a half-removed address stays visible to the rest of the stack and repeat removals are readily reachable from remote input. Under IPv6 SLAAC, an on-link node that sends a Router Advertisement with a short valid lifetime causes the address to expire into the deferred state; a refreshing Router Advertisement is handled by net_if_ipv6_addr_update_lifetime(), which re-arms the lifetime timer without restoring the reference, so the next expiry performs the second, reference-stealing removal. With CONFIG_NET_IPV4_ACD and DHCPv4, repeated spoofed ARP conflicts for the device's lease drive dhcpv4_acd_event_handler() into calling net_if_ipv4_addr_rm() more than once on the same address. Where CONFIG_USERSPACE is enabled, the net_if_ipv4_addr_rm_by_index() and net_if_ipv6_addr_rm_by_index() system calls let an unprivileged user-mode thread trigger the same kernel-side release deterministically. The impact is loss of a configured IP address while connections are still using it, and a corrupted reference count that propagates: the slot can be handed to a newly added address, and when the stale holder later tears down, net_if_addr_unref() resolves the address by value and drops the new incarnation's only reference, removing a live address a second time. The affected objects are static array slots and the counter is never decremented below zero, so there is no memory corruption or information disclosure; the consequence is denial of service against the device's network connectivity, repeatable at the attacker's pace. The fix resolves the address before touching the reference count and returns false without calling net_if_addr_unref() when the address is already marked as removed, so the public removal API can no longer consume a reference it does not own.
-
CVE-2026-95827
The IPv6 neighbour discovery code in subsys/net/ip/ipv6_nbr.c keeps packets that are waiting for address resolution on a per-neighbour pending queue. A queued packet holds two references: the one handed over by the sender and a second one taken by net_ipv6_send_ns() when the packet is put on the queue. The Router Advertisement handler handle_ra_input() popped each packet off the queue and passed it to net_send_data(), relying on a trailing nbr_clear_ns_pending() call to release the queue reference — but that call then found the queue already empty and released nothing, so one reference was leaked for every pending packet flushed on an RA. Router Advertisements are unauthenticated link-local ICMPv6 messages, so any node on the same link can trigger the flush. An attacker can make the leak deterministic by advertising itself as a router without a source link-layer address option, so the victim's neighbour entry stays INCOMPLETE, then inducing the victim to transmit through that next hop (the packet is queued and a neighbour solicitation is sent), leaving the solicitation unanswered and sending a further RA to drain the queue. Each round leaks one struct net_pkt together with its net_buf chain. Leaked packets are never returned to the fixed-size pools, so the loss is permanent for the lifetime of the boot. With CONFIG_NET_PKT_TX_COUNT defaulting to 4 (14 with CONFIG_NET_PKT_LARGE_DEFAULTS), a small number of repetitions exhausts the transmit packet slab and the node stops being able to send any traffic. The impact is availability only: the reference count is always too high, never too low, so no use-after-free or memory corruption arises. The fix moves the flush into a helper, nbr_send_pending(), which releases the queue reference for every packet it takes off the queue before handing the remaining reference to net_send_data(), and releases that one too if the send is refused.
-
CVE-2026-77350
pmo-platform is a modular PMO and release-management platform for Claude Code. Prior to 3.69.1, core/hooks/block-credential-reads.sh, core/hooks/block-egress.sh, core/hooks/block-fs-boundary.sh, and core/hooks/block-mcp-writes.sh pin PATH to /usr/bin:/bin, hard-code jq as /usr/bin/jq, and return success when jq is unavailable there. On systems where jq exists only in a Homebrew path, including affected pre-Sequoia or Homebrew-only macOS installations, the hooks can skip credential-read, network-egress, filesystem-boundary, and MCP-write controls. Native deny rules independently protect some credential paths but do not backstop the egress or MCP controls. An attacker who can influence a prompt, repository, or evaluation flow can use the skipped checks to expose secrets or perform unauthorized tool actions. This issue is fixed in version 3.69.1.
-
CVE-2026-77389
pmo-platform is a modular PMO and release-management platform for Claude Code. Prior to 3.73, core/hooks/block-fragile-refs.sh and core/hooks/block-gh-path-leak.sh source co-shipped awk primitives from core/hooks/lib/ and degrade when a required primitive is absent. In a partial or corrupted deployment, block-gh-path-leak.sh skips its entire path-leak check, while block-fragile-refs.sh skips only the positional-issue-reference detector and continues running its other detectors. Both hooks operate in a warning posture, so the impact is a skipped warning rather than bypass of an enforcing control, with no established exfiltration or code-execution primitive. This issue is fixed in version 3.73.
-
CVE-2026-62320
Netatalk is a Free and Open Source file server suite for Unix-like operating systems. From 1.5.0 until 4.5.1, dsi_tcp_open in libatalk/dsi/dsi_tcp.c silently limits the bytes read from an unauthenticated DSI frame received on TCP/548 to server_quantum when the attacker-controlled dsi_len is larger, without rejecting the frame or draining the declared remainder. The same silent dsi_len truncation pattern exists in libatalk/dsi/dsi_stream.c. The leftover bytes remain queued on the TCP connection and the next dsi_stream_receive call interprets them as a new DSI header and payload, allowing a remote unauthenticated attacker to smuggle one or more additional attacker-controlled DSI frames into the same connection. This issue is fixed in version 4.5.1.
-
CVE-2026-62319
Netatalk is a Free and Open Source file server suite for Unix-like operating systems. From 1.5.0 until 4.5.1, dsi_stream_receive in libatalk/dsi/dsi_stream.c independently clamps dsi_doff and dsi_len for DSIWrite requests without enforcing that the data offset is no greater than the declared frame length. An authenticated AFP client can send dsi_doff greater than dsi_len, causing the receiver to consume bytes from a queued subsequent frame before libatalk/dsi/dsi_write.c detects the inconsistency. The irreversible stream desynchronization lets the attacker control the next parsed DSI header and payload and corrupts protocol framing for the connection. This issue is fixed in version 4.5.1.
-
CVE-2026-49390
Netatalk is a Free and Open Source file server suite for Unix-like operating systems. From 3.0.0 until 4.5.0, getoption_int in libatalk/util/netatalk_conf.c accepts an out-of-range server quantum value from afp.conf without applying the documented fallback to the default. An administrator can unintentionally configure a very large allocation whose failure causes AFP_PANIC at startup or a very small command buffer that disrupts DSI framing, leaving afpd unavailable until the configuration is corrected and the daemon is restarted. The option is read locally at startup and cannot be changed through the AFP protocol. This issue is fixed in version 4.5.0.
-
CVE-2026-133212
@fastify/reply-from is a Fastify plugin that forwards an incoming HTTP request to an upstream server. In versions before 12.9.1, when the plugin forwards a request whose body is not written to the upstream connection, such as a GET, HEAD, or TRACE request that carried a body, or a request with a falsy JSON body, it still forwards the client's Content-Length header while sending no body. The upstream server then treats the bytes of the following request as the unfinished body, which an attacker can use to desynchronize the connection and smuggle a request. The issue is fixed in @fastify/reply-from 12.9.1, and users should upgrade to 12.9.1 or later.
-
CVE-2026-77331
memtomem-stm is a short-term memory proxy gateway with proactive memory surfacing for AI agents. From 0.1.2 until 0.1.34, concurrent surfacing RPC failures can start more than one reconnect for the same dirty session because concurrent LTM recovery is not uniquely owned by a session generation. Cancellation during a post-reconnect retry can also leave the replacement transport reusable after an interrupted protocol exchange, causing session-state integrity loss or local service degradation. The duplicate-reconnect subcase was introduced in 0.1.24, while the cancellation-recovery subcase affects the broader listed range. This issue is fixed in version 0.1.34.
-
CVE-2026-77332
memtomem-stm is a short-term memory proxy gateway with proactive memory surfacing for AI agents. From 0.1.0 until 0.1.34, untrusted long-term and working-memory fields can introduce Markdown or HTML-compatible structure into the surfaced-memory envelope because those fields are not serialized as inert data. A crafted memory can create misleading structural boundaries or visually conceal surfaced content, affecting how a consuming agent interprets the memory context. Ordinary natural-language persuasion remains a residual model-level risk and is outside the structural serialization flaw. This issue is fixed in version 0.1.34.
-
CVE-2026-77329
Tapo provides Rust and Python APIs and an MCP server for TP-Link Tapo smart devices. From 0.3.1 until 0.5.0, the Streamable HTTP implementation in tapo-mcp/src/server.rs calls disable_allowed_hosts() without replacement Host or Origin validation, while TAPO_MCP_API_KEY remains optional for a listener that can bind to all interfaces. A malicious website can use DNS rebinding to send same-origin browser requests to the victim's local or LAN Tapo MCP listener when it runs without an API key. The attacker can invoke exposed tools including control_device for smart-device actuation and take_snapshot for configured cameras. Deployments with TAPO_MCP_API_KEY set are not affected by the unauthenticated path, and exploitation requires a running listener configured with real Tapo credentials and devices, plus victim interaction with the malicious site. This issue is fixed in version 0.5.0.
-
CVE-2026-133629
A vulnerability was determined in Sunnet CTMS up to 10.11. Impacted is an unknown function of the file /cltcms/api/external-training/ext-course of the component External Training Application Form Feature. This manipulation of the argument memberids causes improper control of resource identifiers. Remote exploitation of the attack is possible. The exploit has been publicly disclosed and may be utilized. The vendor was contacted early about this disclosure but did not respond in any way.
-
CVE-2026-106024
Mongoose Embedded Web Server / Networking Library before 7.24 contains a WebSocket fragmentation validation vulnerability that allows remote attackers to spoof message metadata by sending a single CONTINUATION frame with FIN=1 as the first data frame instead of a required TEXT/BINARY start frame, violating RFC 6455 sequencing rules. Attackers can craft this malformed frame to make mg_ws_cb() in src/ws.c misinterpret the first byte of payload as saved FIN/opcode flags, causing the application to receive a dispatched MG_EV_WS_MSG event with attacker-controlled flags and data truncated by one byte without closing the connection.
-
CVE-2026-133702
free5GC PCF through 1.4.6 contains a resource leak in postAppSessCtxProcedure that leaves PCC rules and QoS data in the SM policy decision after rejected Npcf_PolicyAuthorization create requests. Authenticated AFs knowing an active PDU session's UE address can send repeated failing requests with distinct AF application identifiers to grow PCF memory and pollute policy sent to the SMF.
-
CVE-2026-133705
Hanko before 3.2.0 contains an open redirect vulnerability in Saml.IsAllowedRedirect that allows unauthenticated attackers to bypass the redirect allowlist by appending attacker-controlled domain suffixes to allowed hostnames. Attackers can lure victims through a crafted /saml/auth login link to capture the one-time hanko_token and redeem it without PKCE to hijack the victim's session.
-
CVE-2026-133700
free5GC SMF through 1.4.6 contains a resource leak vulnerability in HandlePDUSessionSMContextCreate that leaves SM contexts unreleased when UDM access token retrieval fails. Attackers can trigger repeated PDU session establishment attempts while the NRF is unreachable, causing empty 200 responses and persistent SMF memory growth.
-
CVE-2026-101340
Net::DNS uses a 16-bit query ID to match responses to the corresponding requests. In versions up to and including 1.57, the ID is obtained from Perl's built-in rand() function. An attacker that can observe the sequence of query IDs is able to expose the internal state of the PNRG and hence predict query IDs before they are transmitted. The victim Net::DNS instance is then exposed to forged responses injected by the attacker. Note that the deduplication cache in 1.57 has no impact on the feasibility of this attack.
-
CVE-2026-131167
pretix offloads certain processes to background tasks that are run by a different server to avoid HTTP workers being clogged with processes that take longer than a usual web request. The browser periodically polls the server to see if the task is already finished using a UUID-shaped task ID. Once the task is completed, pretix resumes to the next step of the operation based on the result of the task. The system does however not check if the result stems from the correct type of task. Therefore, an attacker can supply the UUID of a task of a different type and pretix will continue with the result from the wrong task. In most cases, this will just result in an error because the return types do not match up, but in some cases this can be abused to extract objects the attacker usually should not see. Specifically, a vector was discovered that allowed to abuse this for unauthorized access to other people's ticket purchases. The attacker can, however, not control which purchases they get access to.
-
CVE-2026-131170
Our payment integration with Stripe did not properly validate payment status responses. An attacker could use a successful payment status response from one payment and supply it to the system for a different payment, gaining access to multiple valid tickets with only one payment. However, the attack requires knowledge of a valid charge ID from the same event which we believe is usually not exposed to users. To simplify attack surface, the fix removes support for incoming webhooks regarding payments created through Stripe's old "source" API. pretix stopped creating payments using this API in 2024.1 for most payment methods and in 2024.9 for all payment methods.
-
CVE-2026-86586
Certain EZVIZ devices are affected by a replay attack vulnerability (CWE-294). An attacker may intercept communication traffic and replay specific control commands, causing the device to execute unauthorized operations.
-
CVE-2026-59599
Android on supported Pixel devices before security patch level 2026-11-05 allows elevation of privilege in the hypervisor subcomponent in the kernel, aka A-519434175.
-
CVE-2026-73860
CRIU is a checkpoint and restore tool for Linux. Prior to 4.2.1, CRIU injects a syscall;int3 sequence at the start of the first executable virtual memory area and resumes the target process to execute parasite code. Malicious code in the target process can keep an rseq critical section armed over that injection point so the kernel's restartable-sequence abort redirects execution to an attacker-controlled callback when CRIU resumes the process. In container snapshot environments, the hijacked parasite runs with seccomp disabled and a relaxed AppArmor profile and supplies trusted data to the privileged CRIU process, which can enable higher-privilege code execution. This issue is fixed in version 4.2.1.
CVE records published in 2026 (ids starting CVE-2026-), sorted by how likely they are to be exploited.