CVE-2026-19935 is a use-after-free vulnerability in the Zephyr Bluetooth LE host's handling of L2CAP connection-oriented channel (CoC) data when using dynamic PSMs (0x0080-0x00FF). The issue arises because the channel teardown process does not cancel a queued work item (rx_work) that references the channel, allowing a remote unauthenticated peer to trigger use-after-free by sending a data K-frame followed immediately by an L2CAP Disconnect Request. This can lead to memory corruption or workqueue malfunction. The vulnerability affects Zephyr versions from 2.0.0 through 4.4.2 and has a CVSS 3.1 score of 7.5 (high severity).
CVE-2026-19737 is a memory-safety vulnerability in the Zephyr project's i2s_esp32 driver affecting versions from 4.4.0 up to but not including 4.5.0. The issue arises when the driver dereferences NULL pointers for I2S directions not implemented by the device, leading to a kernel-mode read from address zero. This can be triggered by an unprivileged user-mode thread with access to the I2S device, causing a system-wide denial of service due to a fatal exception. The vulnerability does not allow information disclosure or code execution. A fix has been merged that adds proper pointer checks and returns an error for unsupported directions.
CVE-2026-18418 is a low-severity vulnerability in the Zephyr project affecting the zbus proxy agent IPC backend. It involves improper handling of a non-NUL-terminated channel_name field in inter-domain frames, which can cause out-of-bounds reads during logging. This may lead to limited memory disclosure in logs or a potential fatal fault. The vulnerability requires specific configuration and conditions, including control over a peer domain's firmware and logging at warning level or above.
CVE-2026-19736 is a vulnerability in the NXP MCUX TRNG entropy driver used in Zephyr OS on certain i.MX RT5xx and RT6xx SoCs. The driver passes the requested byte count directly to the vendor SDK's TRNG_GetRandomData() function, which rounds entropy requests up to multiples of 128 bytes and copies data in 32-bit words. Requests for entropy with lengths not divisible by four cause an underflow in the SDK's buffer size check, resulting in up to 127 bytes of memory being overwritten beyond the caller's buffer. This overflow can corrupt kernel memory or adjacent user memory, potentially causing crashes or privilege escalation. The issue affects Zephyr versions from 4.3.0 through 4.4.2 on affected hardware. A patch is available that ensures only word-aligned requests are passed to the SDK, preventing overflow.
CVE-2026-19735 is a medium-severity vulnerability in the Zephyr project's TCP initial sequence number (ISN) generation. Due to improper handling of failures in the cryptographic random source, the unique key used for ISN generation can remain all zeros, making ISNs predictable. This weakens the protection against blind TCP connection spoofing and data injection. The issue affects Zephyr versions from 3.7.0 through 4.4.2. A fix has been implemented to ensure key generation only latches on success and falls back to a less secure random source if necessary.
CVE-2026-19577 is a vulnerability in the Zephyr project affecting the IPv6 neighbor cache handling. It involves an out-of-bounds read when resolving the next-hop link-layer address for IPv6 packets. An unauthenticated attacker on the same link can trigger this by causing unresolved neighbor cache entries and then sending packets to those neighbors. This can lead to a denial of service via device panic if assertions are enabled, or information disclosure of static RAM contents if assertions are disabled.
CVE-2026-19576 is a vulnerability in the Goodix GT9xx input driver of the Zephyr project. The driver reads a touch point count from the controller and uses it unchecked as a loop bound to fill a stack array sized for a maximum of 5 touch points. A malicious or faulty controller can report more points, causing a stack buffer overflow with attacker-controlled data. This can lead to kernel privilege code execution, out-of-bounds reads, and system crashes. The vulnerability requires physical or local hardware access to the I2C touch controller. A fix is available that clamps the reported touch count to the configured maximum before indexing the array.
CVE-2026-19669 is a high-severity memory safety vulnerability in the Zephyr RTOS affecting user-mode syscall verifiers for fuel gauge device properties. It arises from unchecked variable-length arrays sized by an attacker-controlled length parameter, leading to out-of-bounds writes on the kernel stack. This can allow a user-mode thread with appropriate permissions to overwrite kernel memory, potentially resulting in kernel code execution or system crashes.
The luoxy-baileys npm package, a fork of the Baileys WhatsApp Web library, contains malicious code that automatically subscribes the user's authenticated WhatsApp account to certain WhatsApp Channels (newsletters) chosen by the publisher. This auto-subscription occurs without explicit user consent, triggered 90 seconds after a WhatsApp connection is established. The list of channels to follow is partially hardcoded and partially fetched at runtime from a publisher-controlled GitHub URL, allowing dynamic modification of subscribed channels. The package does not exfiltrate credentials or install persistent backdoors, but it covertly inflates the publisher's channel subscriber counts. This behavior is enabled by default and undocumented except for a comparison table in the README. A user can opt out by setting a configuration flag to false.
The hridoy-ultimate npm package version 8.0.0 is an unofficial Facebook chat client containing malicious behaviors. It forwards user Facebook credentials and 2FA secrets to a hardcoded third-party server, exposing user accounts to full compromise. Additionally, it dynamically pulls and executes code from an unpinned GitHub repository during installation, allowing arbitrary code execution without integrity checks. A remote-control WebSocket client is present but opt-in and gated by configuration.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the SysUserController getUserDetailByUserId handler. This flaw allows any authenticated user to access other users' personal details by supplying arbitrary userId values. Exposed information includes real names, usernames, emails, phone numbers, birthdays, employee numbers, department paths, and posts. The vulnerability has a medium severity with a CVSS score of 4.3.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the SysCommentController exportXls handler. This flaw allows any authenticated user, including those with low privileges, to export all comments from the system. Attackers can access the /sys/comment/exportXls endpoint to download every sys_comment row, exposing comment text and user IDs for records they should not be able to access.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the SysDepartRoleController exportXls handler. This flaw allows any authenticated user, including those with minimal default roles, to export department role data. The exposed data includes role names, codes, descriptions, and the users who created them. The vulnerability has a CVSS score of 4.3, indicating medium severity.
JeecgBoot versions up to 3.9.5 have a missing authorization vulnerability in the PUT /sys/dict/editDictByLowAppId endpoint. This flaw allows any authenticated user to modify low-code application dictionaries by supplying a dictionary's low_app_id via parameters or headers. The vulnerability enables renaming dictionaries and replacing their items without proper authorization.
JeecgBoot versions through 3.9.5 have an insecure direct object reference vulnerability in the AiragBaseApiController. This flaw allows authenticated users to read AI chat variables belonging to other users by manipulating the username parameter in a specific API endpoint. The vulnerability involves sending POST requests to /airag/api/getChatVariable with a target appId, username, and variable name to access stored chat memory values in Redis. The CVSS score is 4.3, indicating a medium severity issue.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the getTenantPackInfo handler. This flaw allows any authenticated user to read product pack membership information of other tenants. Low-privileged attackers can exploit this by supplying a tenantId and fixed packCode values to disclose administrator usernames, real names, phone numbers, and departments. The vulnerability has a medium severity rating with a CVSS score of 4.3.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the editThirdAppConfig handler. This flaw allows any authenticated user, including low-privileged attackers, to modify configurations of third-party integrations such as DingTalk, WeCom, or Feishu. Attackers can replace sensitive parameters like client id, client secret, agent id, and corp id, potentially redirecting directory synchronization and messaging to attacker-controlled applications or disrupting these services.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the SysMessageTemplateController delete handler. This flaw allows any authenticated user, including low-privileged attackers, to delete message templates by obtaining template IDs from an unprotected list endpoint. Deleting these templates can disrupt system notices and workflow reminders.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the AiragAppController's queryById handler. This flaw allows low-privileged authenticated users to access AI application configurations belonging to other users. Attackers can enumerate application IDs via the unprotected /airag/app/listDict endpoint and then query each ID to retrieve sensitive configuration details such as system prompts, memory prompts, model IDs, knowledge base IDs, and plugin bindings.
JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the AiragPromptsController. This flaw allows any authenticated user, including low-privileged attackers, to delete AI prompt templates belonging to other users by exploiting unprotected API endpoints. Specifically, attackers can retrieve template IDs from the unguarded GET /airag/prompts/list endpoint and then delete those templates via DELETE /airag/prompts/delete. The vulnerability results in denial of availability of prompt templates but does not impact confidentiality.