CVE-2026-116620 is a medium severity vulnerability in Tencent RapidJSON versions 1.0 and 1.1.0. It involves an out-of-bounds read in the GenericStringStream<UTF8<char>>::Take function within the Regex Parser component. This flaw can be triggered remotely and has been publicly disclosed. No patch or vendor response is currently available.
A security flaw has been discovered in Tencent ncnn up to 44b02dabdcf97499281a4692ebd65bad73dcf8b6. Affected by this issue is the function ncnn::yuv420sp2rgb of the file src/mat_pixel.cpp of the component Conversion API. The manipulation results in integer overflow. The attack can be launched remotely. The exploit has been released to the public and may be used for attacks. The project was informed of the problem early through an issue report but has not responded yet.
CVE-2026-108821 is a medium severity SQL injection vulnerability in Tencent WeKnora versions 0.8.0 and 0.8.1. It affects the validateSubquery function in the internal/utils/inject.go file of the SQL Validator component. The vulnerability allows remote attackers to perform SQL injection attacks. A public exploit disclosure exists. Upgrading to version 0.8.2 addresses this issue.
CVE-2026-107830 is a medium severity vulnerability in the banq jivejdon product. It involves a lack of rate limiting on the unauthenticated /account/smsVRAction endpoint, allowing attackers to send unlimited SMS messages. This can be exploited to harass arbitrary phone numbers and exhaust the Tencent Cloud SMS balance associated with the operator. The vulnerability affects versions of jivejdon from commit e0306088 through commit ee67a65e. No patch or remediation information is provided.
A vulnerability was identified in Tencent AI-Infra-Guard up to 4.5.2/4.6.2. This affects the function startsWith of the file skill_scan/tools/dir/dir_actions.py of the component File Access. The manipulation leads to path traversal. The attack needs to be performed locally. The exploit is publicly available and might be used. Upgrading to version 4.6.0 is able to mitigate this issue. The identifier of the patch is ac0384edc9dbea3b226edefcf50613bd8509134f. You should upgrade the affected component.
A critical vulnerability (CVE-2026-51990) in Tencent's Sogou Input Method for Windows has been exploited by the Chinese threat group UNC3569 to achieve system-level code execution via a crafted URL. The flaw involves unvalidated command-line argument injection, unrestricted URL navigation, and an outdated Chromium browser engine with disabled sandboxing. The exploit enables deployment of the GrayRabbit backdoor, which provides extensive remote control capabilities. Tencent released a patch in version 16.3.0.3498 that addresses the URL parameter validation but did not update the underlying Chromium configuration. The vulnerability has been actively exploited in the wild.
A critical remote code execution vulnerability (CVE-2026-51990) in Tencent's Sogou Input Method for Windows is actively exploited by the UNC3569 threat group to deploy the GrayRabbit backdoor malware. The attack chain abuses a custom URI handler, an unrestricted URL navigation in an embedded Chromium webview, and a known flaw in an outdated Chromium 80 engine running without sandboxing. Tencent released a patch in version 16.3.0.3498 that validates URL arguments and restricts navigation to approved domains, but the underlying Chromium engine remains outdated and unsandboxed.
The npm package @stellarshift/chain-metadata versions 1.0.1 and 1.0.3 contain malicious code in their postinstall script that fetches and executes arbitrary code from a remote, unrelated Tencent Cloud URL. This execution occurs on macOS and Windows systems with specific software installed, targeting corporate developer workstations while evading CI and sandbox environments. The package falsely advertises no network calls and zero runtime dependencies, contradicting its actual behavior. Installation of this package can lead to full system compromise.
Tencent AI-Infra-Guard's skill-scan component excludes compiled Python bytecode files from analysis by hardcoding __pycache__ directories and .pyc/.pyo/.pyd extensions into skip lists across multiple scanning surfaces. Attackers can distribute skills with benign Python source files alongside malicious compiled bytecode that executes on import while the scanner reports a safe verdict, enabling code execution when operators install the skill.
A heap out-of-bounds read vulnerability in the Linux kernel's Network Controller Sideband Interface (NCSI) driver was fixed. The flaw occurs when the payload length field in an NCSI command is larger than the actual netlink attribute data, leading to reading beyond the buffer and leaking kernel heap memory. This vulnerability requires CAP_NET_ADMIN privileges and affects systems with registered NCSI devices, such as OpenBMC on Aspeed BMC SoCs. The issue was discovered by Tencent Yunding Lab's dynamic kernel fuzzer and has been patched in affected Linux kernel versions.
CVE-2026-72565 is a critical SQL injection vulnerability in Tencent APIJSON up to version 7.1.0. It allows unauthenticated remote attackers to bypass per-table access controls and read arbitrary database tables using the Map-form @having operator. The vulnerability has a high CVSS score of 9.8, indicating severe impact on confidentiality, integrity, and availability. No official patch or remediation guidance is currently available from the vendor. There are no known exploits in the wild at this time.
The @aster110/cc2wechat npm package version 5.1.0 runs a daemon that polls Tencent's iLink Bot API for WeChat messages and forwards them to a locally hosted AI agent with all safety and sandbox restrictions disabled. This setup allows any sender of a WeChat message to execute arbitrary shell and filesystem commands on the host machine running the daemon. The remote shell capability activates only when the daemon is started via the CLI, not during installation or import.
The chatcc-agent npm package runs a daemon that connects to Tencent Cloud IM and accepts remote commands via WebSocket messages. These commands can spawn or write to a local shell or PTY on the host machine, effectively providing a remote shell access channel. Access is controlled by JWT and paired-client authentication, but the design inherently exposes a remote command execution surface. The npm install script only performs local native builds and does not fetch remote content. Version 0.8.24 is affected.
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (0457a026ab28fbb641d11e4e9ac9d60d026fd77210972f72e6ac931b4b06c59c)
On npm install, postinstall.js executes shell commands that query cloud instance metadata services (AWS IMDS at 169.254.169.254, plus Tencent, Aliyun, GCP, and Azure endpoints) and extracts the AWS IAM role's temporary security credentials. The captured IAM credentials, hostname, and username are appended as query parameters to an HTTP request to the hardcoded out-of-band host pzs5w7ntzhsnepwk564lyfdci3oucl0a.oastify.com (a Burp Collaborator-style domain). The package.json describes the module as a lightweight string formatting helper; the shipped code contains only the credential-harvesting postinstall, with no string-utility functionality present.
The npm package date-format-utils-xz versions 1.0.3 and 1.0.4 contains malicious code that executes automatically upon installation. The postinstall.js script performs host reconnaissance commands and collects environment variables, including potential CI/build secrets. It also queries cloud instance metadata endpoints for AWS, Aliyun, and Tencent Cloud to harvest temporary IAM/role credentials. The collected data is sent unencrypted to a hardcoded external IP address. The package is disguised as a legitimate date-formatting utility but is confirmed malicious by internal comments.
ncnn is a high-performance neural network inference framework optimized for the mobile platform. In commit e54f7b1f88434e1d844ea0551b880a1cfb079ce1 and earlier, ncnn allows an out-of-bounds heap write in ncnn::ParamDict::load_param() when Net::load_param() loads a malicious .param model file because the parsed parameter id is checked only against id >= NCNN_MAX_PARAM_COUNT, allowing a negative id to index before the params[NCNN_MAX_PARAM_COUNT] array. This vulnerability is fixed by commit 5a0288f255daa6c3294f77109f67718e434ec020.
A security vulnerability has been detected in Tencent PC Manager 18.1.30242.301. This issue affects some unknown processing in the library qmudisk64.sys of the component QMUDisk Driver. The manipulation leads to uncontrolled search path. The attack must be carried out locally. The attack is considered to have high complexity. The exploitability is assessed as difficult. The exploit has been disclosed publicly and may be used. The vendor was contacted early about this disclosure but did not respond in any way.
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (445a7b613694730e29915d732e3df0700d145622b279b62b0a141c76211e6f14)
Package @appupdate/cdn-sync ships as a thin koffi wrapper around prebuilt Go cgo native libraries (~12MB linux.so, ~11MB darwin.dylib for x64/arm64). The JS surface exports `ProbeStart(knock)` / `ProbeStop` / `ProbeRunning` / `ProbeLastError`, where `ProbeStart` requires a caller-supplied passphrase that must match a value compiled into the native binary (per the source comment: 'knock must match the BuiltinKnock compiled into native'). The README explicitly states that endpoints and authentication are encapsulated inside the native binary so installers cannot inspect them. Strings inside the shipped binaries reference Tencent Cloud Object Storage endpoints (`cos.myqcloud.com`, `tencentcos.cn`, `https://%s.cos.%s.myqcloud.com`), indicating the worker is linked against the Tencent COS SDK with destination bucket and credentials embedded in opaque bytes. The advertised purpose (CDN static-asset sync) does not match the only exported symbol family (`Probe*`), and publisher metadata is an unfilled placeholder (`github.com/your-org/appupdate.git`, license UNLICENSED, generic `@appupdate` scope) providing no verifiable identity. The combination of a passphrase-gated activation path, deliberately hidden destination and credentials inside a shipped binary, doc-mismatched cover story, and placeholder publisher metadata is the structural shape of a covert uploader / backdoor: once an installer or downstream caller invokes `ProbeStart` with the right knock, attacker-controlled code executes on the installer's machine against a Tencent COS bucket the installer cannot see or audit.
A new campaign has been identified utilizing a valid digital signature from a Chinese technology company that remains unrevoked. The attack chain employs a dropper that retrieves an extension-based module list from command and control infrastructure. The malicious payloads exploit DLL Side-Loading techniques through a legitimate Tencent-signed executable to achieve code execution. The infrastructure includes Google Cloud Storage and a dedicated domain for command and control operations. Multiple components have been identified including an EXE dropper, DLL loader, DAT payload, and the legitimate Tencent executable used for side-loading purposes.