vLLM is an inference and serving engine for large language models (LLMs). Prior to version 0.14.1, a Server-Side Request Forgery (SSRF) vulnerability exists in the `MediaConnector` class within the vLLM project's multimodal feature set. The load_from_url and load_from_url_async methods obtain and process media from URLs provided by users, using different Python parsing libraries when restricting the target host. These two parsing libraries have different interpretations of backslashes, which allows the host name restriction to be bypassed. This allows an attacker to coerce the vLLM server into making arbitrary requests to internal network resources. This vulnerability is particularly critical in containerized environments like `llm-d`, where a compromised vLLM pod could be used to scan the internal network, interact with other pods, and potentially cause denial of service or access sensitive data. For example, an attacker could make the vLLM pod send malicious requests to an internal `llm-d` management endpoint, leading to system instability by falsely reporting metrics like the KV cache state. Version 0.14.1 contains a patch for the issue.
PyTorch is a Python package that provides tensor computation. Prior to version 2.10.0, a vulnerability in PyTorch's `weights_only` unpickler allows an attacker to craft a malicious checkpoint file (`.pth`) that, when loaded with `torch.load(..., weights_only=True)`, can corrupt memory and potentially lead to arbitrary code execution. Version 2.10.0 fixes the issue.
RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine. In version 0.23.1 and possibly earlier versions, the MinerU parser contains a "Zip Slip" vulnerability, allowing an attacker to overwrite arbitrary files on the server (leading to Remote Code Execution) via a malicious ZIP archive. The MinerUParser class retrieves and extracts ZIP files from an external source (mineru_server_url). The extraction logic in `_extract_zip_no_root` fails to sanitize filenames within the ZIP archive. Commit 64c75d558e4a17a4a48953b4c201526431d8338f contains a patch for the issue.
An Authentication Bypass Using an Alternate Path or Channel vulnerability [CWE-288] vulnerability in Fortinet FortiAnalyzer 7.6.0 through 7.6.5, FortiAnalyzer 7.4.0 through 7.4.9, FortiAnalyzer 7.2.0 through 7.2.11, FortiAnalyzer 7.0.0 through 7.0.15, FortiManager 7.6.0 through 7.6.5, FortiManager 7.4.0 through 7.4.9, FortiManager 7.2.0 through 7.2.11, FortiManager 7.0.0 through 7.0.15, FortiOS 7.6.0 through 7.6.5, FortiOS 7.4.0 through 7.4.10, FortiOS 7.2.0 through 7.2.12, FortiOS 7.0.0 through 7.0.18, FortiProxy 7.6.0 through 7.6.4, FortiProxy 7.4.0 through 7.4.12, FortiProxy 7.2.0 through 7.2.15, FortiProxy 7.0.0 through 7.0.22, FortiWeb 8.0.0 through 8.0.3, FortiWeb 7.6.0 through 7.6.6, FortiWeb 7.4.0 through 7.4.11 may allow an attacker with a FortiCloud account and a registered device to log into other devices registered to other accounts, if FortiCloud SSO authentication is enabled on those devices.
Suricata is a network IDS, IPS and NSM engine. Prior to versions 8.0.3 and 7.0.14, various inefficiencies in xff handling, especially for alerts not triggered in a tx, can lead to severe slowdowns. Versions 8.0.3 and 7.0.14 contain a patch. As a workaround, disable XFF support in the eve configuration. The setting is disabled by default.
Suricata is a network IDS, IPS and NSM engine. While saving a dataset a stack buffer is used to prepare the data. Prior to versions 8.0.3 and 7.0.14, if the data in the dataset is too large, this can result in a stack overflow. Versions 8.0.3 and 7.0.14 contain a patch. As a workaround, do not use rules with datasets `save` nor `state` options.
Suricata is a network IDS, IPS and NSM engine. Starting in version 8.0.0 and prior to version 8.0.3, inefficiency in http1 headers parsing can lead to slowdown over multiple packets. Version 8.0.3 patches the issue. No known workarounds are available.
Suricata is a network IDS, IPS and NSM engine. Prior to version 8.0.3 and 7.0.14, an unsigned integer overflow can lead to a heap use-after-free condition when generating excessive amounts of alerts for a single packet. Versions 8.0.3 and 7.0.14 contain a patch. As a workaround, do not run untrusted rulesets or run with less than 65536 signatures that can match on the same packet.
Suricata is a network IDS, IPS and NSM engine. Starting in version 8.0.0 and prior to version 8.0.3, Suricata can crash with a stack overflow. Version 8.0.3 patches the issue. As a workaround, use default values for `request-body-limit` and `response-body-limit`.
Suricata is a network IDS, IPS and NSM engine. Prior to versions 8.0.3 and 7.0.14, crafted DCERPC traffic can cause Suricata to expand a buffer w/o limits, leading to memory exhaustion and the process getting killed. While reported for DCERPC over UDP, it is believed that DCERPC over TCP and SMB are also vulnerable. DCERPC/TCP in the default configuration should not be vulnerable as the default stream depth is limited to 1MiB. Versions 8.0.3 and 7.0.14 contain a patch. Some workarounds are available. For DCERPC/UDP, disable the parser. For DCERPC/TCP, the `stream.reassembly.depth` setting will limit the amount of data that can be buffered. For DCERPC/SMB, the `stream.reassembly.depth` can be used as well, but is set to unlimited by default. Imposing a limit here may lead to loss of visibility in SMB.