CWE-76
Improper Neutralization of Equivalent Special Elements
BaseDraftLikelihood: High
Description
The product correctly neutralizes certain special elements, but it improperly neutralizes equivalent special elements.
The product may have a fixed list of special characters it believes is complete. However, there may be alternate encodings, or representations that also have the same meaning. For example, the product may filter out a leading slash (/) to prevent absolute path names, but does not account for a tilde (~) followed by a user name, which on some *nix systems could be expanded to an absolute pathname. Alternately, the product might filter a dangerous "-e" command-line switch when calling an external program, but it might not account for "--exec" or other switches that have the same semantics.
Hierarchy (View 1000)
Parents
Children
none
CVEs mapped to this weakness (1)
| CVE | Sev | Risk | CVSS | EPSS | KEV | Published | Description |
|---|---|---|---|---|---|---|---|
| CVE-2024-34359 | Cri | 0.60 | 9.6 | 0.62 | May 14, 2024 | llama-cpp-python is the Python bindings for llama.cpp. `llama-cpp-python` depends on class `Llama` in `llama.py` to load `.gguf` llama.cpp or Latency Machine Learning Models. The `__init__` constructor built in the `Llama` takes several parameters to configure the loading and running of the model. Other than `NUMA, LoRa settings`, `loading tokenizers,` and `hardware settings`, `__init__` also loads the `chat template` from targeted `.gguf` 's Metadata and furtherly parses it to `llama_chat_format.Jinja2ChatFormatter.to_chat_handler()` to construct the `self.chat_handler` for this model. Nevertheless, `Jinja2ChatFormatter` parse the `chat template` within the Metadate with sandbox-less `jinja2.Environment`, which is furthermore rendered in `__call__` to construct the `prompt` of interaction. This allows `jinja2` Server Side Template Injection which leads to remote code execution by a carefully constructed payload. |