Arnold Chan
@slazTrusted contributorCompletionist
2 followers2,100 downloads4,383 copies51 likes26,162 views
7,555 detections
Filters
Last updated
All Time
Detection languages
5,650
1,006
394
169
98
Categories
2,356
1,807
1,233
1,013
924
Platforms
4,191
1,622
1,250
676
676
Products / Services
2,820
1,003
991
720
485
MITRE Techniques
2,087
1,520
1,413
1,070
536
CVEs
38
33
24
22
21
IDS Classtypes
477
179
104
79
52
IDS Protocols
554
156
144
86
18
Detects a multi-stage authentication bypass attempt against LiteLLM MCP endpoints. The rule tracks an initial malformed Authorization Bearer token request to the /mcp/ endpoint, followed by a server response returning a 200 OK status code accompanied by an 'mcp-session-id' header, confirming the exploitation of CVE-2026-59822.
Detects requests to a LiteLLM proxy that have been assigned the 'PROXY_ADMIN' role despite the absence of valid authentication credentials (Authorization headers, JWT tokens, or OAuth2 tokens). This pattern typically indicates that the LiteLLM instance is misconfigured, specifically that a 'master_key' is missing, causing the proxy to default to a permissive mode where every request is treated as having administrative privileges.
Detects the use of the default, hardcoded 'sk-1234' secret key for HS256 JWT signing in LiteLLM proxy logs. This configuration is insecure and allows for the forgery of arbitrary user session tokens.
Detects the use of the default, hardcoded 'sk-1234' secret key for HS256 JWT signing in LiteLLM proxy logs. This configuration is insecure and allows for the forgery of arbitrary user session tokens.
Detects incoming HTTP requests to sensitive LiteLLM endpoints (/mcp and /guardrails) which are typically intended for internal or controlled use. Exposing these to the public internet may allow unauthorized individuals to interact with, enumerate, or exploit the LiteLLM framework's capabilities.
Detects the creation or update of LiteLLM custom code guardrails that include suspicious Python snippets, specifically 'import os' and 'os.popen', which are indicative of attempts to execute arbitrary OS commands via guardrail logic (CVE-2026-59821).
Detects the creation or update of LiteLLM custom code guardrails that include suspicious Python snippets, specifically 'import os' and 'os.popen', which are indicative of attempts to execute arbitrary OS commands via guardrail logic (CVE-2026-59821).
Detects the use of the Windows Management Instrumentation Command-line (WMIC) utility to execute processes, particularly when a remote node target is specified or when using common WMI process creation arguments. This pattern is commonly used by adversaries for lateral movement and remote code execution while attempting to blend into administrative activity. Legitimate management tools have been excluded from the scope to reduce noise.
Detects the use of the Windows Management Instrumentation Command-line (WMIC) utility to execute processes, particularly when a remote node target is specified or when using common WMI process creation arguments. This pattern is commonly used by adversaries for lateral movement and remote code execution while attempting to blend into administrative activity. Legitimate management tools have been excluded from the scope to reduce noise.
Detects the use of PowerShell cmdlets such as Invoke-WmiMethod, Invoke-CimMethod, Get-CimInstance, or Get-WmiObject to interact with the Win32_Process class for process creation on local or remote systems. This behavior is frequently associated with administrative activity but is also commonly used by adversaries for remote command execution and lateral movement.
Page 539 of 756