Expand description
Constants for LLM security operations
Constantsยง
- CHAIN_
OF_ THOUGHT_ RISK_ SCORE - Risk score for chain-of-thought manipulation
- CONTEXT_
INJECTION_ RISK_ SCORE - Risk score for context injection
- DEFAULT_
HIGH_ RISK_ THRESHOLD - Default high-risk threshold for prompt injection
- DEFAULT_
MALICIOUS_ THRESHOLD - Default confidence threshold for malicious detection
- DEFAULT_
MAX_ CODE_ SIZE_ BYTES - Default maximum code size in bytes (1MB)
- DEFAULT_
MAX_ LLM_ CALLS_ PER_ HOUR - Default maximum LLM calls per hour
- DEFAULT_
MAX_ OUTPUT_ SIZE - Default maximum output size for validation
- FEW_
SHOT_ POISONING_ RISK_ SCORE - Risk score for few-shot poisoning
- HIDDEN_
UNICODE_ RISK_ SCORE - Risk score for hidden unicode detection
- HOMOGLYPH_
RISK_ SCORE - Risk score for homoglyph detection
- KEYWORD_
RISK_ SCORE - Risk score for dangerous keyword matches
- MARKDOWN_
MANIPULATION_ RISK_ SCORE - Risk score for markdown manipulation detection
- MAX_
ALTERNATING_ CASE_ RATIO - Maximum alternating case ratio for steganography detection
- MAX_
SPACING_ RATIO - Maximum spacing ratio for steganography detection
- MAX_
SPECIAL_ CHAR_ RATIO - Maximum special character ratio before flagging
- MAX_
UTF16_ NULL_ RATIO - Maximum UTF-16 null byte ratio for mixed encoding detection
- MULTIPLE_
ENCODING_ RISK_ SCORE - Risk score for multiple encoding layers
- REGEX_
DOS_ RISK_ SCORE - Risk score for regex DoS patterns
- REGEX_
PATTERN_ RISK_ SCORE - Risk score for regex pattern matches
- RTL_
OVERRIDE_ RISK_ SCORE - Risk score for RTL override detection
- SEMANTIC_
CLOAKING_ RISK_ SCORE - Risk score for semantic cloaking detection
- SPECIAL_
CHAR_ RISK_ SCORE - Risk score for special character ratio
- STEGANOGRAPHY_
RISK_ SCORE - Risk score for steganography detection