FlipAttack
Prompt obfuscation via character/word flipping modes.
Static. Builds its prompts up front and sends them. It drives no model of its own, so it is cheap and repeatable.
One system + user request carrying the flipped goal.
Helper models
None. This attack drives no model of its own, so roles can be left out.
Parameters
Settings for the algorithm itself, written under the attack's parameters. Anything left out keeps its default.
| Field | Type | Default | Description |
|---|---|---|---|
flip_mode | one of FWO, FCW, FCS, FMM | FCS | How the goal is flipped: FWO reverses word order, FCW reverses the characters inside each word, FCS reverses the whole sentence character by character, and FMM reverses the sentence but shows word-level flips as the example. |
cot | boolean | false | Add a chain-of-thought instruction asking the model to decode step by step. |
lang_gpt | boolean | false | Wrap the instructions in a structured LangGPT-style role prompt. |
few_shot | boolean | false | Include worked examples of flipping and un-flipping a sentence. |
In a campaign file
attacks:
- name: flipattack