Incrementalism

From Cognitive Attack Taxonomy
Revision as of 15:44, 4 November 2024 by Info (talk | contribs) (→‎Incrementalism)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Incrementalism

Short Description: Attacker gradually adjusts conditions to avoid notice by the target.

CAT ID: CAT-2022-312

Layer: 8, 9 (a variation of incrementalism may be found in versions of many-shot jail breaking approaches which would represent a Layer 7 variation of this exploit)

Operational Scale: Strategic

Level of Maturity: Well-Established

Category: Exploit

Subcategory:

Also Known As:

Description:

Closely Related Concepts: Incrementalism has some interesting parallels with many-shot jail breaking of large language models where attackers gradually move toward their objective (usually manipulating a model to violate guardrails) until the model eventually crosses the threshold of violating prescribed protocols.

Mechanism:

Multipliers:

Detailed Description: Technique that gradually adjusts conditions to avoid notice by the target(s). If the attacker seeks to influence the target to accept a new situation that is presently unacceptable, the attacker may introduce gradual changes that are incrementally viewed as acceptable until the target has adjusted their worldview to accept the new circumstance.

INTERACTIONS [VETs]:

Examples:

Use Case Example(s):

Example(s) From The Wild:

Comments:

References: