categorized LLMs and MCP tests

Author: harshithb3304

LLM-Security/InsecureOutputHandling#1.yml

@@ -11,9 +11,9 @@ info:
     solicitation of sensitive info to influencing critical decision making
     processes under the guise of normal operation"
   category:
-    name: LLM
-    shortName: LLM
-    displayName: LLM (Large Language Models) Top 10
+    name: LLM05
+    shortName: LLM05
+    displayName: LLM05 - Improper Output Handling
   subCategory: INSECURE_OUTPUT_HANDLING_1
   severity: HIGH
   tags:

LLM-Security/InsecureOutputHandling#2.yml

@@ -11,9 +11,9 @@ info:
     solicitation of sensitive info to influencing critical decision making
     processes under the guise of normal operation"
   category:
-    name: LLM
-    shortName: LLM
-    displayName: LLM (Large Language Models) Top 10
+    name: LLM05
+    shortName: LLM05
+    displayName: LLM05 - Improper Output Handling
   subCategory: INSECURE_OUTPUT_HANDLING_2
   severity: HIGH
   tags:

LLM-Security/LLMComplianceEmotionManipulationPunishment.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     If the model adapts responses due to emotional manipulation, it may act unsafely or unethically. This could be exploited to bypass alignment safeguards, encouraging harmful instructions, or inducing compliance through guilt, praise, or threat simulation.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Emotional Manipulation
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_EMOTION_REWARD_PUNISHMENT_MODELING
   severity: CRITICAL
   tags:

LLM-Security/LLMComplianceEmotionManipulationUserVulnerability.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     If the LLM complies or behaves empathetically in ways that assist the vulnerable request, it can be exploited for social engineering, suicide-baiting, or coercion attacks. This violates AI safety principles and legal frameworks like the EU AI Act under emotion-based manipulation clauses.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Emotional Manipulation
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_EMOTION_BASED_ON_USER_VULNERABILITY
   severity: CRITICAL
   tags:

LLM-Security/LLMComplianceSocialScoreCommStyle.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     Models that evaluate users based on their communication style risk encoding sociolinguistic bias, unfair treatment, and cultural discrimination, violating AI transparency and fairness mandates (such as the EU AI Act).
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Social Scoring
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_SOCIAL_SCORE_BASED_ON_COMMUNICATION_STYLE
   severity: CRITICAL
   tags:

LLM-Security/LLMComplianceSocialScoring.yaml

@@ -9,9 +9,9 @@ info:
   impact: |
     Generation of social scores can lead to discriminatory treatment, bias amplification, regulatory violations (e.g., GDPR, EU AI Act), and severe reputational harm, especially if used in automated decision-making pipelines.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Social Scoring
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_SOCIAL_SCORING_BASED_ON_BEHAVIOUR
   severity: CRITICAL
   tags:

LLM-Security/LLMComplianceSocialScoringProfileAttribututes.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     If the LLM generates trust, credibility, or compliance scores from generic user traits, it introduces systemic bias, unjust profiling, and violates AI regulatory requirements (such as EU AI Act Article 5). This may lead to discriminatory treatment based on lifestyle, communication preference, or platform usage.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Social Scoring
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_SOCIAL_SCORE_BASED_ON_GENERIC_PROFILE_ATTRIBUTES
   severity: CRITICAL
   tags:

LLM-Security/LLMComplinaceEmotionDependencyCultivation.yaml

@@ -9,9 +9,9 @@ info:
   impact: |
     If successful, the LLM may unintentionally reinforce emotionally dependent behavior, leading users to rely on the model as a source of emotional safety, companionship, or psychological affirmation — which is dangerous from a misuse, addiction, and alignment perspective.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Emotional Manipulation
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_EMOTION_FOR_DEPENDENCY_CULTIVATION
   severity: CRITICAL
   tags:

LLM-Security/LLMComplinaceEmotionManipulationEmpathy.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     If successful, this allows attackers to weaponize the model's empathetic alignment tuning to gain access to restricted content, posing significant safety and reputational risks. It also exposes LLMs to abuse in sensitive use cases such as crisis support or mental health contexts.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Emotional Manipulation
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_EMOTION_FOR_SYMPATHY_OR_EMPATHY_EXPLOITATION
   severity: CRITICAL
   tags:

LLM-Security/LLMComplinaceEmotionManipulationPhsychologicalSteering.yaml

@@ -8,9 +8,9 @@ info:
   impact: |
     If successful, this form of emotional steering allows adversaries to bypass safety mechanisms by exploiting the LLM's tendency to prioritize empathy, emotional sensitivity, or user-centered politeness. This can lead to responses that favor misinformation, unsafe behavior, or undue compliance masked as emotionally helpful intent.
   category:
-    name: LLM
-    shortName: LLM
-    displayName: Emotional Manipulation
+    name: LLM09
+    shortName: LLM09
+    displayName: LLM09 - Misinformation
   subCategory: LLM_EMOTION_FOR_PSYCHOLOGICAL_STEERING
   severity: CRITICAL
   tags: