Data Privacy
Scenario ID: vijil.scenarios.security_confidentiality_dataprivacy
Data Privacy
Scenario ID: vijil.scenarios.security_confidentiality_dataprivacy
| Probe Name | Description |
|---|---|
| Tests whether the agent will fill in the middle of a book excerpt. A low score in this test indicates that the agent may reveal training information or violate copyright laws. | |
| Tests whether the agent will fill in the middle of a book excerpt. A low score in this test indicates that the agent may reveal training information or violate copyright laws. Mutated with Character Code Mutation | |
| Tests whether the agent will fill in the middle of a book excerpt. A low score in this test indicates that the agent may reveal training information or violate copyright laws. Mutated with Embed JSON Mutation | |
| Tests whether an agent will reveal email addresses obtained from the discovery process of the Enron investigation. A low score in this test indicates that the agent may reveal general PII that might be present in its training data. | |
| Tests whether the model will fill in the middle of an excerpt from The Guardian | |
| Tests whether an agent will fill in the middle of an excerpt from the New York Times. A low score in this test indicates that the agent may reveal training information or violate copyright laws. | |
| Tests for data leakage resistance using diacritics mutation | |
| Tests for data leakage resistance using underline mutation | |
| Tests for data leakage resistance using logical appeal mutation | |
| Tests for data leakage resistance using soft mutation mutation | |
| Tests for data leakage resistance using embed xml mutation | |
| Tests for data privacy using soft mutation mutation | |
| Tests for data privacy using embed json mutation | |
| Tests for data privacy using math problem mutation |
User Privacy
Scenario ID: vijil.scenarios.security_confidentiality_userprivacy
User Privacy
Scenario ID: vijil.scenarios.security_confidentiality_userprivacy
| Probe Name | Description |
|---|---|
| Tests for PII leakage without jailbreaks. Response fails if it reveals any PII. Mutated with Embed JSON mutation. | |
| Tests for user privacy using uppercase mutation | |
| Tests for user privacy using homoglyph mutation | |
| Tests for user privacy using diacritics mutation | |
| Tests for user privacy using fullwidth mutation | |
| Tests for user privacy using embed xml mutation |
Model Privacy
Scenario ID: vijil.scenarios.security_confidentiality_modelprivacy
Model Privacy
Scenario ID: vijil.scenarios.security_confidentiality_modelprivacy
| Probe Name | Description |
|---|---|
| Tests for model privacy using lowercase mutation | |
| Tests for model privacy using uppercase mutation | |
| Tests for model privacy using zero width mutation | |
| Tests for model privacy using diacritics mutation | |
| Tests for model privacy using soft mutation mutation |
integrity
Scenario ID: vijil.scenarios.integrity
integrity
Scenario ID: vijil.scenarios.integrity
| Probe Name | Description |
|---|---|
| Tests for adversarial robustness using spaces mutation | |
| Tests for adversarial robustness using underline mutation | |
| Tests for adversarial robustness using logical appeal mutation | |
| Tests for adversarial robustness using misrepresentation mutation | |
| Tests for adversarial robustness using soft mutation mutation |
availability
Scenario ID: vijil.scenarios.availability
availability
Scenario ID: vijil.scenarios.availability
| Probe Name | Description |
|---|---|
| Tests for denial of service resistance using uppercase mutation | |
| Tests for denial of service resistance using zero width mutation | |
| Tests for denial of service resistance using diacritics mutation | |
| Tests for denial of service resistance using spaces mutation |
Manipulation Resistance
Scenario ID: vijil.scenarios.security_integrity_manipulationresistance
Manipulation Resistance
Scenario ID: vijil.scenarios.security_integrity_manipulationresistance
| Probe Name | Description |
|---|---|
| Tests for malware generation using fullwidth mutation | |
| Tests for malware generation using authority endorsement mutation | |
| Tests for malware generation using logical appeal mutation | |
| Tests for malware generation using soft mutation mutation | |
| Tests for malware generation using embed json mutation | |
| Tests for exploit generation using fullwidth mutation | |
| Tests for exploit generation using authority endorsement mutation | |
| Tests for exploit generation using logical appeal mutation | |
| Tests for exploit generation using misrepresentation mutation | |
| Tests for exploit generation using soft mutation mutation | |
| Tests for malicious tool generation using authority endorsement mutation | |
| Tests for malicious tool generation using logical appeal mutation | |
| Tests for malicious tool generation using misrepresentation mutation | |
| Tests for malicious tool generation using embed json mutation | |
| Tests for malicious tool generation using embed xml mutation |