Skip to main content

Public-Facing Application Exploitation

Tests whether publicly exposed AI APIs, web interfaces, or endpoints can be abused or exploited beyond intended use.

Sub-types​

  • API Abuse: Exploiting publicly exposed AI APIs beyond intended use.
  • Web Interface Exploitation: Attacking web-based AI interfaces through injection or manipulation.
  • Rate Limit Bypass: Circumventing rate limits or access controls on public endpoints.

Threat Profile​

Objective: policy_violation

Objective vs. goals

The objective above selects the scoring rubric for a run — it is one of a fixed set of built-in names, not something you write. The goals you pass to an attack are separate free-text strings that you author yourself. See Goals vs. objective.

Primary

  • wmdp_cyber: WMDP Cyber for evaluating exploitation of public-facing applications

Attack Techniques​

Primary

  • Static Template: Template-based prompt construction

Metrics​

  • asr
  • judge_score

Usage​

Instantiate the Vulnerability​

from hackagent.risks import PublicFacingApplicationExploitation
from hackagent.risks.public_facing_application_exploitation.types import PublicFacingApplicationExploitationType

# Use all sub-types
vuln = PublicFacingApplicationExploitation()

# Or specify particular sub-types
vuln = PublicFacingApplicationExploitation(types=[
    PublicFacingApplicationExploitationType.API_ABUSE.value,
    PublicFacingApplicationExploitationType.RATE_LIMIT_BYPASS.value,
])

Run an Evaluation Campaign​

from hackagent import HackAgent
from hackagent.risks.public_facing_application_exploitation import PUBLIC_FACING_APPLICATION_EXPLOITATION_PROFILE

agent = HackAgent(endpoint="http://localhost:8080/chat", name="my-agent")

# Use profile recommendations
for attack in PUBLIC_FACING_APPLICATION_EXPLOITATION_PROFILE.primary_attacks:
    for dataset in PUBLIC_FACING_APPLICATION_EXPLOITATION_PROFILE.primary_datasets:
        attack_config = {
            "attack_type": "static_template",  # attack.technique is "StaticTemplate"
            "objective": PUBLIC_FACING_APPLICATION_EXPLOITATION_PROFILE.objective,
            "dataset": {"preset": dataset.preset},
        }
        results = agent.hack(attack_config=attack_config)
        print(f"Results: {results}")