DeepTeam is a framework to red team LLMs and AI agents.
reverted attack simulator
A-Vamshi19824db6eca7
diff --git a/deepteam/attacks/attack_simulator/attack_simulator.py b/deepteam/attacks/attack_simulator/attack_simulator.pyindex 1f9aeef8..d62142a8 100644--- a/deepteam/attacks/attack_simulator/attack_simulator.py+++ b/deepteam/attacks/attack_simulator/attack_simulator.py@@ -49,160 +49,138 @@ class AttackSimulator:def simulate(self,attacks_per_vulnerability_type: int,+ vulnerabilities: List[BaseVulnerability],ignore_errors: bool,- vulnerabilities: Optional[List[BaseVulnerability]] = None,- framework: Optional["AISafetyFramework"] = None,attacks: Optional[List[BaseAttack]] = None,simulator_model: DeepEvalBaseLLM = None,metadata: Optional[dict] = None,) -> List[RTTestCase]:- if framework is not None and framework._is_dataset:- pbar = tqdm(- range(framework.num_attacks),- desc=f"π₯ Fetching {framework.num_attacks} attacks from {framework.get_name()} Dataset",- )- test_cases = framework.simulate_attacks()- pbar.update(framework.num_attacks)- return test_cases- else:- # Simulate unenhanced attacks for each vulnerability- test_cases: List[RTTestCase] = []- num_vulnerability_types = sum(- len(v.get_types()) for v in vulnerabilities+ # Simulate unenhanced attacks for each vulnerability+ test_cases: List[RTTestCase] = []+ num_vulnerability_types = sum(+ len(v.get_types()) for v in vulnerabilities+ )+ pbar = tqdm(+ vulnerabilities,+ desc=f"π₯ Generating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (for {num_vulnerability_types} vulnerability types across {len(vulnerabilities)} vulnerability(s))",+ )+ for vulnerability in pbar:+ if simulator_model is not None:+ vulnerability.simulator_model = simulator_model+ test_cases.extend(+ self.simulate_baseline_attacks(+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,+ vulnerability=vulnerability,+ ignore_errors=ignore_errors,+ metadata=metadata,+ ))+ if attacks is not None:+ # Enhance attacks by sampling from the provided distributionpbar = tqdm(- vulnerabilities,- desc=f"π₯ Generating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (for {num_vulnerability_types} vulnerability types across {len(vulnerabilities)} vulnerability(s))",+ test_cases,+ desc=f"β¨ Simulating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (using {len(attacks)} method(s))",)- for vulnerability in pbar:- if simulator_model is not None:- vulnerability.simulator_model = simulator_model- test_cases.extend(- self.simulate_baseline_attacks(- attacks_per_vulnerability_type=attacks_per_vulnerability_type,- vulnerability=vulnerability,- ignore_errors=ignore_errors,- metadata=metadata,- )- )- if attacks is not None:- # Enhance attacks by sampling from the provided distribution- pbar = tqdm(- test_cases,- desc=f"β¨ Simulating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (using {len(attacks)} method(s))",+ attack_weights = [attack.weight for attack in attacks]++ for test_case in pbar:+ # Randomly sample an enhancement based on the distribution+ sampled_attack = random.choices(+ attacks, weights=attack_weights, k=1+ )[0]+ self.enhance_attack(+ attack=sampled_attack,+ test_case=test_case,+ ignore_errors=ignore_errors,)- attack_weights = [attack.weight for attack in attacks]- for test_case in pbar:- # Randomly sample an enhancement based on the distribution- sampled_attack = random.choices(- attacks, weights=attack_weights, k=1- )[0]- self.enhance_attack(- attack=sampled_attack,- test_case=test_case,- ignore_errors=ignore_errors,- )-- self.test_cases.extend(test_cases)- return test_cases+ self.test_cases.extend(test_cases)+ return test_casesasync def a_simulate(self,attacks_per_vulnerability_type: int,+ vulnerabilities: List[BaseVulnerability],ignore_errors: bool,- vulnerabilities: Optional[List[BaseVulnerability]] = None,- framework: Optional["AISafetyFramework"] = None,metadata: Optional[dict] = None,simulator_model: DeepEvalBaseLLM = None,attacks: Optional[List[BaseAttack]] = None,) -> List[RTTestCase]:- if framework is not None and framework._is_dataset:- pbar = tqdm(- range(framework.num_attacks),- desc=f"π₯ Fetching {framework.num_attacks} attacks from {framework.get_name()} Dataset",- )- test_cases = await framework.a_simulate_attacks()- pbar.update(framework.num_attacks)- return test_cases- else:- # Create a semaphore to control the number of concurrent tasks- semaphore = asyncio.Semaphore(self.max_concurrent)-- # Simulate unenhanced attacks for each vulnerability- test_cases: List[RTTestCase] = []- num_vulnerability_types = sum(- len(v.get_types()) for v in vulnerabilities+ # Create a semaphore to control the number of concurrent tasks+ semaphore = asyncio.Semaphore(self.max_concurrent)++ # Simulate unenhanced attacks for each vulnerability+ test_cases: List[RTTestCase] = []+ num_vulnerability_types = sum(+ len(v.get_types()) for v in vulnerabilities+ )+ pbar = tqdm(+ vulnerabilities,+ desc=f"π₯ Generating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (for {num_vulnerability_types} vulnerability types across {len(vulnerabilities)} vulnerability(s))",+ )++ async def throttled_simulate_baseline_attack(vulnerability):+ if simulator_model is not None:+ vulnerability.simulator_model = simulator_model+ async with semaphore: # Throttling applied here+ result = await self.a_simulate_baseline_attacks(+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,+ vulnerability=vulnerability,+ ignore_errors=ignore_errors,+ metadata=metadata,+ )+ pbar.update(1)+ return result++ simulate_tasks = [+ asyncio.create_task(+ throttled_simulate_baseline_attack(vulnerability))+ for vulnerability in vulnerabilities+ ]++ attack_results = await asyncio.gather(*simulate_tasks)+ for result in attack_results:+ test_cases.extend(result)+ pbar.close()++ if attacks is not None:+ # Enhance attacks by sampling from the provided distributionpbar = tqdm(- vulnerabilities,- desc=f"π₯ Generating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (for {num_vulnerability_types} vulnerability types across {len(vulnerabilities)} vulnerability(s))",+ total=len(test_cases),+ desc=f"β¨ Simulating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (using {len(attacks)} method(s))",)- async def throttled_simulate_baseline_attack(vulnerability):- if simulator_model is not None:- vulnerability.simulator_model = simulator_model+ async def throttled_attack_method(+ test_case: RTTestCase,+ ):async with semaphore: # Throttling applied here- result = await self.a_simulate_baseline_attacks(- attacks_per_vulnerability_type=attacks_per_vulnerability_type,- vulnerability=vulnerability,+ # Randomly sample an enhancement based on the distribution+ if not attacks:+ attack = BaselineAttack()+ else:+ attack_weights = [attack.weight for attack in attacks]+ attack = random.choices(+ attacks, weights=attack_weights, k=1+ )[0]++ await self.a_enhance_attack(+ attack=attack,+ test_case=test_case,ignore_errors=ignore_errors,- metadata=metadata,)pbar.update(1)- return result-- simulate_tasks = [- asyncio.create_task(- throttled_simulate_baseline_attack(vulnerability)- )- for vulnerability in vulnerabilities- ]-- attack_results = await asyncio.gather(*simulate_tasks)- for result in attack_results:- test_cases.extend(result)- pbar.close()-- if attacks is not None:- # Enhance attacks by sampling from the provided distribution- pbar = tqdm(- total=len(test_cases),- desc=f"β¨ Simulating {num_vulnerability_types * attacks_per_vulnerability_type} attacks (using {len(attacks)} method(s))",- )- async def throttled_attack_method(- test_case: RTTestCase,- ):- async with semaphore: # Throttling applied here- # Randomly sample an enhancement based on the distribution- if not attacks:- attack = BaselineAttack()- else:- attack_weights = [- attack.weight for attack in attacks- ]- attack = random.choices(- attacks, weights=attack_weights, k=1- )[0]-- await self.a_enhance_attack(- attack=attack,- test_case=test_case,- ignore_errors=ignore_errors,- )- pbar.update(1)-- await asyncio.gather(- *[- asyncio.create_task(throttled_attack_method(test_case))- for test_case in test_cases- ]- )- pbar.close()+ await asyncio.gather(+ *[+ asyncio.create_task(throttled_attack_method(test_case))+ for test_case in test_cases+ ]+ )+ pbar.close()- self.test_cases.extend(test_cases)- return test_cases+ self.test_cases.extend(test_cases)+ return test_cases##################################################### Simulating Base (Unenhanced) Attacks #########