| Name |
Type |
Description |
Notes |
| address |
str |
Function's virtual address, hex-encoded |
|
| categories |
List[str] |
Distinct execution categories evidenced by this function |
|
| confidence |
str |
High when a direct name match was found, medium when the function only calls into execution APIs |
|
| direct_matches |
List[ExecutionDirectMatch] |
Matches against the function's own name |
[optional] |
| evidence_count |
int |
Total number of direct matches and execution calls |
|
| executes |
bool |
Whether this function evidences executing code rather than only supporting it |
|
| execution_calls |
List[ExecutionCall] |
Matches against names this function calls |
[optional] |
| function_id |
int |
ID of the function the finding was reported in |
|
| function_name |
str |
Name of the function the finding was reported in |
|
| function_size |
int |
Size of the function in bytes |
|
| sources |
List[str] |
Distinct execution sources evidenced by this function |
|
| verification |
ExecutionVerification |
LLM verdict checking this finding against its decompilation. Present only when the run verified this finding. |
[optional] |
from revengai.models.execution_finding import ExecutionFinding
# TODO update the JSON string below
json = "{}"
# create an instance of ExecutionFinding from a JSON string
execution_finding_instance = ExecutionFinding.from_json(json)
# print the JSON string representation of the object
print(ExecutionFinding.to_json())
# convert the object into a dict
execution_finding_dict = execution_finding_instance.to_dict()
# create an instance of ExecutionFinding from a dict
execution_finding_from_dict = ExecutionFinding.from_dict(execution_finding_dict)
[Back to Model list] [Back to API list] [Back to README]