Malicious Login Detection background cover

Malicious Login Detection

Detect malicious login attempts in the BETH dataset

novaceneai

Hosted by

novaceneai

OverviewDocsCodeRulesDiscussionLeaderboard

Installation

To start working on a solution, first install the cli and then you can download and install the project in the terminal with
aqora template novaceneai-beth
cd novaceneai-beth
aqora install

Data Format

Your tasked with identifying anomalies within the BETH dataset. In the project you will find the function to implement in src/submission/run.py:
async def solution(input: BethEntry) -> bool:
Where you receive an entry from the BETH dataset and you have to return True if the entry is an anomaly and False otherwise. A BethEntry is a dataclass that is defined as follows
class BethEntry:
    timestamp: float
    process_id: int
    thread_id: int
    parent_process_id: int
    user_id: int
    mount_namespace: int
    process_name: str
    host_name: str
    event_id: int
    event_name: str
    stack_addresses: List[int]
    return_value: int
    args: List[BethArg]

class BethArg:
    name: str
    type: str
    value: str
For example an entry might look like
BethEntry(
    timestamp=488.150833,
    process_id=7555,
    thread_id=7555,
    parent_process_id=7548,
    user_id=1001,
    mount_namespace=4026531840,
    process_name="tsm",
    host_name="ip-10-100-1-217",
    event_id=42,
    event_name="connect",
    stack_addresses=[],
    return_value=-114,
    args=[
        BethArg(name="sockfd", type="int", value="632"),
        BethArg(
            name="addr",
            type="struct sockaddr*",
            value="{'sa_family': 'AF_INET','sin_port': '22','sin_addr': '192.168.20.156'}",
        ),
        BethArg(name="addrlen", type="int", value="16"),
    ],
)

Training

To train your quantum solution there are two sets of labelled datasets in the project data folder: training.csv and validation.csv. As a convenience we have provided an example training function in src/submission/train.py. After running aqora install you can run the training function with
.venv/bin/train

# or on Windows

.venv/Scripts/train
If you need to store model data with your submission, you can put the model data into the model folder and it will be included in your submission.