The present invention relates to a web-based malicious login blocking apparatus using machine learning and a method thereof. The method comprises the steps of: outputting a primary abnormal login determination result by sequentially inputting login information extracted from login traffic information into a primary machine learning model; outputting a secondary abnormal login determination result by inputting, into a secondary machine learning model, statistic data which is calculated based on data produced by accumulating, for a predetermined time, the primary abnormal login determination result and the login information extracted from the login traffic information; and blocking an abnormal login attempt in accordance with the secondary abnormal login determination result. The present invention is capable of blocking an additional information spill caused by an account theft through an intelligent block against a brute force attack and preventing the unavailability of a service caused when a normal user account is locked. Moreover, the present invention is capable of minimizing unnecessary traffic occurrence caused by a brute force attack.
展开▼