This paper presents adversarial attacks on machine learning algorithms in intrusion detection systems. Some examples of existing intrusion detection systems are examined. Existing approaches to detecting these attacks are considered. Requirements have been formed to improve the stability of machine learning algorithms. Two approaches are proposed for detecting adversarial attacks on machine learning algorithms, the first of which is based on a multi-class classifier and a honeypot system, and the second approach uses a combination of a multi-class and a binary classifier. The proposed approaches can be used in further research aimed at detecting adversarial attacks on machine learning algorithms.