Review of malicious code detection in data mining applications: challenges, algorithms, and future direction

Review of malicious code detection in data mining applications: challenges, algorithms, and future direction

연구 분야: Databases

학회: Cluster Computing

In an era where machine learning critically underpins business operations, detecting vulnerabilities introduced by malicious code has become increasingly essential. Although prior research has extensively explored malicious code within machine learning algorithms, a targeted analysis specifically designed to identify and address these threats remains necessary. This paper presents an exhaustive literature review, focusing on the key processes of insertion, recognition, decision-making, and selection of malicious codes. We aim to uncover architectural weaknesses in data mining applications that amplify system vulnerabilities. Leveraging an integrative review covering publications from 2008 to 2024, we synthesize insights from a diverse array of academic and digital sources, examining 167 pertinent articles. This rigorous approach reveals the nuanced effects of malicious code on feature selection algorithms, crucial for maintaining data integrity. Our findings indicate that malicious code can significantly disrupt various sectors, including industrial, telecommunications, and biological data mining, adversely affecting clustering, classification, and regression algorithms. However, an encouraging outcome is observed in advanced feature selection algorithms that demonstrate resilience by effectively filtering out irrelevant data inputs. The paper concludes with a strong call for the development of sophisticated detection methods, which are vital for mitigating the growing risks associated with malicious code. It stresses the importance of proactive algorithm identification and classification to preserve the efficacy of data mining. Current challenges in accurately classifying machine learning algorithms raise concerns about data privacy, security, and potential biases. Ongoing research is crucial for improving data interoperability and algorithm transparency, thereby strengthening the defense mechanisms of machine learning applications against the complex and evolving landscape of cyber threats.