Detection of ARP Spoofing with Optimized False Alarm Using Deep Learning Based Absolute Thresholding

The growth of the internet has transformed lives worldwide, enabling new opportunities for communication, information access, and business. However, this expansion has also created avenues for cybercriminals to exploit security flaws, compromise sensitive information, and disrupt online activities. Online fraud, identity theft, and ransomware attacks have become commonplace, posing significant threats to individuals and organizations. The need for IDS is crucial for detecting and preventing threats in real-time. NIDS monitor network traffic to detect and prevent attacks, serving as a watchdog that alerts administrators to potential threats and helps keep networks safe and secure. Traditional signature-based detection methods have become ineffective against advanced cyber threats. Fortunately, recent developments in ML and DL have significantly enhanced NIDS performance, enabling them to identify emerging threats with greater precision and reduce false alarms. By observing other papers Vanlalruata Hnamte [20] and Md. Alamin Talukder <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[7]</sup> approach achieving an impressive 99.99% accuracy in detecting threats. Traditional NIDS struggled with class imbalance, computational complexity, and generalization issues, resulting to biased models and inefficiency. The proposed Deep Learning-based Static Thresholding (DLST) method addresses these challenges by integrating SMOTE for data balancing and dropout layers to prevent overfitting, improving detection accuracy and reducing false positives.