Digital advertising remains the backbone of the modern internet economy, yet it is perpetually undermined by ad-click fraud. Automated botnets and organized click farms generate deceptive traffic that drains budgets and erodes ecosystem trust. To counter this, this dissertation proposes a holistic remediation strategy combining traffic-level and infrastructure-level interventions through three novel systems. The first, ECHO, is a proactive remediation tool that hijacks malware update mechanisms to deliver "cleanup" payloads. In tests on 702 Android malware samples, ECHO successfully neutralized 523 infections, effectively shrinking the available botnet pool. Building on this, GRAPHIUS provides an automated pipeline to detect the abuse of residential IP (ResIP) proxies. By utilizing graph-based learning to identify topological signatures of fraud, it achieved 98.3% accuracy across real-world datasets from major proxy providers and search engines. Finally, COSEC offers a semantic detection layer that quantifies an "incoherence index" for search sessions. By analyzing temporal and behavioral features, COSEC distinguishes legitimate users from fraudulent patterns with 95.8% precision and 92.4% recall. Together, these systems provide an end-to-end defense—dismantling botnet infrastructure, flagging proxy abuse, and filtering fraudulent traffic—to secure the digital advertising landscape against its most persistent and costly threats.