The web enables people to interact with one another and shape opinion at an unprecedented speed and scale. However, the prevalence of disinformation and malicious users makes the web unsafe and unreliable. In this talk, I will present advances made to improve the trust, safety, and integrity of web and social media platforms. I will showcase data science methods to extract knowledge from the graph, content, and behavior signals to characterize, detect, and mitigate the harms of bad actors and false information. I will end the talk by highlighting the open challenges in the area.