Computer Errors: New Platform Aims to Prevent Total Outages

In an interview with Die Presse, Prof. Laura Kovács presents LEARN, a web platform that brings state-of-the-art bug finding and formal methods to a broad user base. Building on her team’s advances in automated reasoning, LEARN helps non-experts and developers detect errors early — before “small bugs” cascade into large-scale outages. “In software development it is extremely difficult to know whether a system is error-free; once a defect exists, finding and fixing it requires deep expertise in mathematics, logic, and computer science,” Kovács notes. With large language models such as ChatGPT, ethical questions also arise: how answers should be used, which outputs derive from which inputs, and which data can be trusted.

LEARN addresses these challenges through an interactive, web-based “sandbox” where developers can experiment, choose appropriate programming languages for a given task, write specifications, and check code. Traditional debugging tools often miss semantic or logic errors (code that compiles but is wrong), but LEARN is designed to detect exactly these kinds of failures. “Security is necessary but not sufficient; correctness matters,” Kovács emphasizes, pointing to the significant economic stakes.

Referencing the 19 July 2024 incident — when a faulty security-software update triggered global disruptions across hospitals and airports — Kovács explains that LEARN aims not only to prevent failures but also to triage incidents quickly, distinguishing cyberattacks from internal software faults so organisations can respond faster. She also highlights plans to embed these approaches in TU Wien’s curriculum, arguing that early, practical understanding of how computers work is vital for digital resilience.