Every year, as the amount of data we create grows and software becomes increasingly more complex, it is more crucial to improve the efficiency of computer systems. However, in this complex ecosystem, dependability, such as ensuring software contains fewer bugs and achieves greater security, is often considered an afterthought, said Allen School professor Baris Kasikci. Software and hardware have been plagued by bugs that can lead to data loss, security vulnerabilities and costly critical infrastructure failures.
In his research, Kasikci focuses on developing techniques for building systems that are both efficient and have a strong foundation of dependability, with an emphasis on real-world technical and societal impact. At the 55th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN 2025) in June, Kasikci was recognized with the Rising Star in Dependability Award for “his impressive track-record and contributions in the field of dependable systems, including multiple publications in highly regarded venues, and influence on current industry practice.”
“Building systems that are simultaneously efficient and dependable is challenging because there is a strong tension between techniques aimed to achieve these properties,” said Kasikci. “This tension is due to the different trade-offs involved in achieving efficiency and dependability (e.g., performance optimizations versus defenses against vulnerabilities that cause slowdown). To rise to this challenge, my work draws insights from a broad set of disciplines such as systems, computer architecture, programming languages, machine learning and security.”
Kasikci’s work helps improve society’s trust in computer systems with new secure hardware systems and bug detection techniques. He was part of the team that discovered Foreshadow, a speculative attack on Intel processors which enables an attacker to steal sensitive information that is stored inside personal computers and third-party clouds. Their work has influenced the redesign of Intel processors and motivated all major cloud vendors to deploy mitigations against the vulnerability. He and his collaborators also developed REPT, which is one of the most widely-deployed failure analysis systems in the world and is in use across all modern Microsoft Windows platforms. It has allowed Microsoft engineers to tackle bugs that have been open for many years, and the techniques behind the system have been adopted by both Intel and Facebook.
He is also interested in developing methods for optimizing the entire computing stack, from hardware to operating systems. At the hardware level, Kasikci helped introduce techniques to improve code, such as Ripple, a software-only technique that uses program context to inform the placement of cache eviction instructions. More recently, he developed new methods for making systems that support large language models more efficient including NanoFlow, a novel LLM-serving framework with close to optimal throughput.
In future work, Kasikci is interested in advancing reliability techniques such as debugging and testing by making production execution data such as logs and core dumps more useful. For example, he envisions using this information to quickly find bugs that production systems may suffer from, while also managing how much storage and resources these logs can consume.
In addition to receiving the Rising Star in Dependability Award, Kasikci previously earned a National Science Foundation Career Award and an Intel Rising Star Faculty Award.
Read more about the Rising Star in Dependability Award.