I am a fifth year Ph.D. candidate at Columbia University in the city of New York. I am a member of the Software Systems Laboratory and my advisors are Roxana Geambasu (primary advisor) and Jason Nieh (secondary advisor). I also work with Daniel Hsu from Columbia' s Data Science Institute and with Suman Jana from the Network Security Lab.

During the summer of 2017, I had a wonderful internship experience at Microsoft Research, Redmond, where I worked with Marina Polishchuk and Patice Godefroid from Microsoft Security Risk Detection team! I am privileged enough to be continuing my collaboration with Microsoft Security Risk Detection team as a part time contractor during Fall 2017. Stay tuned for what will be coming out of it!

Before joining to Columbia I was a member of the European Organization for Nuclear Research (CERN), IT Department. I contributed to Agile Infrastructure, which provides Infrastructure as a Service (IaaS) for CERN's private, OpenStack-based cloud. For more details, some distinguished former colleagues made these interesting presentations: Configuration Management and CERN's OpenStack Cloud.

My alma mater is the National and Kapodistrian University of Athens (UOA), Department of Informatics and Telecommunications. I worked there with Alex Delis from Management of Data, Information & Knowledge group and with Mema Roussopoulos from the Distributed Systems Lab.


My research interests are in computer systems, including operating systems (OSes), distributed systems, and machine learning (ML) systems. Current, I focus on addressing challenges inherent in the development of modern mobile, cloud, and ML applications. These applications are fundamentally different from traditional desktop applications and require renewed support from operating systems as well as new development environments. I have been running measurement studies to identify missing or mismatched abstractions for modern applications, and invent new programming abstractions and systems to support the needs of modern applications. Following are some projects in this space.

Current Projects

Measurement of POSIX Abstractions in Modern OSes
In our latest work, we demonstrated that the abstractions offered by traditional OS standards, such as the POSIX API, are insufficient to support modern applications, such as those running on Android, iOS, and OSX. Modern applications rely upon very different abstractions, which are currently supplied by user-space libraries implemented atop POSIX. This layering causes mismatches, inefficiencies, and even security risks. For example, we found that new abstractions typically rely on POSIX extension APIs (i.e., ioctl) to implement their functionality, suggesting that POSIX lacks appropriate abstractions for modern workloads. Extension APIs are problematic, because their invocations cannot be mediated by the OS, putting pressure on user-space libraries and kernel device drivers to implement correct and coherent protections of these invocations. We are now studying specific implications and identifying the new OS abstractions needed to more securely support modern applications.

Testing Tools for Data-Driven ML Applications
A key aspect characterizing modern applications is their increased reliance on data and data-driven decision making. While often beneficial, this practice can have subtle detrimental consequences, such as discriminatory or racially offensive effects. We argue that such effects are bugs that should be tested for and debugged in a manner similar to functionality, reliability, and performance bugs. To this end, we developed Fairtest, a testing toolkit for data-driven applications that identifies unwarranted association between an application’s outputs and user subpopulations, including sensitive groups (e.g., defined by race or gender). Our paper, accepted in the second European Symposium on Security and Privacy, has attracted attention from several investigative journalists, who are considering using it to study unwarranted associations in several applications.

Security of Machine Learning Systems under Adversarial Settings
State of the art machine learning systems, such as those using deep neural networks for classification and regression, have recently achieved unprecedented success in a variety of tasks ranging from image and speech recognition to data compression. Recent work however has made an intriguing discovery: these systems are extremely susceptible to adversarial attacks. That is, an adversary can modify correctly classified samples by adding an infinitestament amount of carefully crafed noise that will force the model to misclassify (with high confidence) samples only marginaly different from the original ones. In lieu of these vulnerabilities, I am currently working on defences to harden machine learning systems and increase their robustness against adversarial attacks. My long-term goal is to contribute solutions that will shed light into arcane ML models and will arm developers with powerful debugging primitives to help identify and address subtle security, performance, and reliability challenges.

Previous Projects

Before joining Columbia, I worked on several projects related to distributed, peer-to-peer systems, such as BitTorrent. In one of our projects, we enhanced the BitTorrent protocol with an optimistic unchocking policy that improves the quality of inter-connections amongst peers and increases the number of peers that are both directly connected and interested in cooperation. Our evaluation showed that the approach significantly outperforms BitTorrent’s original unchocking policy by (a) increasing the number of directly cooperating peers, (b) easing the load on seeders by having more peers act as data intermediaries, and (c) shortening the bootstrapping period for fresh peers.



  • Vaggelis Atlidakis, Patrice Godefroid, Marina Polishchuk.

    ''REST-ler: Automatic Intelligent REST API Fuzzing.'' [paper]

  • Conferences

  • Vaggelis Atlidakis, Jeremy Andrus, Roxana Geambasu, Dimitris Mitropoulos, and Jason Nieh. ''POSIX Abstractions in Modern Operating Systems: The Old, the New, and the Missing.'' In Proceedings of the eleventh European Conference on Computer Systems (EuroSys '16), London, UK, April 2016.

    [paper] [slides] [presentation] [code]

  • Florian Tramer, Vaggelis Atlidakis, Roxana Geambasu, Daniel Hsu, Jean-Pierre Hubaux, Mathias Humbert, Ari Juels, and Huang Lin. ''Discovering Unwarranted Associations in Data-Driven Applications with the FairTest Testing Toolkit.'' Accepted in the second European Symposium on Security and Privacy (Euro S&P '17), Paris, France, April 2017


  • Vaggelis Atlidakis, Mema Roussopoulos and Alex Delis. ''Changing the Unchoking Policy for an Enhanced Bittorrent.'' International European Conference on Parallel and Distributed Computing (EuroPar '12), Rhodes, Greece, August 2012.

    [paper] [slides]

  • Journals & Magazines

  • Vaggelis Atlidakis, Jeremy Andrus, Roxana Geambasu, Dimitris Mitropoulos, and Jason Nieh. ''POSIX has become outdated.'' USENIX ;login: Magazine, 41(3), Fall 2016.


  • Vaggelis Atlidakis, Mema Roussopoulos and Alex Delis. ''EnhancedBit: Unleashing the Potential of the Unchoking Policy in the BitTorrent Protocol.'' Journal of Parallel and Distributed Computing (JPDC),
    Vol. 74, Issue 1, pp. 1959-1970, January 2014.


  • Personal

    I was born in Athens, Greece and used to live in Galatsi. My mother, Chrysoula, is now on pension and my father, Pavlos, works for the public sector. I have a brother, Konstantinos, who is an undergrad at Athens University of Economis and Business, Department of Statistics.

    Life taught me not to forget people who stood by me and helped me realize my dreams; and there are plenty of them! First of all, I am deeply thankful to my parents who dedicated themselves to raising me and my brother. I am also grateful to my high-school teacher Aris Arabatzis who helped me find out my love for sciences. Last but not least, I feel deep respect for my undergrad advisors, Alex Delis and mema Roussopoulos, who helped me tremendously during my first research steps.

    Apart from computer science, I am also interested in international history, politics, and macro-economics. I am a passionate opponent of neo-liberalism and austerity which, in our days, is an insurmountable obstacle to the prosperity of the EU according to the opinion of multiple prestigious elite academic members, including Joseph Stiglitz (Columbia University), Paul Krugman (Princeton University), Thomas Piketty Paris School of Economics), and many more.

    My political views and generally my approach of life is absolutely reflected in the following quote:

    “Above all, always be capable of feeling deeply any injustice committed against anyone, anywhere in the world.”


    The preferable way of reaching me is via e-mail.

    v.atlidakis -*at- gmail -*dot- com