Lowering the cost of anonymization

a PhD thesis

Acknowledgements

Ugh, emotions.
(Martha Wells, Network Effect)

Pursuing a part-time PhD project, in a large company without any established process nor precedent for doing so was, unsurprisingly, not exactly straightforward. It required persistent optimism, a fair amount of perseverance, and a lot of unearned privilege. Above all, though, it required the support of a great many people, without which this whole endeavor would been entirely impossible. I cannot possibly do all of them justice in just a few words, but this certainly will not stop me from trying.

First of all, obviously, I would like to thank David Basin, my advisor. He decided to bet on me despite my part-time constraints, and having a project fairly distinct from the rest of his group: I am immensely grateful for the mentorship, trust, and independence he provided me during these four years. His encouragements to slow down when I was over-stretching myself, and his telling me with no hesitation that he would hire me full-time if things at Google got too heated, are the kind of show of support that I wish for any PhD student.

Thanks as well to Ashwin Machanavajjhala, Carmela Troncoso, and Dennis Hofheinz, who kindly accepted to be a part of my dissertation committee. I am also grateful to André Miede and Tino Wagner, whose LaTeX packages made it surprisingly painless to produce a beautiful-looking thesis.

Laura Arhire supported me from the very start, and was instrumental in my growth as an engineer; I can only dream of becoming as good a manager as she is one day. She, and Lorenzo Martignoni, gave me a level of freedom that most people only dream of: I thank them both for their trust and kindness. Peter Dickman tried to tell me that doing a part-time PhD was a bad idea, then when I still persevered, he provided me with much-needed help to collect all necessary approvals, and convinced me not to give up in the face of seemingly endless bureaucracy. His mentorship and support during the past four years are a major reason why I somehow have not been fired yet; I am very grateful for that.

To obtain necessary approvals, Michel Benard’s assistance was also crucial; I am thankful that he took so much of his time to help me navigate internal processes. I am also very grateful to Cyrill Osterwalder, Niels Provos, and Gerhard Eschelbeck for being supportive of this project, and unlocking the necessary funding. Thanks to Pern Hui Chia for being the main engineering lead for the initial project and for reviewing so much of my code and drafts. Many thanks to Bryant Gipson as well for being such a fantastic cheerleader for my work.

If support from leaders made it possible to kick-start this project, support from peers is what made it sustainable and exhilarating throughout. I was very lucky to work with excellent researchers, who taught me a lot about research, writing, and teamwork: thanks to Andreas Lochbihler, Balazs Pejo, Elisabeth Krahmer, and Esfandiar Mohammadi, for being lovely people to work with and for making scientific research engaging and fun.

Speaking of research, I am also thankful to the many people who reviewed parts of our drafts, answered our emails about their work, helped us with thorny issues, or alerted us of typos and mistakes on the published thesis: Aaron Johnson, Adam Groce, Adam Smith, Alain Forget, Aleksandra Korolova, Aleksei Triastcyn, Alex Kulesza, Antoine Amarilli, Ashwin Machanavajjhala, Borja Balle, Bryant Gipson, Chao Li, Chinmoy Mandayam, Christophe de Cannière, Kareem Amin, Christoph Dibak, Daniel Kifer, Dennis Kraft, Henk Brozius, Ilya Mironov, Iosif Pinelis, Ismaël Bouya, Jakob Dambon, Jessica Staddon, Jonathan Katz, Kunal Talwar, Lea Kissner, Lorenzo Martignoni, Marc Jeanmougin, Michael Lugo, Mikhail Berlyant, Nina Taft, Pablo Rauzy, Patricia Mirabile, Patrick Scheibe, Peter Dickman, Pierangela Samarati, Rebecca Balebako, Rees Davison, Roberto Tauraso Sebastian Meiser, Theresia Weise, Úlfar Erlingsson, Virgile Andreani, Yu-Xiang Wang, Yusuke Kawamoto, as well as Stack Exchange users egreg, Mark, PSPACEhard, usul, and I apolologize if I missed anyone. I am also thankful to all anonymous reviewers, in particular those from IEEE S&P who decided to reject the first version of our cardinality estimators paper: their decision was the main motivator for me to come back to the drawing board and significantly strenghten the result; it was an excellent illustration of the review process working as intended and producing good outcomes.

Alessandro, Alejandro, Benjamin, Ioana, Lucas, Marius, Nevena, Paweł, Régis, Robert, Roman, Simone, Victor, and Yannis: you made the office a fun place to be, and the endless (and impressively creative!) trash talk at the foosball table is the main reason why I am looking forward for the office to fully reopen. I am also grateful to the Memecademy (yes) for making work a little bit more hilarious and many times more absurd, and for honoring me with the most treasured awards I ever received. Special hat tip to Heradon Douglas’ creative genius. Thanks also to Patrick and Sven for being very pleasant officemates at the university.

Cloud DLP, the team I got to work on at the beginning of this project, was a fantastic opportunity to build and ship software for reidentification and risk analysis to the outside world; I am especially grateful to Jordanna Chord for being an inspiring leader, to Scott Ellis for helping me understand the difference between theory and practice, to Felipe Hoffa to take me to fun industry conferences and encouraging me to invest in external outreach, and to Emily Rossetti and Michael Daub for being excellent teammates.

My work in the Anonymization team has been, and still is, incredibly fun and rewarding. This is due in small part to the terrific problem space, but mostly to the exceptional people who compose it. Heartfelt thanks to Miguel Guevara, whose efforts to tackle such an uncharted problem space early on are nothing short of amazing and inspiring; much of what we do today has been made possible by your initial vision and leadership. Many thanks to Daniel Simmons-Marengo and Milinda Perera for their impossible kindness, and for their principled and rigorous approach to privacy consulting work. (Oh, and Daniel, thank you for the many excellent book recommendations, please keep them coming!) Thank you to Chao Li, Sławek Goryczka, and Pern Hui Chia for making this group survive against all odds when we were under an absurd load. I learned so much from all of you. Special thanks for the trust you put in me when you convinced me to take on the leadership hat for Anonymization consulting. It was the best and most terrible decision. I would do the same mistake again in a heartbeat.

Newer folks to the anonymization world, it is such an honor and a pleasure to work alongside you. Thank you to Janina Voigt and Yurii Sushko: you two are excellent leaders; what we have accomplished this past year is a testament to your team-building skills. Special thanks to Raine Serrano, Aishe Krishnamurthy, and Jessica Colnago: I would have completely drowned under the load if it were not for your exceptional work making anonymization consulting better for everyone. Many thanks to Christoph Dibak, David Marn, Dennis Kraft, Maria Telyatnikova, Miraç Vuslat Başaran, and Sasha Kulankhina; working with you and helping you ramp up and become anonymization experts in such a short time is probably the most impactful and rewarding thing I have done while at Google. (And Miracç, you turning Privacy on Beam from my ugly proof-of-concept to a production-ready open-source library still brings a tear to my eye when I think of it.) Thanks to Alain Forget, Vadym Doroshenko, and Xinyu Ye; I almost want to apologize to you for not having spent as much time as I would have liked working with you, but the fantastic work that you are all doing clearly shows that you do not, in fact, need my help. To all of you, I am awe-struck by your enthusiasm, eagerness to learn, technical depth, and kindness. I cannot wait to see how you grow, and what you accomplish in the future.

Google has been an exceptional place to grow, not only by improving technical and interpersonal skills, but also as a human being. Examples of extraordinary leaders abound, but I especially would like to thank Lea Kissner for being such an inspiring example of compassionate leadership, unbelievably kind and effective communication skills, and principled decision-making. Google is also where I learned the crucial importance of speaking truth to power and holding leaders accountable for their decisions. I am especially grateful to Irene Knapp and Liz Fong-Jones for giving me both the inspiration and the practical tools to organize for ethical principles. Thanks to them, I now know that you should, in fact, meet your heroes: if you choose them well, they will still somehow be as kind and inspiring in real life as you imagined they would be.

This naturally leads me to extend a wholehearted thank you to allies and accomplices of the past couple of years. Thanks to everyone making the Zürich site a fairer place to work: André Nogueira, Eli Stevens, Miriam Berger, Pedro Gonnet, Tom McAdam, Urs Zbinden, and others whom I forget or who prefer not to be named, for their involvement, advice, and/or support. Extra-special thanks to Adrien Kunysz and Raphaël Jamet for being at my side in October 2019; I would probably have exploded from all the stress of that absurd week if it were not for your support. Rebecca, thank you for the many discussions about work and life, and your stellar advice on both sides; I am lucky to be friends with you.

I benefited from an excellent work environment also thanks to my colleagues in temp, vendors, and contractors positions: cafe staff, baristas, massage therapists, cleaners, QA engineers, and many others. These folks hold the company together, and play a critical role in building and maintaining our core products (content moderation, rating, internationalization…), yet they are still kept as second-class citizens by the company. They do not get the pay nor benefits they deserve, and the lack of recognition they get is insulting: I cannot find a single email welcoming them to the team, or wishing them well when they leave, even when we worked alongside them every single day for months. As a result, I can only list a few folks I particularly remember for their kindness: Chef Steph, Moussa Dabre, Mohammad Bibaoune, Violeta Boneva, Wanapa Inthayard. Thanks to them and to all the others.

Having two jobs, even when one has exceptional benefits, sometimes presents quite a challenge for the body and brain. I am particularly grateful to Isabelle Frank-Ziem, Johann Glessgen, and Philippe Stöckli, whose professional expertise helped me not fall apart when things got particularly tough.

The real treasure, of course, is the friends we made along the way. A million hugs to Streambed folks, the most wholesome online community there ever was; I am so grateful for all of you, and I can only hope to be as kind and strong as you lot one day. Much love towards Postcrossing enthusiasts as well: you are what makes the world absurd and beautiful. Thanks to everyone who ever sent me kind words for my blog posts or other stuff I put online, you have no idea how much it means. If you are reading these lines, go take a couple of minutes to send a few nice words to your favorite online content creators. Go on! This thesis will still be there when you come back. Speaking of which, thank you to Confused Bi-Product of a Misinformed Culture, whose mixes were the main soundtrack for my work.

Tabletop games are one of the main things that kept me sane these four years. Wilson, thank you for continuing to invite me to play even though I only have time once in a blue moon. Sam, Marco, and Praveena, thanks for the fun times rolling dice together. Cheers to the cast of Critical Role, who made me be enthusiastic about playing D&D again, and provided me with countless hours of much-needed entertainment during confinement. Many thanks to Andrei, Ioana, Lisa, Lucas, and Robert for being such a fun group to play with, and special thanks to Sylvain for being such a devious, fantastic DM.

The trick to having a social life outside of your work in Switzerland is to find weird, absurd, joyful niche communities: you end up meeting exceptionally lovely people. Maggie, thank you for the many drunken discussions about everything and anything. Gabriella, thank you for the ridiculous “five-star” experiences, for your trust, and for your great approach to life. Marco, thank you for being your beautiful, opinionated self, I aspire to be 1% as cool as you some day. Dan, thank you for your openness to try novel and scary things, and for the many fun evenings during confinement. I am lucky to have you in my chosen family.

Last but not least. Chloe, my love, thank you so much for being in my life. Thank you for the many fun moments, thank you for the constant support, thank you for the cathartic rants about thesis writing, thank you for the mind-boggling level of trust, thank you for the positive approach to relationships and life, thank you for the much-needed cuddles. Here’s to many more years of swimming up rivers, cooking delicious food, waking up lazily on Sunday mornings, drinking cocktails, and dreaming about the future.

Remerciements

Cette thèse est dédicacée à Mamine, feu ma grand-mère paternelle. J’aspire à faire preuve d’une fraction de sa gentillesse légendaire. Nombreux sont les gens m’ayant poussé à faire une thèse, mais c’est elle qui m’a donné le gros de ma motivation et persévérance.

Je suis reconnaissant d’avoir une famille qui m’a soutenu tout du long: l’amour, c’est de continuer à manifester son soutien et affection même lorsqu’on ne comprend pas certains choix de vie ; j’ai une chance infinie à ce niveau. Papa, merci pour m’avoir donné le virus du syndicalisme, et pour continuer à me supporter même quand je déblatère ce genre d’ânerie. Maman, merci mille fois pour l’amour débordant, les longues discussions philosophiques, les hectolitres de thé, et le regard résolument positif sur la vie. Carline et Violaine, merci pour le soutien constant et l’apport de zen lorsque j’en avais sacrément besoin. Merci à Françoise pour les conseils et l’hospitalité légendaire, merci à Hervé pour m’avoir donné le goût pour l’informatique il y a une vingtaine d’années, merci à Nelly et Margot pour être les cousines les plus adorables et enthousiasmantes de l’univers, merci à Frédi, Cécile, Papi et Mamike pour les chouettes et nombreuses fêtes et vacances en famille. Je vous aime fort!

Merci à tous mes potes de l’ENS et adjacents pour m’avoir injecté une quantité adéquate de pression sociale pour que je fasse une thèse. Vous aviez raison : c’était un projet absurde, mais quand même bien marrant. Merci aux Grotas pour les chouettes discussions politiques qui m’ont aidé à faire la part des choses et à ne pas trop devenir lobotomisé par la culture d’entreprise. Tout plein de remerciements à a3nm, Elarnon, louis, Mc, Machin, olasd, pintoch, Robin, Yap, et tous les autres gens d’IRC pour vos principes et pour les discussions mathématiques ou philosophiques aussi interminables qu’enrichissantes, et pensées spéciales pour ceux d’entre vous qui m’ont suivi dans la poudreuse en hiver!

Alex, Alma, Nicolas, et Shauna, merci pour les visites, les jeux de société et les nombreux chouettes moments. Noélie, merci pour les visites, toujours plaisantes même lorsqu’on se fait surprendre par l’apocalypse. Pauline, merci pour l’hospitalité, les millions de cartes postales réjouissantes, et le soutien moral permanent. Manon, merci pour les jeux et la confiance qui va avec, je suis fier de t’avoir connue avant que tu ne deviennes une auteure mondialement célèbre. Juliette, merci pour la confiance et pour l’inspiration, je suis fier et chanceux de te compter parmi mes amies. Patricia, merci pour le soutien, les recommandations de lecture, et les excellents conseils sentimentaux. Laure, merci aussi pour la confiance, les vacances random et les discussions à cœur ouvert m’ont probablement évité de perdre pied plus d’une fois, merci tout plein pour être une personne aussi exceptionnellement enthousiasmante. Rien que lister tous les gens qui m’ont accompagné me met la larme à l’œil… Je ne sais pas ce que j’ai fait pour mériter d’être aussi bien entouré.

Danksagung

Sfé, Pascal und Tricia, ihr seid der Grund, warum ich mich zum ersten Mal in Zürich wie zuhause gefühlt habe. Tausend Dank für die Abendessen, Spiele une Organisation der Veranstaltungen. Olga, danke für dein Vertrauen une die vielen lustigen Abende. Danke, dass ihr alle so seltsame und augezeichnete Menschen seid.

Annerose und Bernd, vielen Dank für die herausragende Gastfreundlichkeit. Das Gefühl ein Familienmitgleid zu sein war sowohl sehr berührend als auch entscheidend, um meine Doktorarbeit fertig zu schreiben.

Schließlich, Sia, Liebe meines Lebens. Ich weiß nicht, wie ich dir nur annähernd für die ständige Liebe, Unterstützung une Vertrauen in diesen Jahren danken könnte. Du warst in den dunkelsten Momenten für mich da, und du hast die besten Zeiten tausendmal freudiger gemacht. Ich freue mich auf viele weitere Kuschelstunden, ausgefallene Abendessen, Autoreisen, Massagen, Star-Trek-Abende und Träume von der Zukunft.

All opinions here are my own, not my employers.
I'm always glad to get feedback! If you'd like to contact me, please do so via e-mail (se.niatnofsed@neimad) or Twitter (@TedOnPrivacy).