(Last updated: April 2023)
Welcome! I am Javier, a PhD Candidate in AI Security in the Dept. of Computing at Imperial College London. My current interests are (not limited to) adversarial machine learning, aiming to investigate the security of machine learning algorithms; bilevel optimization problems; Generative Adversarial Networks (GANs); and federated learning. I focus on data poisoning attacks, where attackers can manipulate training data collected from untrusted sources to degrade the ML algorithm’s performance. I am part of the Resilient Information Systems Security (RISS) Group under the supervision of Prof Emil C. Lupu and Dr Luis Muñoz González.
For a high-level and fun presentation of my research interests in data poisoning, you can have a look at this video:
I have worked as a teaching assistant in several courses in ML, deep learning, and probabilistic methods at Imperial College London. I did a research internship in summer 2022 at IBM Research on ML security and machine unlearning.
I have extensive experience in prototyping ML algorithms in Python and PyTorch. My background is also in Telecom Engineering. In 2021 I was included in the Santander-CIDOB 35 under 35 List:
If you want to know fun facts about me, you can have a look at this video:
I am passionate about travelling. I am eager to learn, analytical, deliberative and achiever. I am always interested in challenges and open to new collaborations. Drop me an email or DM me on LinkedIn to connect!
PhD in Machine Learning Security, (exp.) 2023
Imperial College London
MRes (Hons) in Multimedia and Communications, 2017
Universidad Carlos III de Madrid
MSc (Hons) in Telecommunications Engineering, 2017
Universidad Carlos III de Madrid
BEng (Hons) in Telecommunications Engineering, 2015
Universidad Carlos III de Madrid
Researching the security of machine learning in the Dept. of Computing: analyzing the vulnerabilities of machine learning algorithms, designing effective attacks and evaluating their impact, and proposing defenses that can help these algorithms to be more robust to adversaries. Special focus on data poisoning attacks. Techniques developed based on bilevel optimization and Generative Adversarial Networks.
Assisted in the supervision of 2 MSc (one of them passed with Distinction), 1 MEng, and 1 Undergraduate Research Opportunities Programme (UROP) student research projects, and 1 group project (5 students) on data poisoning attacks against machine learning.
Research in antennas, passive electromagnetic sensors, and IoT applications in the Dept. of Signal Theory and Communications.
We proposed a novel low-cost and portable IoT reader for passive wireless electromagnetic sensors. An interesting application is the remote measurement of harmful substances. To our knowledge, it was the first wireless reader of passive electromagnetic sensors including IoT functionalities (“An IoT Reader for Wireless Passive Electromagnetic Sensors”).
We also designed a novel contactless sensing system composed of a metamaterial-inspired sensor and a reader antenna, in order to detect substances in short distances in real time. This led to a low-cost, replaceable, battery-free and fully passive solution. Moreover, thanks to the short-reading range, the sensor avoids external interferences and undesired radiations (“"A Contactless Dielectric Constant Sensing System Based on a Split-Ring Resonator-Loaded Monopole”; “A Contactless System for the Dielectric Characterization of Liquid Drops”).
We also collaborated with an important Spanish telecommunications company (Prodetel, S.A.) in an R&D project. We designed an innovative multiband feeder for aeronautic telemetry applications.
80%
80%
90%
70%
70%
80%
80%
80%
70%
80%
60%
50%
80%
50%
50%
80%
40%
80%
80%
80%
100%
80%