Fairness in Machine Learning: Metrics, Trade-offs, and Evaluation

Fairness is not a property that can be established by checking whether one metric exceeds a universal threshold. It is a system-level question about people, decisions, benefits, harms, institutions, and the evidence used to justify an intervention. Machine-learning metrics still matter. They can reveal differences in selection rates, error rates, calibration, and performance across groups. But choosing a metric is already a policy decision: it determines which differences count as harms and which trade-offs receive attention. ...

October 15, 2023 · 12 min · Akshat Gupta

GlyphNet: Homoglyph Domain Detection with CNNs

Paper · Code · Dataset Authors: Akshat Gupta, Laxman Singh Tomar, Ridhima Garg Research overview GlyphNet studies visual detection of homoglyph domains using an attention-based CNN and a dataset of real and synthetically altered domain images. The paper reports 0.93 accuracy, F1 and AUC on its own test set. These results describe the paper experiment; the separately rendered public release is detailed under Paper dataset and public release. Introduction In cyber security, attackers employ different attacks to infiltrate our systems and networks, with the objective varying from stealing crucial information to inflicting system damage. One such deceptive attack is the homoglyph attack, which involves an attacker trying to fool humans and computer systems by using characters and symbols that may appear visually similar to characters used in real domain and process names but are different. ...

March 1, 2023 · 5 min · Akshat Gupta
Intel Machine Learning Workshop at Dr. Akhilesh Das Gupta Institute

Intel Machine Learning Workshop Recap

Date of Event: September 6, 2019 Venue: Dr. Akhilesh Das Gupta Institute of Technology & Management Event Overview: In an exhilarating day filled with insights and hands-on activities, the Machine Learning workshop organized by Intel turned out to be a landmark event for aspiring ML enthusiasts. Our auditorium, buzzing with the energy of keen learners, became a crucible for innovation and deep understanding in the rapidly evolving field of Machine Learning. ...

September 6, 2019 · 2 min · Akshat Gupta
NLP in Healthcare workshop at Poornima University

Role of Natural Language Processing in Healthcare

Date of Event: May 05, 2020 Venue: Online Event Overview: In May 2020, Poornima University in Rajasthan hosted a groundbreaking workshop focusing on the Role of Natural Language Processing (NLP) in Healthcare. This event provided a unique platform for healthcare professionals, technologists, and students to explore the intersection of advanced linguistic technology and healthcare applications. Highlights of the Workshop: Introduction to NLP in Healthcare: The workshop started with an overview of NLP and its evolving role in the healthcare sector, emphasizing how it transforms patient care and medical data analysis. ...

May 5, 2020 · 2 min · Akshat Gupta
Smart India Hackathon 2017

Smart India Hackathon 2017

Date: April 2017 Location: Chennai, Tamil Nadu, India Organized by: Government of India Smart India Hackathon (SIH) is one of the world’s largest open innovation hackathons, organized by the Government of India to provide students a platform to solve pressing problems faced by government ministries and industries. Problem Statement Our team worked on a problem statement issued by the Ministry of Earth Sciences — one of the most technically demanding tracks in the 2017 edition, involving the processing and analysis of earth observation and environmental data. ...

April 28, 2017 · 2 min · Akshat Gupta

AI Workshop in Ujjain

Date of Event: December 17, 2019 Venue: Ujjain Event Overview: Ujjain witnessed an enriching educational event in December 2019, with a workshop dedicated to introducing the fundamentals of Artificial Intelligence (AI). Designed for beginners and enthusiasts alike, this event served as a primer to the world of AI, attracting a diverse audience from students to professionals keen on understanding this cutting-edge technology. Highlights of the Workshop: Fundamentals of AI: The workshop focused on the core principles of AI, providing a comprehensive introduction to its basic concepts and terminologies. ...

December 17, 2019 · 2 min · Akshat Gupta

Model Extraction Attacks: Threat Models and Defences

A prediction API protects model files from direct download, but it still exposes information about the model through its outputs. With enough carefully chosen queries, an adversary may train a substitute that imitates the service, infer properties of the target, or—in narrower settings—recover parameters. These outcomes are often grouped under “model stealing”, although they are not equivalent. A surrogate that matches task accuracy may disagree with the target on many inputs. A high-fidelity copy need not contain the same parameters. Extracting memorised training records is a privacy attack, not proof that the model itself was copied. ...

September 15, 2024 · 12 min · Akshat Gupta