Ashish Kundu

Title: Ashish Kundu: Innovator in Federated Learning and Cybersecurity

Introduction

Ashish Kundu is a prominent inventor based in San Jose, CA, known for his contributions to the fields of federated learning and cybersecurity. With a total of 4 patents, Kundu has made significant strides in developing innovative solutions that address complex challenges in technology.

Latest Patents

One of Kundu's latest patents is titled "Uncheatable Federated Learning." In this invention, a device identifies a plurality of nodes within a distributed or federated learning system. The device receives model training results from these nodes and determines whether any particular node or subset of nodes has provided fraudulent results. Based on this determination, the device initiates corrective measures in accordance with a specified policy.

Another notable patent is "Automatic Construction of Attack Graphs Using Large Language Models." This method involves a computing system that sends vulnerability information to a large language model to extract structured information. The system then builds a natural language query based on this structured information and sends it to the language model to generate a portion of an attack graph from the vulnerability data.

Career Highlights

Ashish Kundu is currently employed at Cisco Technology, Inc., where he applies his expertise in developing advanced technologies. His work focuses on enhancing security measures and improving the efficiency of machine learning systems.

Collaborations

Kundu has collaborated with notable colleagues, including Ramana Rao V R Kompella and Renascence Tarafder Prapty. These collaborations have contributed to the advancement of innovative solutions in their respective fields.

Conclusion

Ashish Kundu's work exemplifies the spirit of innovation in technology, particularly in federated learning and cybersecurity. His patents reflect a commitment to addressing critical challenges in these areas, making him a valuable contributor to the tech industry.