2017 USC Viterbi Graduate Hackathon

| April 6, 2017

Five teams competed against each other over spring break at the USC Viterbi Graduate Hackathon to solve problems with facial clustering and IoT security

Participating students in the 2017 USC Graduate Student Hackathon

Over spring break, amidst drawings of designs and outlines of complex codes and algorithms, five teams of graduate students competed against each other in the inaugural USC Viterbi Graduate Hackathon Series.

Although the competition was located at the USC Information Sciences Institute (ISI) in Marina Del Rey overlooking Pier 44, each team managed to keep their eyes glued to the prize: a $1000 reward for the top team of each division. The teams spent three days and two nights, from March 14-16, working on different ways to address this year’s prompts: “Clustering Faces within a Face Data Set” and “Internet of Things (IoT) Home Inspector Challenge.” Three teams were assigned to the facial clustering prompt that focused on creating the best clustering technique to identify facial structure, and the other two teams were placed into the IoT Home Inspector Challenge prompt that deals with creating a technical solution to guard consumers against security attacks through the vulnerabilities of IoT devices in their homes, such as home surveillance cameras or baby monitors. Dr. Yue (Rex) Wu mentored the teams focusing on clustering faces within a face dataset with his intensive research background in image processing, pattern recognition, and cryptography and encryption. Dr. Clifford Neuman, the Director of the Center for Computer Systems Security at ISI, mentored the IoT teams with his expertise in the area of computer security and information infrastructure.

1st place team for facial clustering (Fanghao Luo, Tianshu Cheng, Vishnupriya Ravibalan and Jianfa Lin) and mentors

The teams were judged separately based on their prompt via a 20-minute presentation, followed by a brief demonstration. The facial clustering teams were judged on their program’s clustering technique using an objective metric to measure cluster performance. IoT teams were judged on how well they detect devices and the spatial rules they apply to segregate traffic to the rest of the network.

Judges from different fields of computer science and security were gathered to oversee the event, including USC alumnus, Henry “Hal” Mecke, who graduated from USC with a computer engineering degree through the progressive degree program. After getting to know each team, Mecke stated that he was “looking forward to what they’re going to produce. Everyone seems like they were getting where they wanted to be, so it’ll be exciting to see what they come up with.”

Mecke himself had competed in coding marathons and projects when he was a student at USC.

The USC Viterbi Graduate Hackathon also included students with previous hackathon experience. Anshuli Patil, a frequent hackathon participant, claimed her most memorable moment in this hackathon was “figuring out to how to improve the recall score.” As part of the facial clustering team, Patil help created a program to cluster images of the same individual together based on facial recognition, similar to Google Photos.

1st place team for IoT (Navaneeth Krishnan Yadunandanan, Karthik Chindalur Sridhara, Reshma Bhatia, Kaushal Shah and Prathamesh Naik) with mentors

Another hackathon veteran, Kaushal Shah, worked on identifying security vulnerability in IoT devices. Shah claims the USC Viterbi Graduate Hackathon is “very exciting because you can create products in a short time frame while social networking with like-minded individuals that you can possibly collaborate with in the future to make a difference in people’s lives.”

Shah and his team – Navaneeth Krishnan Yadunandanan, Karthik Chindalur Sridhara, Reshma Bhatia and Prathamesh Naik – were able to create a program that identifies personal IoT devices before creating a list of websites that the device can access. Based on the preapproved list, it restricts certain websites to prevent hackers from infiltrating IoT devices through a series of firewalls. The code is also compatible for machine learning, meaning it can learn and identify any random or new device that comes up in the market.

This real life application would protect an emerging line of IoT kitchen devices such as the Samsung Flex Refrigerator with Family Hub that allows users to stream music, share photos, and search new recipes online. More detrimental results can occur in medical devices like insulin pumps. Johnson & Johnson recently released a warning of the vulnerability of the insulin pump, OneTouch Ping, to hackers. Whereas the Flex Refrigerator may allow hackers to trigger the ice making mechanism at questionable hours of the night, hackers can change the insulin quantity dose in the insulin pump, which has severe detrimental effects on the patient, including the possibility of death. In light of the recent WikiLeaks documents revealing the CIA’s software used to break into various IoT devices, including smartphones and computers, the programs created in the IoT Home Inspector Challenge could very well prohibit hacking software from reaching home and personal IoT devices by completely blocking out websites not deemed appropriate or relevant to the device.

Shah states his favorite memory of the USC Viterbi Graduate Hackathon was the thrill he felt “identifying a lot of bugs really quickly just before the deadlines.” Luckily, Shah and his teammates were able to fix their program, winning first place in the IoT Home Inspector Challenge division. The team winning first place in the facial clustering division included: Fanghao Luo, Tianshu Cheng, Vishnupriya Ravibalan and Jianfa Lin.