Initially from the small Balkan nation of Montenegro, Strahinja (Strajo) Janjusevic says his life has unfolded in sudden methods, for which he’s deeply grateful. After graduating from highschool, he was chosen to symbolize his nation in the US, learning cyber operations and pc science on the U.S. Naval Academy in Annapolis, Maryland. He has since continued his cybersecurity research and is at the moment a second-year grasp’s pupil within the Expertise and Coverage Program (TPP), hosted by the MIT Institute for Knowledge, Methods, and Society (IDSS). His analysis with the MIT Laboratory for Data and Determination Methods (LIDS) and the MIT Maritime Consortium group goals to enhance the cybersecurity of crucial maritime infrastructure utilizing synthetic intelligence, contemplating each the expertise and coverage frameworks of options.
“My present analysis focuses on making use of AI methods to cybersecurity issues and analyzing the coverage implications of those developments, particularly within the context of maritime cybersecurity,” says Janjusevic. “Representing my nation on the highest ranges of schooling and trade has given me a novel perspective on cybersecurity challenges.”
Janjusevic’s pathway from Montenegro to Maryland was created by a program that enables chosen college students from allied nations to attend the U.S. Naval Academy. Janjusevic graduated with a twin bachelor’s diploma in cyber operations and pc science. His undergraduate expertise offered alternatives to collaborate with the U.S. navy and the Nationwide Safety Company, exposing him to high-level cybersecurity operations and fueling his curiosity in tackling advanced cybersecurity challenges. Throughout his undergraduate research, he additionally interned with Microsoft, creating instruments for cloud incident response, and with NASA, visualizing photo voltaic knowledge for analysis purposes.
Following his commencement, he realized that he nonetheless wanted extra information, notably within the space of AI and cybersecurity. TPP appealed to him instantly due to its twin emphasis on rigorous engineering innovation and the coverage evaluation wanted to deploy it successfully. Janjusevic’s experiences at TPP have been a giant change from his time on the U.S. Naval Academy, with a unique tempo and surroundings. He has particularly appreciated with the ability to broaden his understanding about quite a lot of analysis domains and apply the self-discipline and information he earned throughout his time on the academy.
“My TPP expertise has been superb,” says Janjusevic. “The cohort is de facto small, so it looks like a household, and everyone seems to be engaged on numerous, high-impact issues.”
Mitigating the dangers of rising applied sciences
Janjusevic’s thesis brings collectively disciplines of cybersecurity, AI and deep studying, and management concept and physics, specializing in securing maritime cyber-physical programs — specifically, massive legacy ships. The hacking of those ships’ networks may end up in substantial harm to nationwide safety, in addition to critical financial results.
“Strajo is working to outsmart maritime GPS spoofing,” says Saurabh Amin, the Edmund Okay. Turner Professor in Civil Engineering. “Such assaults have already lured vessels off beam in contested waters. His method layers physics-based trajectory fashions with deep studying, catching threats that no single technique can detect alone. His experience has been very useful in advancing our work on menace modeling and assault detection.”
The analysis makes use of superior menace modeling and vessel dynamics to coach AI programs to differentiate between authentic maneuvers and spoofed alerts. It entails constructing a framework that employs an inner LSTM (lengthy short-term reminiscence) autoencoder to research sign integrity, whereas concurrently utilizing a physics-based forecaster to foretell the vessel’s motion primarily based on environmental components like wind and the ocean state. By evaluating these predictions towards reported GPS positions, the system can successfully distinguish between pure sensor noise and malicious spoofing assaults. This hybrid framework is designed to empower, not exchange, human operators, offering verified navigation knowledge that enables watch standers to differentiate technical glitches from strategic cyberattacks.
Sanjusevic has been in a position to improve his tutorial analysis with trade expertise. In summer time 2025, he interned with the Community Detection group on the AI cybersecurity firm Vectra AI. There, he investigated potential threats new applied sciences can carry, notably AI brokers and the mannequin context protocol (MCP) — the rising normal for AI agent communication. His analysis demonstrated how this expertise might be repurposed for autonomous hacking operations and superior command and management. This work on the safety dangers of agentic AI was lately offered within the preprint, “Hiding within the AI Site visitors: Abusing MCP for LLM-Powered Agentic Pink Teaming.”
“I used to be in a position to achieve sensible insights and hands-on expertise into how a knowledge science group makes use of AI fashions to detect anomalies in a community,” says Janjusevic. “This work inside trade straight knowledgeable the anomaly detection fashions in my analysis.”
Worldwide coverage perspective
“Strajo brings not only a excessive stage of intelligence and vitality to his work on cyber-physical safety for service provider vessels, but additionally a robust intuition from his Navy coaching that resonates deeply with the analysis effort and grounds it in actionable coverage,” says Fotini Christia, the Ford Worldwide Professor of the Social Sciences, director of IDSS and the MIT Sociotechnical Methods Analysis Heart, and a frontrunner of the MIT Maritime Consortium.
Janjusevic participates within the cybersecurity efforts of the Maritime Consortium, a collaboration between academia, trade, and regulatory companies targeted on creating technological options, trade requirements, and insurance policies. The consortium contains cooperation with some worldwide members, together with from Singapore and South Korea.
“In AI cybersecurity, the coverage factor is de facto essential, as the sector is so fast-moving and the implications of hacking will be so harmful,” says Janjusevic. “I believe there’s nonetheless numerous want for coverage work on this house.”
Janjusevic can also be at the moment serving to to prepare two upcoming main conferences: the Harvard European Convention in February, which is able to convene officers and diplomats from throughout the globe, and the Expertise and Nationwide Safety Convention in April, a collaboration of Harvard and MIT that brings collectively high leaders from authorities, trade, and academia to deal with crucial challenges in nationwide safety.
“I’m striving to discover a place the place I can affect and advance the cybersecurity subject with AI, whereas on the identical time main collaboration and innovation between the US and Montenegro,” says Janjusevic. “My objective is to be a bridge between Europe and the U.S. on this house of nationwide safety, AI, and cybersecurity, bringing my information to each side.”