Stefano Longari

profile

stefano.longari@polimi.it
02 2399 3443






I am currently an assistant professor (RTDa) at Politecnico di Milano, working in the system security group at NECST Laboratory inside the Dipartimento di Elettronica, Informazione e Bioingegneria. The focus of my research revolves around offensive and defensive techniques for the security of cyber-physical systems and transportation systems, e.g., automotive, space, industry 4.0, and critical infrastructure.




Teaching:

Human and Physical Aspects of Security

Available to Politecnico di Milano's Computer Science and Engineering master students and to Bocconi's Cyber Risk Strategy and Governance master students.

Program

Course under construction. Topics will include:
- The threat modeling process, common threat modeling frameworks
- The human factor in security: Social engineering, its principles, digital and physical mediums, mitigations and countermeasures.
- Cyber-physical systems security, the differences between IT and OT, technology use cases, and countermeasures.

Resources TBD

Advanced Research Topics in cybersecurity

Available to Politecnico di Milano's Computer Science and Engineering Ph.D. students and optionally to master students.

Program

The goal of the course is that of providing the students with an understanding of the latest years cutting edge research challenges and solutions in the cybersecurity field. The first half of the course will be held as frontal lectures, while the second half of the course will be held as a flipped classroom with paper discussions.
The topics we will discuss about are: Economics of cybercrime, Software vulnerabilities, Malware analysis, Hardware and embedded systems security, Cyber-physical and critical systems security, Cyberwarfare and cyberdefense, Machine learning for security and security of machine learning

Research:

I am particularly interested in the area of cyber-physical systems and how they can be defended against threats by incorporating techniques from multiple fields. My goal, in studying both defensive measures and new attack methods, is to gain a comprehensive understanding of the threat model of this evolving field. My research interests mainly delve into the security of industrial and manufacturing automation, land - mainly automotive - and air transportation systems, space and satellite systems, and overall critical cyber-physical infrastructures.

If you are interested in a research thesis project on the aforementioned topics, feel free to contact me.

Publication Highlights

Janus: A Trusted Execution Environment Approach for Attack Detection in Industrial Robot Controllers (IEEE TETC 2024)

The evolution of our work on monitoring mechanisms for cyber-physical systems based on the use of Trusted Execution Environments, to guarantee the integrity of the attack detection algorithm even in case the controller's software is compromised, while not requiring external hardware for its detection process.

Preprint / Link

Evaluating the Impact of Privacy-Preserving Federated Learning on CAN Intrusion Detection (IEEE VTC 2024)

Our initial project focusing on the federation of ML/DL algorithms for intrusion detection on CAN. The goal of the paper is to understand the limitiations and requirements to federate CANdito (an evolution of CANnolo, see below) while maintaining detection capabilities and low network overhead.

Preprint / To be published

CANflict: Exploiting Peripheral Conflicts for Data-Link Layer Attacks on Automotive Networks (ACM CCS 2022)

One of our team's most effective attack approaches. We demonstrate how it is possible to generate advanced data-link layer attacks against automotive networks (CAN) without requiring physical access to the vehicle, previously an assumption for that kind of attacks.

Preprint / Link

CANnolo: An anomaly detection system based on LSTM autoencoders for controller area network (IEEE TNSM 2020)

The beginning of our research team work on automotive intrusion detection based on deep learning. Our IDS attempts to reconstruct the last CAN packets with a specific ID, and depending on its reconstruction error defines whether a sequence is anomalous or not.

Preprint / Link