Attacking Yourself First: Inside Nosy Neighbor
The Setup: A Simple, Devastating Question
Imagine a car built by a dozen different suppliers — one team builds the engine control unit, another the infotainment system, another the anti-lock brakes — all wired into the same vehicle, each running on its own tight schedule of milliseconds. Now ask: how do you know none of those components can secretly learn what the others are doing? Not by reading their code, not by breaching their network — just by watching, from the outside, how long each one takes to finish its job. If timing alone can leak a secret, how would you ever catch it before someone else does?
The Breakthrough: Building the Attack to Find the Weakness
Here's what a team of researchers from Washington State University, the University of Colorado Colorado Springs, and Metro State University actually did.
They designed an attack — deliberately, on purpose — called NosyNeighbor, aimed at "partitioned" time- and safety-critical computing systems: the kind that run in avionics, medical devices like pacemakers, and anti-lock braking systems in cars. These systems are built the way modern software increasingly is, in separate, independently designed components — engine control, cockpit infotainment, flight navigation — packaged together and deployed as one coherent unit, each partition handling its own stage of an overall task within a strict time budget.
NosyNeighbor doesn't touch the software or algorithms of the system it's attacking. It's a side-channel attack: it sits outside the system, observes how long tasks take to run in one partition, and infers what's happening in others purely from that timing information. In experimental testing, it identified which specific task was executing in a target partition with roughly 73% precision under normal operating conditions — accurate enough, the researchers say, to represent a genuine safety risk, since an adversary only needs one correct inference to disrupt a safety-critical operation.
The work, led by co-author Monowar Hasan, an assistant professor in WSU's School of Electrical Engineering & Computer Science, and Vijay Banjaree, a former WSU postdoctoral researcher now at Argonne National Laboratory, was published in ACM Transactions on Cyber-Physical Systems.
Why It's Bigger Than It Looks
The instinct to build an attack in order to defend against one isn't new — it's the whole premise of red-teaming and penetration testing. What makes NosyNeighbor notable is where that instinct gets applied: not to a website or a corporate network, but to the increasingly modular, multi-vendor architecture underneath physical systems people trust with their lives. As Hasan points out, modern safety-critical systems are becoming more "containerized," with components from different vendors designed separately and only integrated at the end — which is efficient for manufacturing, but multiplies the number of components that have to trust each other without fully knowing each other's internals.
That's the deeper implication. If a side-channel can leak information between partitions that were never meant to share it, then the entire premise of siloed, vendor-separated safety systems — the idea that isolating components also isolates risk — needs re-examining, especially as supply-chain complexity keeps increasing.
The Part Nobody Talks About: This Is a Proof of Concept, Not a Fix
It's worth being precise about what NosyNeighbor actually delivers: a demonstration that the vulnerability exists, not a patch that closes it. The research grew out of Hasan's broader work on detecting and mitigating information leakage in time-critical cyber-physical systems, supported by a National Science Foundation CAREER award — work that is fundamentally about building better defenses, but whose current output is an attack, not a shield. Knowing that a 73%-precision side-channel is possible tells engineers where to look; it doesn't yet tell them exactly how to stop it, and closing a timing side-channel without breaking the strict deadlines these systems depend on is its own hard problem.
The Meta-Twist: The Danger Isn't the Leak, It's the Timing
The scariest part of NosyNeighbor isn't that it can guess what another partition is doing — it's what that guess can be used for immediately afterward. Hasan describes the real risk plainly: if an adversary knows a certain task is running at a certain time, they can disable that service, and in a safety-critical system, that's not an inconvenience — it's potentially catastrophic. His own example is stark: an anti-lock braking module disabled by an adversary at the exact moment a car is going through an intersection. The vulnerability isn't abstract information leakage in the way a stolen password is; it's a timing window that, once known, can be exploited in real time against a physical system that can't simply be logged out of.
Conclusion: Isolation Was Never as Isolated as It Looked
NosyNeighbor's real contribution isn't the 73% figure — it's the reminder that "separate" doesn't automatically mean "secure." Partitioned systems were built to contain failures and simplify multi-vendor development, not to withstand an adversary specifically hunting for what leaks between the seams. As Hasan puts it, thinking like an attacker is how you find out what the worst case actually looks like — and in systems where the worst case is a disabled brake at an intersection, finding that out in a lab is far better than finding it out on the road.
References:
https://www.eurekalert.org/news-releases/1137797
https://techxplore.com/news/2026-07-scientists-cyberattack-cyberattacks.html
https://www.miragenews.com/cybersecurity-experts-develop-nosyneighbor-1717630/