Our method
Most detection work tries to see the threat better. We start from the opposite end.
Every effective threat has one property it cannot give up. That property is usually what makes it hard to detect, and it is almost never where attention goes, because attention follows the threat itself.
We look for the condition the threat depends on rather than the threat. A faint signal that only one thing can produce is worth more than a strong signal that a hundred things produce.
This is a method, not a slogan. It is why our system is simpler than the systems it outperforms.
The problem
Small unmanned aircraft are inexpensive, widely available, and hard to see. Sites built to hold back vehicles and people now face something an order of magnitude smaller and faster than anything their sensors were specified against.
Detection is the hard part. An interceptor is only as good as the warning that triggers it, and the warning has to arrive early enough to be acted on.
There is a second problem, and it is quieter. A system that acts without detecting cannot say what it faced. When it works, nothing appears to have happened. When it fails, you learn from the damage. Either way the event leaves no account of itself, and a site that cannot say what crossed it cannot plan against it, argue for the budget to stop it, or tell whether last quarter was quiet or merely unobserved.
What we build
Hiltra Defense develops electro-optical detection systems for counter-drone applications.
The method is the subject of a patent application filed in 2026. We do not describe it publicly. Technical material is available to evaluating organizations under a non-disclosure agreement.
Design principles
Deterministic
The detection decision comes from explicit logic, not from a trained model. The same input produces the same output every time, and the reason for any alarm can be stated. Behavior can be characterized before deployment rather than inferred from trial results.
Additive
The system runs alongside the radar, radio-frequency and camera coverage a site already operates, and reports into existing infrastructure. Nothing has to be removed for it to be useful.
Deployable along a whole perimeter
A perimeter is only as protected as its least covered segment. The system is built from commercially available components, so covering the full line is a purchasing decision rather than a program.
Auditable
Each detection is reported as a discrete event, timestamped and carrying the basis for the decision, in a form a command system can store and replay. We hold nothing ourselves. The record stays in your infrastructure, under your rules. A blind interception that worked and a night with no attempt at all look identical afterwards, and only a detection event tells them apart.
Where we are
- Patent application filed in 2026.
- Detection demonstrated in controlled experiments.
- Field evaluation underway with defense and security organizations.
Team
- Nitsan Sharon Co-founder and CEO
- Prof. Amir Rosenthal Co-founder. Electrical and Computer Engineering, Technion.
- Gil Gelbert Co-founder. Doctoral researcher, Electrical and Computer Engineering, Technion.
Contact
Evaluating organizations can request technical material under a non-disclosure agreement. What that involves.