Encrypted RFID Backscatter With ML Decoding

ISEF Category: Embedded Systems

Subcategory: Networking and Data Communications  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A tiny tag can talk back without a battery. That sounds like magic, but it works because the tag changes how it reflects a radio signal. The hard part is making that reflection both readable and secure. That gives you a real research problem, not just a demo.

What Is It?

Backscatter communication sends data by changing reflections instead of generating a full radio transmission. Think of it like flashing a mirror with patterns instead of using a flashlight. A passive tag can do this with very little power, which is why RFID labels and smart sensors can work for so long without batteries.

An encrypted backscatter scheme adds a security layer. The tag or reader scrambles the communication in a planned way, then the reader decodes it with an algorithm, often a machine learning model. In simple terms, the reader has to learn the tag’s pattern through noise, interference, and changing distance. Your project can ask how well that decoding works when you change reflector design, channel conditions, or the model itself.

Why This Is a Good Topic

This topic gives you a strong mix of hardware, signal processing, and security. You can test real variables, measure decoding accuracy, and compare designs with numbers instead of opinions. It connects to supply chain tracking, smart packaging, access control, and low-power sensing. You can also scale the project to your skill level, from a basic RF demo to a deeper study of modulation, classification, and error rates.

Research Questions

• How does reflector geometry affect reader-side decoding accuracy?
• What is the effect of distance between tag and reader on bit error rate?
• Does adding encryption-specific randomness reduce classification accuracy under fixed noise?
• To what extent does a machine learning decoder outperform a rule-based decoder for weak backscatter signals?
• Which channel conditions cause the largest drop in tag identification accuracy?
• How does the choice of feature set change decoding performance across different reflector designs?

Basic Materials

• Software-defined radio platform, such as a USRP or HackRF if available.
• Passive RFID-style tag or custom backscatter reflector prototypes.
• Assorted conductive tape, foil, copper mesh, or printed reflector materials.
• Breadboard and jumper wires for low-power prototype control logic.
• Laptop with signal capture and analysis software.
• Calipers or ruler for repeatable tag placement.
• Tripod or fixed mounting setup to hold reader and tag steady.
• Notebook or spreadsheet for recording trials and labels.

Advanced Materials

• Software-defined radio reader with wideband capture capability.
• Passive backscatter tag prototypes with switchable reflector states.
• Vector network analyzer or spectrum analyzer for antenna and reflection checks.
• RF anechoic or semi-controlled test space, if available.
• Signal generator for controlled carrier injection.
• Shielded cables, attenuators, and RF connectors for calibration.
• Microcontroller or FPGA board for tag-side control experiments.
• Precision antenna mounts and distance rail for repeatable geometry.

Software & Tools

• GNU Radio: Builds and tests signal chains for capture, filtering, and decoding.
• Python: Cleans data, trains models, and runs statistical tests.
• scikit-learn: Trains baseline classifiers and compares decoding methods.
• ImageJ: Measures and compares captured spectrogram images if you convert signals into visual features.
• MATLAB: Supports signal processing and model comparison if your school has a license.

Experiment Steps

1. Define the exact backscatter behavior you want to encode, such as state changes, symbol patterns, or encrypted packets.
2. Choose one tag design variable to change first, such as reflector shape, spacing, or switching pattern.
3. Plan a reader pipeline that turns raw RF captures into features you can analyze consistently.
4. Build a labeled dataset that includes both clean and noisy captures so you can test generalization.
5. Select at least one baseline decoder and one machine learning decoder, then compare them with the same metric.
6. Design controls that separate signal loss, geometry effects, and true decoding errors.

Common Pitfalls

• Changing tag position between trials, which mixes geometry effects with decoding errors.
• Training a model on one channel setup and testing it on a different one, which makes accuracy look better than it is.
• Using only clean captures, which hides how the scheme fails in real noisy conditions.
• Skipping a baseline decoder, which makes it hard to prove the machine learning method adds value.
• Ignoring calibration drift in the reader or antennas, which can shift the signal features over time.

What Makes This Competitive

A strong version of this project does more than report accuracy. It compares multiple tag designs, multiple channel conditions, and at least one simple baseline against the machine learning decoder. The best entries also measure confusion patterns, bit error rates, or mutual information, not just percent correct. If you can test whether the scheme still works after a realistic attack or interference source, your project gets much stronger.

Project Variations

• Test the same backscatter scheme with different reflector materials, such as foil, copper tape, or printed conductive ink.
• Compare a rule-based decoder, a random forest, and a neural network on the same captured RF data.
• Study how the scheme performs when the tag is moved through different indoor environments, such as open space, near metal, or near people.

Learn More

• NIH PubMed: Search for review articles on RFID security, backscatter communication, and low-power wireless sensing.
• NASA Glenn Research Center educational materials: Search for wireless communication and RF basics articles and talks.
• NOAA National Severe Storms Laboratory data and signal resources: Useful for general signal analysis practice and noisy measurement thinking.
• MIT OpenCourseWare: Search for courses in signals and systems, communications, and machine learning that include lecture notes and assignments.
• IEEE Xplore or ACM Digital Library: Search for recent peer-reviewed papers on encrypted backscatter, ambient backscatter, and RFID decoding.