RAMBO air-gap attack turns RAM into a radio transmitter

Estimated reading time: 10 minutes

Key Takeaways

• Speed: The RAMBO paper reports leaking data from an air-gapped computer at 1,000 bits per second using software-generated radio signals.
• Equipment: An attacker needs software-defined radio (SDR) hardware and a simple off-the-shelf antenna to pick up the signals from a distance. The malware must already be on the target machine.
• What can leak: The paper says the channel can carry files, images, keylogging data, biometric information and encryption keys.
• Paper trail: Mordechai Guri is the sole author. The arXiv page says a version was accepted to the Nordic Conference on Secure IT Systems (NordSec) 2023. The arXiv preprint (v1) was submitted on September 3, 2024.
• Defense and evidence: The paper's countermeasures include Faraday enclosures. The retrieved sources are the author's own paper and a university record, so there is no broad independent expert consensus yet.

The RAMBO air-gap attack is a research method that turns a computer's memory into a tiny radio transmitter. The paper says malware can use it to leak data from an air-gapped computer at up to 1,000 bits per second, using only software-generated radio signals and an off-the-shelf antenna on the receiving end. The claim comes from the arXiv abstract page for the RAMBO paper.

An air-gapped computer is one that is kept off the internet and off other networks. People use this setup to protect their most sensitive data. RAMBO shows, in a research setting, that cutting the network cable does not cut every path out.

For founders and operators, the message is plain. "Air-gapped" no longer guarantees safety on its own. Sensitive environments need layers: strong controls on what gets in, monitoring, and defenses that think about hardware and physical space, not just network isolation.

How the RAMBO Air-Gap Attack Sends Data Without a Network

The paper is titled "RAMBO: Leaking Secrets from Air-Gap Computers by Spelling Covert Radio Signals from Computer RAM." RAM is the short-term memory a computer uses while it works. The paper focuses on the memory buses, which are the pathways that move data to and from that memory.

The core idea is simple to state. Malware on a compromised computer can make the memory buses generate radio signals. Those signals can carry secret data out of the room, even though the computer has no network connection.

The attack in four steps

Here is the flow as the paper's abstract describes it:

1. Malware is already on the air-gapped computer. The paper starts from a compromised machine. RAMBO is a way to get data out, not a way to break in.
2. The malware encodes the data into radio signals. It does this in software, using the memory buses as the source of the signals.
3. A receiver nearby picks up the raw signals. The attacker uses SDR hardware and a simple off-the-shelf antenna, from a distance.
4. The attacker decodes the signals. The radio signals are translated back into binary information.

SDR hardware is a radio receiver where software, not fixed circuits, does much of the work of tuning and decoding. That makes it flexible and fairly easy to get. The paper stresses that the antenna is a simple, off-the-shelf one, not custom gear.

Why this is called a covert channel

A covert channel is a hidden way to move information that the system was never designed to use. Your computer's RAM is meant to store data, not broadcast it. RAMBO turns it into a transmitter anyway.

That is why the paper describes an "out-of-band" threat. The data does not travel over Wi-Fi, Ethernet or Bluetooth. It travels over a side path that normal network tools may not watch.

The paper's abstract also says the work covers design, implementation, related work and evaluation results.

Source: arXiv abstract page for RAMBO (arXiv:2409.02292)

Air-Gapped RAM Leakage: What Kinds of Data Can Escape

The paper does not limit the attack to short text messages. It says malware can encode files, images, keylogging data, biometric information and encryption keys into the radio signals. You can read that list in the full paper on arXiv (v1).

That list matters because each item is valuable in a different way:

• Files can hold documents, source code or customer records.
• Images can hold scans, photos or screenshots.
• Keylogging data captures what someone types, including passwords.
• Biometric information covers data tied to a person's body, such as fingerprint or face templates.
• Encryption keys can unlock other protected data.

Encryption keys deserve special attention. A small key can open a lot of other information. A leaked key can turn a locked archive into an open one.

What 1,000 bits per second means in plain terms

The paper reports a rate of 1,000 bits per second. That is slow compared with a home internet connection. But the rate says nothing about how dangerous the data is.

Here is some simple arithmetic to give a sense of scale. These are my own back-of-the-envelope numbers, not results from the paper. They assume a perfect channel with no errors or extra overhead:

• 1,000 bits per second equals 125 bytes per second.
• A 256-bit encryption key is 32 bytes, or 256 bits. At the paper's rate, that is a fraction of a second of transmission.
• A 1-megabyte file is about 8 million bits. At 1,000 bits per second, that would take about 8,000 seconds, or a bit over two hours.

So the channel is a poor fit for stealing huge databases. It is a better fit for small, high-value items such as keys, passwords and short documents. Real-world conditions would likely slow things down, but the paper's own evaluation is the place to check that.

Why slow leaks still matter

Attackers often care more about being quiet than being fast. A slow channel that avoids normal network monitoring can be attractive. An air-gapped room that holds a few secrets may be worth hours of patient listening.

This is a reasoned implication, not a claim from the paper. The paper shows the channel works at the stated rate. Whether anyone uses it in the real world is a separate question that these sources do not answer.

Source: full paper on arXiv (v1) and arXiv abstract page for RAMBO

Mordechai Guri's RAMBO Paper: From NordSec 2023 to arXiv in 2024

The RAMBO paper has a single listed author, Mordechai Guri. It is filed under Cryptography and Security (cs.CR) on arXiv as arXiv:2409.02292. Its DOI is 10.48550/arXiv.2409.02292.

The paper's publication path is easy to mix up, so here is a clear timeline.

Three dates worth keeping straight

• 2023: The arXiv comments field says this is a "version of this work accepted to Nordic Conference on Secure IT Systems, 2023." The Ben-Gurion University research portal's record ties the work to Secure IT Systems / NordSec 2023. The timeline for this article dates that conference record to November 2023, with month-level precision only.
• 2024 (research record): The same university portal indexes the work as a 2024 research output. It links the paper to Springer LNCS volume 14324.
• September 3, 2024: Version 1 of the paper was submitted to arXiv, according to the arXiv submission history. The timestamp is 21:06:04 UTC.

So the arXiv posting did not introduce the work. It appeared after the conference acceptance and the 2023 conference record. Readers who see "2024" on arXiv should not assume the research itself began that year.

What each source supports

It helps to match each fact to the right page:

• The arXiv abstract page supports the paper's title, the 1,000 bits per second figure, the SDR and antenna setup, the conference acceptance comment and the September 3, 2024 submission date.
• The full paper (v1) supports the list of data types and the countermeasure discussion.
• The Ben-Gurion University research portal supports the indexing of the work as a 2024 output tied to NordSec 2023 and Springer LNCS volume 14324.

The portal page is a researcher profile. I used it for the indexing and volume details only. It does not add technical findings.

Source: arXiv abstract page for RAMBO and Ben-Gurion University research portal, Mordechai Guri

RAM Radio Exfiltration: What the Evidence Shows and What It Doesn't

Exfiltration means getting data out of a place where it should stay. In this case, the place is an air-gapped computer, and the exit is a radio signal from RAM. So far, the evidence for RAM radio exfiltration of this kind is the paper itself.

The evidence is mostly the author's own

The sources retrieved for this article are the researcher's paper and a university research record. They show the work exists, where it was accepted and what it claims. They are not independent tests by other labs.

That means there is no broad independent expert consensus in the material I reviewed. I could not find outside analysts weighing in with their own measurements or reviews, so I will not invent a consensus.

What the sources do consistently show is this:

• The paper presents a realistic covert-channel demonstration.
• It has practical defensive implications, since it proposes countermeasures.
• The strongest evidence is the paper and its conference indexing, not outside replication.

A conference acceptance means the work passed that conference's review. It is a useful signal. It is not the same as other teams repeating the experiment.

Limits of this article

This article relies on the paper's abstract, its listing details and the sources named above. I did not have details such as exact transmission distances, test setups or error rates, so I am not stating them. If you need those numbers for a risk assessment, read the full paper on arXiv directly.

Be careful with two shortcuts:

• Don't read "demonstrated" as "happening." A research demonstration shows something is possible. It does not show anyone is using it.
• Don't read "no consensus" as "no risk." It only means independent confirmation is thin so far.

Source: arXiv abstract page for RAMBO, full paper on arXiv (v1) and Ben-Gurion University research portal, Mordechai Guri

Faraday Enclosures and Other Ways to Shrink the Risk

The paper does not only describe the attack. It also discusses ways to reduce it. The full paper points to separation controls, including Faraday enclosures, to reduce radio-frequency leakage.

A Faraday enclosure is a conductive shell, often metal, that blocks or weakens radio waves passing in or out. If the signal cannot leave the enclosure, a receiver outside has little to pick up. This fits the problem, because RAMBO's whole channel is radio.

Why shielding fits this threat

Network firewalls cannot stop a radio signal that never touches the network. Shielding works on the physical path instead. It tackles the side channel directly.

The tradeoff is cost and convenience. Shielded rooms or cabinets are harder to set up than a software setting. They make sense only for the most sensitive systems.

Other layers (general practice, not from the paper)

The paper's countermeasure discussion, as retrieved here, names Faraday enclosures. The ideas below are general security practice that fits the same logic. They are my suggestions, not claims from the paper:

• Block the entry path. RAMBO needs malware on the machine first. Strict control over USB drives, software installs and who can touch the system lowers that chance.
• Keep distance and access limits. The attacker needs a receiver within range. Controlling who can be physically near sensitive machines reduces opportunities.
• Watch the environment. Monitoring of the physical space and of the machine's behavior can add warning signs. The paper's abstract does not claim a specific detection method, so treat this as a general principle.
• Separate roles. Keep your most valuable keys and files on as few machines as possible.

No single layer is perfect. The point of layering is that an attacker has to beat several of them.

Source: full paper on arXiv (v1)

What This Means for Startups and Small Businesses

Most startups do not run air-gapped computers. But some do, and many more keep a few high-value systems apart on purpose. Think of a machine that stores signing keys, holds backups of customer data or runs a lab instrument.

Here is how to turn the RAMBO research into action without overreacting.

1. Find out whether you rely on an air gap

Ask your team which systems are kept offline for safety. Write them down. If the answer is "none," RAMBO is mostly a lesson in how security claims can be too simple.

2. Treat "air-gapped" as one layer, not a finish line

If you do use an air gap, ask what else protects that machine. The paper's message, in practical terms, is that isolation from the network is not the same as isolation from the world.

3. Guard the way in

RAMBO starts with malware already installed. So the most practical defense is making that hard:

• Limit and log removable media.
• Lock down who can install software.
• Check devices and files before they enter a secure area.
• Keep the secure machine's purpose narrow.

4. Consider physical separation for your crown jewels

For the few systems that would truly hurt if leaked, ask whether shielding is worth it. The paper names Faraday enclosures as one option. For many small firms, a simpler step is to keep those machines in a controlled room away from public or shared spaces.

5. Ask your vendors and advisers sharper questions

If a vendor says its product is "air-gapped" or "isolated," ask what that covers. Does it address physical and radio-based leakage, or only network access? Those are different questions, and the answers show how deep the vendor's thinking goes.

6. Keep the risk in proportion

A research paper is not an active attack on your business. The sources here show a demonstrated method, not proof of real-world use. Spend effort where your actual risk is highest. For most small companies, phishing, weak passwords and unpatched software still come first.

7. Write it into your security plan

Add a line to your risk register. Note that air-gapped systems can leak through covert channels, list the systems it applies to and name who reviews it. Revisit it when more independent research appears.

The practical summary: you do not need to rebuild your office. You do need to stop treating "offline" as "safe."

Source: arXiv abstract page for RAMBO and full paper on arXiv (v1)

Frequently Asked Questions

Source: arXiv abstract page for RAMBO, full paper on arXiv (v1) and Ben-Gurion University research portal, Mordechai Guri