'Adversarial DNA' breeds buffer overflow bugs in PCs

Boffins had to break gene-reading software but were able to remotely exploit a computer


Scientists from the University of Washington have created synthetic DNA that produced malware of a sort.

Detailed in a paper titled “Computer Security, Privacy, and DNA Sequencing: Compromising Computers with Synthesized DNA, Privacy Leaks, and More”, the authors explain that they decided to “synthesize DNA strands that, after sequencing and post-processing, generated a file; when used as input into a vulnerable program, this file yielded an open socket for remote control.”

Lead author computer scientist Tadayoshi Kohno admits that malware-laced DNA isn't a realistic threat, but suggests there's "potential" in the future for it to be used by hackers planning to steal developing drug IP or ransom off sensitive genomic data.

"It's important to understand the security risks before they manifest," Kohno told The Register. "There aren't bad guys knocking at the door to compromise the software systems developed by this community."

University of Washington DNA Security Research

From left, Lee Organick, Karl Koscher and Peter Ney prepare the DNA exploit for sequencing. Credit: Dennis Wise/University of Washington

To make it work, the authors got their hands on the source code of open-source DNA compressor fqzcomp and “inserted a vulnerability into version 4.6 of its source code; a function that processes and compresses DNA reads individually, using a fixed-size buffer to store the compressed data.”

Biologists use fqzcomp to compress digital files containing strings of A, C, T, G. A simple computer command was translated into 176 DNA letters and, after ordering copies of this DNA from a vendor, the researchers fed it to a sequencing machine. After some processing, this digital file is fed to the compression program.

They manually modified the code to decrease the size of the DNA sequence input it was expecting, allowing the command in the DNA to spill over (a buffer overflow attack) and execute a program connected to a remote server with complete control over the machine. About 37.4 per cent of all reads ended up pulling off the buffer overflow.

Readers may at this point think that it's pretty easy to break software when you feed it data that you know in advance will cause it problems. The researchers recognise this, writing that they know their crocked code is “in many ways the 'best possible' environment for an adversary.”

But they also note that “fqzcomp already contains over two dozen static buffers. Our modifications added 54 lines of C++ code and deleted 127 lines from fqzcomp.” The Register imagines that kind of modification could go unnoticed in many-a-lab.

For example, the researchers also evaluated 13 open-source programs for DNA processing and found the same kind of insecurities.

But the paper also points out that synthesising any DNA, never mind stuff designed to disrupt bioinformatics software, is hard and prone to error. Even if you can do the job, you need to get the right sample into the right lab, and need to know what software that lab is running. Or get malformed software into that lab.

All of which is hard. But so was getting Stuxnet across an air-gap into an Iranian centrifuge.

Pass the password

Yaniv Erlich, a computer scientist at Columbia University in New York City who has studied DNA storage but was not involved in the study, told The Register that the exploit "is basically unrealistic". He says a hacker would have to precisely time the DNA exploit to when a sequencing center runs a new compression tech with longer reads for the very first time, but any centre would "rigorously test" a pipeline when any new tech arrives "so short buffers are likely to be detected ahead of time."

He said he's more worried about academic centres using default passwords on their internet-connected DNA sequencers, which could be targets for ransomware.

But Christophe Dessimoz, a computational biologist at University of Lausanne in Switzerland who was also not involved in the study but has studied DNA storage, told The Register that the authors are correct that the "general security hygiene of bioinformatics programs is very low".

"Current tools have been developed to handle natural DNA sequences, which are not expected to contain code executable by computers," he added. Although not surprising, "The work is a nifty stunt."

Someone needs to be thinking about security in the DNA sequencing ecosystem

The authors' main recommendation is that bioinformatics software just hasn't been written with this kind of attack in mind, but seeing as DNA is information encoded in chemicals the authors of such software should wise up to the risks they've demonstrated.

Kohno, who is next studying regulations around DNA sequencing and analysis pipelines, believes it's possible there are other exploits that could take advantage of DNA. He said that while some researchers might not find it surprising now, "hindsight is always 2020".

You can find the paper here [PDF] and the University's explainer and FAQ here. The team will present their peer-reviewed research at next week's annual USENIX Security Symposium in Vancouver, British Columbia, Canada

The second document tries hard to point out that this is all theoretical. “We have no evidence to believe that the security of DNA sequencing or DNA data in general is currently under attack,” the primer says. “Instead, we view these results as a first step toward thinking about computer security in the DNA sequencing ecosystem.” ®

Bootnote

Here, the University of Washington's interdisciplinary team of biologists, computer security and DNA processing pipeline experts used synthetic DNA to take down a system. But is it possible that natural human DNA could also accidentally take down a biological research computer system someday?

We asked some of the researchers if it would be possible.

They thought about it and said it's "unlikely"...

Similar topics


Other stories you might like

  • Chinese-sponsored gang Gallium upgrades to sneaky PingPull RAT
    Broadens targets from telecoms to finance and government orgs

    The Gallium group, believed to be a Chinese state-sponsored team, is going on the warpath with an upgraded remote access trojan (RAT) that threat hunters say is difficult to detect.

    The deployment of this "PingPull" RAT comes as the gang is broadening the types of organizations in its sights from telecommunications companies to financial services firms and government entities across Asia, Southeast Asia, Europe and Africa, according to researchers with Palo Alto Networks' Unit 42 threat intelligence group.

    The backdoor, once in a compromised system, comes in three variants, each of which can communicate with the command-and-control (C2) system in one of three protocols: ICMP, HTTPS and raw TCP. All three PingPull variants have the same functionality, but each creates a custom string of code that it sends to the C2 server, which will use the unique string to identify the compromised system.

    Continue reading
  • Symantec: More malware operators moving in to exploit Follina
    Meanwhile Microsoft still hasn't patched the fatal flaw

    While enterprises are still waiting for Microsoft to issue a fix for the critical "Follina" vulnerability in Windows, yet more malware operators are moving in to exploit it.

    Microsoft late last month acknowledged the remote code execution (RCE) vulnerability – tracked as CVE-2022-30190 – but has yet to deliver a patch for it. The company has outlined workarounds that can be used until a fix becomes available.

    In the meantime, reports of active exploits of the flaw continue to surface. Analysts with Proofpoint's Threat Insight team earlier this month tweeted about a phishing campaign, possibly aligned with a nation-state targeting US and European Union agencies, which uses Follina. The Proofpoint researchers said the malicious spam messages were sent to fewer than 10 Proofpoint product users.

    Continue reading
  • EnemyBot malware adds enterprise flaws to exploit arsenal
    Fast-evolving botnet targets critical VMware, F5 BIG-IP bugs, we're told

    The botnet malware EnemyBot has added exploits to its arsenal, allowing it to infect and spread from enterprise-grade gear.

    What's worse, EnemyBot's core source code, minus its exploits, can be found on GitHub, so any miscreant can use the malware to start crafting their own outbreaks of this software nasty.

    The group behind EnemyBot is Keksec, a collection of experienced developers, also known as Nero and Freakout, that have been around since 2016 and have launched a number of Linux- and Windows-based bots capable of launching distributed denial-of-service (DDoS) attacks and possibly mining cryptocurrency. Securonix first wrote about EnemyBot in March.

    Continue reading
  • Clipminer rakes in $1.7m in crypto hijacking scam
    Crooks divert transactions to own wallets while running mining on the side

    A crew using malware that performs cryptomining and clipboard-hacking operations have made off with at least $1.7 million in stolen cryptocurrency.

    The malware, dubbed Trojan.Clipminer, leverages the compute power of compromised systems to mine for cryptocurrency as well as identify crypto-wallet addresses in clipboard text and replace it to redirect transactions, according to researchers with Symantec's Threat Intelligence Team.

    The first samples of the Windows malware appeared in January 2021 and began to accelerate in their spread the following month, the Symantec researchers wrote in a blog post this week. They also observed that there are several design similarities between Clipminer and KryptoCibule – another cryptomining trojan that, a few months before Clipminer hit the scene, was detected and written about by ESET analysts.

    Continue reading

Biting the hand that feeds IT © 1998–2022