A serious flaw in Zoom’s Keybase secure chat application left copies of images contained in secure communications on Keybase users’ computers after they were supposedly deleted.

The flaw in the encrypted messaging application (CVE-2021-23827) does not expose Keybase users to remote compromise. However, it could put their security, privacy and safety at risk, especially for users living under authoritarian regimes in which apps like Keybase and Signal are increasingly relied on as a way to conduct conversations out of earshot of law enforcement or security services.

The flaw was discovered by researchers from the group Sakura Samurai as part of a bug bounty program offered by Zoom, which acquired Keybase in May, 2020. Zoom said it has fixed the flaw in the latest versions of its software for Windows, macOS and Linux.

Deleted…but not gone

According to researcher John Jackson of Sakura Samurai, the Keybase flaw manifested itself in two ways. First: Jackson discovered that images that were copy and pasted into Keybase chats were not reliably deleted from a temporary folder, /uploadtemps, associated with the client application. “In general, when you would copy and paste in a Keybase chat, the folder would appear in (the uploadtemps) folder and then immediately get deleted,” Jackson told Security Ledger in a phone interview. “But occasionally that wouldn’t happen. Clearly there was some kind of software error – a collision of sorts – where the images were not getting cleared.”

Exploitable Flaw in NPM Private IP App Lurks Everywhere, Anywhere

Discovering that flaw put Sakura Samurai researchers on the hunt for more and they soon struck pay dirt again. Sakura Samurai members Aubrey Cottle (@kirtaner), Robert Willis (@rej_ex) and Jackson Henry (@JacksonHHax) discovered an unencrypted directory, /Cache, associated with the Keybase client that contained a comprehensive record of images from encrypted chat sessions. The application used a custom extension to name the files, but they were easily viewable directly or simply by changing the custom file extension to the PNG image format, Jackson said.

In a statement, a Zoom spokesman said that the company appreciates the work of the researchers and takes privacy and security “very seriously.”

“We addressed the issue identified by the Sakura Samurai researchers on our Keybase platform in version 5.6.0 for Windows and macOS and version 5.6.1 for Linux. Users can help keep themselves secure by applying current updates or downloading the latest Keybase software with all current security updates,” the spokesman said.

Podcast Episode 141: Massive Data Breaches Just Keep Happening. We Talk about Why.

In most cases, the failure to remove files from cache after they were deleted would count as a “low priority” security flaw. However, in the context of an end-to-end encrypted communications application like Keybase, the failure takes on added weight, Jackson wrote.

“An attacker that gains access to a victim machine can potentially obtain sensitive data through gathered photos, especially if the user utilizes Keybase frequently. A user, believing that they are sending photos that can be cleared later, may not realize that sent photos are not cleared from the cache and may send photos of PII or other sensitive data to friends or colleagues.”

Messaging app flaws take on new importance

The flaw takes on even more weight given the recent flight of millions of Internet users to end-to-end encrypted messaging applications like Keybase, Signal and Telegram. Those users were responding to onerous data sharing policies, such as those recently introduced on Facebook’s WhatsApp chat. In countries with oppressive, authoritarian governments, end to end encrypted messaging apps are a lifeline for political dissidents and human rights advocates.

As Cybercrooks Specialize, More Snooping, Less Smash and Grab

As a result of the flaw, however, adversaries who gained access to the laptop or desktop on which the Keybase application was installed could view any images contained in Keybase encrypted chats. The implications of that are clear enough. For example, recent reports say that North Korean state hackers have targeted security researchers via phishing attacks sent via Keybase, Signal and other encrypted applications.

The flaws in Keybase do not affect the Zoom application, Jackson said. Zoom acquired Keybase in May to strengthen the company’s video platform with end-to-end encryption. That acquisition followed reports about security flaws in the Zoom client, including in its in-meeting chat feature.

Jackson said that the Sakura Samurai researchers received a $1,000 bounty from Zoom for their research. He credited the company with being “very responsive” to the group’s vulnerability report.

The increased use of encrypted messaging applications has attracted the attention of security researchers, as well. Last week, for example, a researcher disclosed 13 vulnerabilities in the Telegram secure messaging application that could have allow a remote attacker to compromise any Telegram user. Those issues were patched in Telegram updates released in September and October, 2020.

In this episode of the podcast (#200), sponsored by Digicert: John Jackson, founder of the group Sakura Samurai talks to us about his quest to make hacking groups cool again. Also: we talk with Avesta Hojjati of the firm Digicert about the challenge of managing a growing population of digital certificates and how  automation may be an answer.


Life for independent security researchers has changed a lot in the last 30 years. The modern information security industry grew out of pioneering work by groups like Boston-based L0pht Heavy Industries and the Cult of the Dead Cow, which began in Lubbock, Texas.

After operating for years in the shadows of the software industry and in legal limbo, by the turn of the millennium hackers were coming out of the shadows. And by the end of the first decade of the 21st century, they were free to pursue full fledged careers as bug hunters, with some earning hundreds of thousands of dollars a year through bug bounty programs that have proliferated in the last decade.

Despite that, a stigma still hangs over “hacking” in the mind of the public, law enforcement and policy makers. And, despite the growth of bug bounty programs, red teaming and other “hacking for hire” activities, plenty of blurry lines still separate legal security research from illegal hacking. 

Hacks Both Daring…and Legal

Still, the need for innovative and ethical security work in the public interest has never been greater. The Solar Winds hack exposed the ways in which even sophisticated firms like Microsoft and Google are vulnerable to compromised software supply chain attacks. Consider also the tsunami of “smart” Internet connected devices like cameras, television sets and appliances are working their way into homes and workplaces by the millions. 

Podcast Episode 112: what it takes to be a top bug hunter

John Jackson is the co -founder of Sakura Samurai, an independent security research group. 

What does a 21st century hacking crew look like? Our first guest this week is trying to find out. John Jackson (@johnjhacking) is an independent security researcher and the co-founder of a new hacking group, Sakura Samurai, which includes a diverse array of security pros ranging from a 15 year old Australian teen to Aubrey Cottle, aka @kirtaner, the founder of the group Anonymous. Their goal: to energize the world of ethical hacking with daring and attention getting discoveries that stay on the right side of the double yellow line.

Update: DHS Looking Into Cyber Risk from TCL Smart TVs

In this interview, John and I talk about his recent research including vulnerabilities he helped discover in smart television sets by the Chinese firm TCL, the open source security module Private IP and the United Nations. 

Can PKI Automation Head Off Chaos?

One of the lesser reported sub plots in the recent Solar Winds hack is the use of stolen or compromised digital certificates to facilitate compromises of victim networks and accounts. Stolen certificates played a part in the recent hack of Mimecast, as well as in an attack on employees of a prominent think tank, according to reporting by Reuters and others. 

Avesta Hojjati is the head of Research & Development at Digicert.

How is it that compromised digital certificates are falling into the hands of nation state actors? One reason may be that companies are managing more digital certificates than ever, but using old systems and processes to do so. The result: it is becoming easier and easier for expired or compromised certificates to fly under the radar. 

Our final guest this week, Avesta Hojjati, the  Head of R&D at DigiCert, Inc. thinks we’ve only seen the beginning of this problem. As more and more connected “things” begin to populate our homes and workplaces, certificate management is going to become a critical task – one that few consumers are prepared to handle.

Episode 175: Campaign Security lags. Also: securing Digital Identities in the age of the DeepFake

What’s the solution? Hojjati thinks more and better use of automation is a good place to start. In this conversation, Avesta and I talk about how digital transformation and the growth of the Internet of Things are raising the stakes for proper certificate management and why companies need to be thinking hard about how to scale their current certificate management processes to meet the challenges of the next decade. 


(*) Disclosure: This podcast was sponsored by Digicert. For more information on how Security Ledger works with its sponsors and sponsored content on Security Ledger, check out our About Security Ledger page on sponsorships and sponsor relations.

As always,  you can check our full conversation in our latest Security Ledger podcast at Blubrry. You can also listen to it on iTunes and check us out on SoundCloudStitcherRadio Public and more. Also: if you enjoy this podcast, consider signing up to receive it in your email. Just point your web browser to securityledger.com/subscribe to get notified whenever a new podcast is posted. 

Neopets, a website that allows children to care for “virtual pets,” has exposed a wide range of sensitive data online including credentials needed to access company databases, employee emails, and even repositories containing the proprietary code for the site, according to information shared with The Security Ledger.

The data includes the IP addresses of Neopets visitors, information that could be used to target Neopets users, according to independent researcher John Jackson, who said he discovered the information after scanning the company’s website with a security tool.

Stolen Accounts For Sale

Neopets is a “virtual pet website” that first launched in 1999. It permits users – many of them children – to care for virtual pets and buy virtual items for them using virtual points earned in-game (Neopoints) or with “Neocash” that can be purchased with real-world money, or won in-game. Purchased by Viacom for $160 million in 2005, in 2017, it was acquired by the Chinese company NetDragon.

In an email to The Security Ledger, Jackson said that he noticed Neopets accounts being offered for sale on an online forum. That prompted him to run a scan on the Neopets site using a forensics tool. That scan revealed a Neopets subdomain that exposed the guts of the Neopets website, Jackson said via instant message.

China Using Big Brother-Like System to Track, Monitor Minorities

“We looked through and found employee emails, database credentials and their whole codebase,” he said.

Jackson shared screen shots of the Neopets directory as well as snippets of code captured from the site that suggest credentials were “hard coded,” or embedded in the underlying code of the website. Working with security researcher Nick Sahler, Jackson was able to download Website’s entire codebase, revealing database credentials, employee emails, user IP addresses and private code repositories. The two researchers also uncovered internal IP addresses and the underlying application logic for the entire Neopets application.

Snippet of code from the NeoPets website showing hard coded credentials. (Image courtesy of John Jackson.)
Snippet of code from the Neopets website showing hard coded credentials. (Image courtesy of John Jackson.)

“This is extremely bad because even though we didn’t attempt to access PII (personally identifying information), with these codebases we can undoubtedly do so,” Jackson said. “They need to fix the root issues, otherwise they will suffer yet another threat-actor related breach.”

U.S. Customs Data Breach Is Latest 3rd-Party Risk, Privacy Disaster

Jackson and Sahler said they have reported their findings to Neopets and provided copies of email exchanges with a support tech at the company who said he would pass the issue to “one of our coders.”

Neopets has not yet responded to requests for comment on the researchers’ allegations.

If true, this would be the second serious security incident involving the Neopets site. In 2016, the company acknowledged a breach that spilled usernames, passwords, IP addresses and other personal information for some 27 million users. That breach may have occurred as early as 2013, according to the website HaveIbeenPwned.

The issue appears to be related to a misconfigured Apache web server, Jackson said. Though many web-based applications are hosted on infrastructure owned by cloud providers such as Amazon, Google or Microsoft’s Azure, leaked documents indicate that the 20 year-old Neopets website continues to operate from infrastructure it owns and operates.

Episode 145: Veracode CTO Chris Wysopal and Life After Passwords with Plurilock

Misconfigured web servers are a frequent source of security breaches -whether self-hosted or hosted by a third party. In 2017, credit rating agency Equifax acknowledged that a hole in the Apache Struts platform first identified in March, 2017 and patched in August of that year was used by hackers to compromise a web application and gain access to the information which included names, email addresses and, for US residents, Social Security Numbers. The vulnerability, identified as CVE-2017-5638, was associated with a string of attacks in 2017 and 2018.

High Bar for Collecting Information on Children

The breach could spell legal trouble for Neopets andWebsites and NetDragon. Online firms that manage information on children are held to a high standard under the federal Children’s Online Privacy Protection Act (“COPPA”).

In June, the U.S. Federal Trade Commission (FTC) announced that it reached a settlement with children’s mobile application developer HyperBeard Inc. that included a $4 million fine for COPPA violations for obtaining parental consent before processing children’s personal information for targeted advertising. (HyperBeard ultimately paid just $150,000 of that penalty, citing an inability to pay the full amount.)

In September, 2019 Google and its YouTube subsidiary agreed to pay a record $170 million fine to settle allegations by the Federal Trade Commission and the New York Attorney General that the YouTube video sharing service violated COPPA by illegally collecting personal information from children without their parents’ consent.