IT NEWS

A researcher has found a way to generate a fingerprint of your device from your installed Google Chrome extensions, and then use that fingerprint to track you online.

Fingerprinting is a way of figuring out what makes your device unique and then using that to identify you as you move around the internet. Websites you visit receive a huge amount of information when you land on their portal—it’s a lot more than “just” which web browser you use to load up someone’s site.

What extensions do you have? How does your screen resolution compare with others? If you use a specific, unusual resolution, do you run other extensions alongside it? Do other people? Which versions of those extensions are on board? Is your IP address plain and exposed, or hidden behind a VPN?

How do sites fingerprint my device?

You can see a typical voluntary form of fingerprinting testing here. The site checks for a variety of information related to your device (including the below), and then places a cookie on your PC for four months:

• the User agent header
• the Accept header
• the Connection header
• the Encoding header
• the Language header
• the Upgrade Insecure Requests header
• the Referer header
• the Cache-Control header
• the BuildId of the browser
• the list of plugins
• the platform
• the cookies preferences (allowed or not)
• the Do Not Track preferences (yes, no or not communicated)
• the timezone
• the screen resolution and its color depth

What you often see in tests like this is a high degree of similarity between users for things like content encoding, preference for secure HTTPs requests, supported video formats, and so on.

The numbers start to flatten out for aspects of your PC like plugins, adblocker use, media devices plugged in, and lists of fonts. As you can see, it’s not just that fingerprinting can tell you what browser you use or your screen resolution at a very basic level, it’s all of the additional components too.

There’s lots of ways fingerprinting can provide a very in-depth profile of a device.

You may use one type of browser like 50% of the other people who had their system fingerprinted. However, only 5% may use a specific version of that browser. Of that 5%, only 2% have a certain extension installed. From there, only 0.3% may use a specific version of this extension. And so it goes on…

Even switching your browsers around may not help much, which leads to people coming up with all sorts of workarounds.

Running the gauntlet of web accessible resources

The site determines installed extensions thanks to something called “web accessible resources”. As the researcher explains:

Web-accessible resources are files inside an extension that can be accessed by web pages or other extensions. Extensions typically use this feature to expose images or other assets that need to be loaded in web pages, but any asset included in an extension’s bundle can be made web accessible.

By default no resources are web accessible; only pages or scripts loaded from an extension’s origin can access that extension’s resources. Extension authors can use the web_accessible_resources manifest property to declare which resources are exposed and to what origins.

A webpage can successfully fetch an installed extensions web accessible resource. If the fetch fails it usually means that the extension is not installed.

Visiting the checker site returns a list of potential Chrome extensions, and each entry has a True/False detection flag. In my case, it correctly reported the installed extensions on the test system and informed me what % of users share those extensions.

The project creator explains that the detection does not work for Firefox as “Firefox extension IDs are unique for every browser instance”. They go on to say that the site “only detects extensions from the Chrome web store. Extensions for [Microsoft Edge] can be detected using the same methods but are not supported by this tool”.

Tackling evasive behaviour

Some extensions have ways of not showing up in this kind of fingerprinting test. Are some of the extensions on your device trying to hide? Thanks to something called “Resource timing comparison”, it may not even matter.

In an effort to prevent detection some extensions will generate a secret token thats required to access their web accessible resources. Any fetch operation made without the secret token will result in failure. Although its much more difficult to detect these protected extensions, it’s still possible.

Resources of protected extensions will take longer to fetch than resources of extensions that are not installed. By comparing the timing differences you can accurately determine if the protected extensions are installed.

Avoiding fingerprinting

There’s numerous suggestions for this, but not all of them may be practical for you in your day to day dealings. Suggestions from the Electronic Frontier Foundation include:

• Using a “non-rare” browser, with the caveat that aspects such as fonts and plugins can easily make you identifiable.
• Disabling JavaScript, with the additional caveat that this may break functionality for most websites.
• Making use of the private browsing modes included in most web browsers.

You could also use browsers with dedicated anti-fingerprinting technology running in the background. Whatever you decide, this is by no means an easy problem to address for most people.

The post You can be tracked online using your Chrome browser extensions appeared first on Malwarebytes Labs.

This blog post was authored by Jérôme Segura

We have seen and heard less buzz about ‘Magecart’ during the past several months. While some marketing playbooks continue to rehash the same breaches of yesteryear, we have been wondering if some changes took place in the threat landscape.

One thing we know is that if the Magecart threat actors decided to switch their operations exclusively server-side then the majority of companies, including ours, would lose visibility overnight. This is why we often look up to researchers that work the website cleanups. If something happens, these guys would likely notice it.

We followed the trail on two recent reports that proved to be worthwhile. It allowed us to make a connection to a previous campaign and identify new pieces of a pretty wide infrastructure.

For now we can say that Magecart client-side attacks are still around and that we could easily be missing them if we rely on automated crawlers and sandboxes, at least if we don’t make them more robust.

Newly reported domains linked with ‘anti-VM’ skimmer

On June 12, @rootprivilege tweeted about a hacked stored injected with the host js.staticounter[.]net that looked highly suspicious. When originally captured, the JavaScript appeared to be clean but it was confirmed to be malicious by @AffableKraut who posted a screenshot of the skimmer code.

A few days before @rootprivilege posted about this skimmer, @Sansec tweeted about another new skimmer domain at scanalytic[.]org. Comparing the two which are both on the same ASN (AS29182), we concluded that they are related.

We were able to connect these 2 domains with a previous campaign from November 2021 which was the first instance to our knowledge of a skimmer checking for the use of virtual machines. However, both of them are now devoid of VM detection code. It’s unclear why the threat actors removed it, unless perhaps it caused more issues than benefits.

There are other differences with the newest skimmer sample from @rootsecdev such as different naming schemes for important input fields. As you can see, in the former case these are explicitly referenced (i.e. CcNumber) while in the later iteration the names are generic web terms, making them less obvious.

Additional infrastructure

Using the urlscan.io service, we were able to discover additional infrastructure related to this ongoing campaign. We started our search with any recent submissions that made contact with an IP address belonging to AS29182.

The table below shows hostnames, their IP address and the date they were first seen on urlscan.io. Most of those were previously unknown to us until we recently started this investigation. You can click on the hyperlinks to load the corresponding sandbox pages, but note that a majority of them do not contain the actual skimmer code. This is most likely because the malicious infrastructure detected that urlscan.io’s sandbox was not using genuine residential IP addresses.

Validating skimmer activity

For the domains that are still responding, we can use information collected by urlscan.io and replay the attack using a genuine residential IP address and mimicking a real shopper’s experience. The image below shows the difference between a crawler session via VPN and one done manually with real network settings.

This allows us to confirm beyond doubt that the domains are indeed malicious, although their ASN should already be enough to proactively block them.

Connection with previous skimmer activity

Based on one hash, we can connect these skimmers to past activity going back to at least May 2020. One of the hostnames from our previous blog on the anti-vm skimmer, con[.]digital-speed[.]net, was loading this resource as well.

We can see 3 different themes used by the threat actor to hide their skimmer, named after JavaScript libraries:

• hal-data[.]org/gre/code.js (Angular JS)
• hal-data[.]org/data/ (Logger)
• js.g-livestatic[.]com/theme/main.js (Modernizr)

Less skimmer activity or simply more covert?

There are likely many more skimmer domains on the infrastructure we detailed above, and it is a good idea to keep a close eye on it. Having said that, we have generally seen less skimming attacks during the past several months. Perhaps we have been too focused on the Magento CMS, or our crawlers and sandboxes are being detected because of various checks including at the network level.

As Ben Martin over at Sucuri showed, WordPress with the WooCommerce plugin is outpacing Magento in terms of attacks. In addition, we (as several other companies) can only observe client-side attacks and as such we are oblivious to what happens server-side. Only a handful of researchers who do website cleanups have the visibility into PHP-based skimmers.

While stealing credit cards is still a good business, there are other types of data considerably more worth it. Crypto wallets and similar digital assets are extremely valuable and there is no doubt that clever schemes to rob those are in place beyond phishing for them. For an example of a client-side attack via JavaScript draining crypto assets, check out this blog from Eliya Stein over at Confiant.

Malwarebytes customers are protected against this campaign.

Indicators of Compromise

Skimmer domains

Skimmer IPs

The post Client-side Magecart attacks still around, but more covert appeared first on Malwarebytes Labs.

Last week on Malwarebytes Labs:

• Serious vulnerabilities found in ITarian software, patches available for SaaS products
• Update Chrome now: Four high risk vulnerabilities found
• Taking down the IP2Scam tech support campaign
• Don’t panic! “Unpatchable” Mac vulnerability discovered
• Introducing Malwarebytes Vulnerability Assessment for OneView: How to check for Common Vulnerabilities and Exposures (CVEs)
• “Multiple adversaries” exploiting Confluence vulnerability, warns Microsoft
• Instagram scam steals your selfies to trick your friends
• Karakurt extortion group: Threat profile
• Email compromise leads to healthcare data breach at Kaiser Permanente
• It’s official, today you can say goodbye to Internet Explorer. Or can you?
• Update now!  Microsoft patches Follina, and many other security updates
• Firefox stops advertisers tracking you as you browse, calls itself the most “private and secure major browser”
• Record breaking HTTPS DDoS attack
• Stealthy Symbiote Linux malware is after financial institutions
• Photos of kids taken from spyware-ridden phones found exposed on the internet
• Interpol’s First Light operation smashes crime on a global scale
• Hertzbleed exposes computers’ secret whispers
• ALPHV squeezes victim with dedicated leak site for employees and customers
• Securing the software supply chain, with Kim Lewandowski: Lock and Code S03E13

Stay safe!

The post A week in security (June 13 – June 19) appeared first on Malwarebytes Labs.

Digital currency fraud is a growing issue on social media, and LinkedIn is no different. In fact, according to according to Sean Ragan, the FBI’s special agent in charge of the San Francisco and Sacramento, California, field offices, cryptocurrency scams are big business on LinkedIn.

“It’s a significant threat. This type of fraudulent activity is significant, and there are many potential victims, and there are many past and current victims.”

How cryptocurrency scams work on LinkedIn

Aspects of LinkedIn cryptocurrency scams share similar traits with fraud attempts on other platforms:

• Someone messages you out of the blue. They begin with small talk, and eventually work their way up to cryptocurrency conversation. They claim that, yes, they can help you make big money from certain investments.
• LinkedIn is generally seen as a trusted platform, reinforced by people’s perception as the go-to place for business related dealings. This is one advantage it has over less formal sites.
• Victims are directed to genuine cryptocurrency investment portals. Though no further details are provided in the article, this can go one of two ways. Either the victim invests with their own cash, or the scammer sends them some funds to get started.
• Weeks or months down the line, the scammer has the victim transfer funds to a site controlled by the scammer. At this point, funds are drained and the cash disappears along with the con-artist.

Scammers take the well-worn path to riches

The FBI notes that this type of fraud is on the rise, and draws a parallel with romance scams. In both cases, the end result is the same: loss of funds. However, this style of cryptocurrency fraud has its origins elsewhere and the connection to romance fraud is quite relevant.

This style of attack is called the “pig butcher” scam. It involves a so-called “fattening up” of the pig (target) with messages of affection. Eventually, the same jump-off into cryptocurrency investment takes place. The money, as always, vanishes. One of the key features of this attack is the pretence of accidental communication. Golf is popular, as are messages about luggage and airports.

The tactics used on LinkedIn almost certainly match up in various ways. If they can just get you to the investment site and have you deposit some funds: they’ve got you.

Linkedin take fraudsters to task

The team at LinkedIn point out that 96% of detected fake accounts and 99.1% of spam and scams are caught and removed by automated defences. That’s somewhere in the region of 70 million scam messages removed between July to December in 2021. For comparison, LinkedIn removed around 60 million between January and June of 2019. It also hit a peak of removals between July to December of 2020, with a massive 91 million scams given a time out.

Additionally, 11.9 million fake accounts were stopped at registration between July and December of 2021. Around 4.4 million were restricted proactively, and 127k further accounts were restricted once members reported them.

How to spot a scam on LinkedIn

With regard to cryptocurrency scams themselves, LinkedIn offers the following advice. Be wary of:

• People asking for money who you don’t know in person. This may include sending cash directly, cryptocurrency, gift cards, prizes, and other winnings.
• Job postings which sound too good to be true. Mystery shoppers, personal assistants, company impersonators are all potential red flags. Steer clear of anything which demands money from you up front.
• Romantic gestures on a business-centric platform. This is especially dubious if tied to a brand new account with few or no connections. Keep in mind that established accounts can also be compromised, and used for any of the scam attempts listed above.

Should you experience LinkedIn content you’re not sure about, don’t worry. You can report it directly to LinkedIn to investigate. Stay safe out there!

The post LinkedIn scams are a “significant threat”, warns FBI appeared first on Malwarebytes Labs.

Matthew Gatrel, a 33-year-old man from St. Charles, Illinois, has been sentenced to two years in prison for running websites that provide powerful distributed denial-of-service (DDoS) attacks against internet users and websites. This sentencing resulted in the seizure of his websites, making the internet a little safer from DDoS attacks.

Gatrel was the administrator and owner of DownThem.org and AmpNode.com, two DDoS-for-hire websites with thousands of clients which launched attacks against more than 200,000 targets. He was convicted of three charges, including conspiracy to commit unauthorized impairment of a protected computer, conspiracy to commit wire fraud, and unauthorized impairment of a protected computer.

“Gatrel ran a criminal enterprise designed around launching hundreds of thousands of cyberattacks on behalf of hundreds of customers,” prosecutors wrote in a sentencing memorandum. More from that memorandum:

“He also provided infrastructure and resources for other cybercriminals to run their own businesses launching these same kinds of attacks. These attacks victimized wide swaths of American society and compromised computers around the world.”

Prosecutors said that DownThem.org was a subscription-based service that allowed paying customers to launch DDoS attacks at targets of their choice.

AmpNode.com was a “bulletproof” server hosting service provider “with an emphasis on ‘spoofing’ servers that could be pre-configured with DDoS attack scripts and lists of vulnerable ‘attack amplifiers’ used to launch simultaneous cyberattacks on victims”.

Gatrel’s services helped launch attacks against targets worldwide, including homes, schools, universities, financial institutions, and local government websites. Many clients of AmpNode also operated DDoS-for-hire services.

Prosecutors also said that Gatrel offered expert advice and guidance to clients of both services, ranging from different methods to “down” different types of computers to bypassing DDoS protection services. To get potential clients to buy in, he used DownThem to launch a DDoS attack against these clients’ intended victims and provide proof that their internet connection had been severed.

Juan “Severon” Martinez from Pasadena, California, Gatrel’s co-defendant and criminal partner, pleaded guilty to the unauthorized impairment of a protected computer. He was sentenced to five years’ probation.

The post DDoS-for-hire service provider jailed appeared first on Malwarebytes Labs.

Going on vacation has never been more talked about and anticipated. I mean—for many of us, it’s been a while.

But before you get lost in dreamy thoughts of sun, sea, and sand, you might want to set aside some time to plan on how to keep your devices, and your data, safe while you are relaxing

Your devices need some prepping, too

Before anything else, know which devices you’ll bring and which ones you’ll leave at home. Then make backups of the files in them.

This is also the perfect time to look deeper into what’s on your devices, especially if you haven’t done any spring cleaning due to busyness. So update those apps that need updating and uninstall those that waste space; scan your devices with a trusty malware scanner, and change any duplicate passwords. Then follow these tips:

7 security and privacy tips that fit in your pocket

Ensure your devices have the “Find My Device” feature enabled. This feature isn’t just limited to Apple products, and can really help if you lose your device. You can remotely wipe a device if you lose it or even put a message on the screen with contact details in case it is found.

Be mindful of seasonal scams. Such scams may arrive via email, SMS, or social media. If a service offers rates that are too good to be true, asks for an upfront fee, or demands payments to be wired, avoid it.

Use 2FA. Make sure you lock your accounts behind two-factor authentication (2FA). This additional security measure makes them harder to compromise should someone get hold of your login details.

Turn off Bluetooth connectivity. Many people forget Bluetooth is there. As a rule of thumb, remove it if you don’t use it. But if you can’t, disable it when it’s not in use.

Leave your device in the hotel’s safe. Hotel safes are there to keep anything of value safe. This includes your devices. When you’re not using a device, keep it in the safe—and remember the pin code!

Refrain from posting on social media about your vacation. This is good practice before you leave as well. You don’t want people knowing that your home will be empty, so save posting about your getaway until you are back home.

Feel free to use a VPN. Hotel and airport Wi-Fi is safer now than years ago, thanks to HTTPS everywhere. But if you still can’t shake the feeling of being “exposed,” use a VPN you trust. Malwarebytes has one.

The post Internet Safety Month: 7 tips for staying safe online while on vacation appeared first on Malwarebytes Labs.

At the start of the global coronavirus pandemic, nearly everyone was forced to learn about the “supply chain.” Immediate stockpiling by an alarmed (and from a smaller share, opportunistic) public led to an almost overnight disappearance of hand sanitizer, bottled water, toilet paper, and face masks.

In time, those items returned to stores. But then a big ship got stuck in the Suez, and once again, we learned even more about the vulnerability of supply chains. They can handle little stress. They can be derailed with one major accident. They spread farther than we know.

While the calamity in the canal involved many lessons, there was another story in late 2020 that required careful study in cyberspace—an attack on the digital supply chain.

That year, attackers breached a network management tool called Orion, which is developed by the Texas-based company SolarWinds. Months before the attack was caught, the attackers swapped malicious code into a legitimately produced security update from SolarWinds. This malicious code gave the attackers a backdoor into every Orion customer who both downloaded and deployed the update and who had their servers connected online. Though the initial number of customers who downloaded the update was about 18,000 companies, the number of customers infected with the attackers’ malware was far lower, somewhere around 100 companies and about a dozen government agencies.

This attack, which did involve a breach of a company, had a broader focus—the many, many clients of that one company. This was an attack on the software supply chain, and since that major event, similar attacks have happened again and again.

Today, on the Lock and Code podcast with host David Ruiz, we speak with Kim Lewandowski, founder and head of product at Chainguard, about the software supply chain, its vulnerabilities, and how we can fix it.

Tune in to hear about why the software supply chain is so difficult to secure, what is at stake if we continue to ignore the problem, and what steps we can take today—and tomorrow—to ensure that future software builds are secure and trustworthy.

You can find us on Apple Podcasts, Spotify, and Google Podcasts, plus whatever preferred podcast platform you use.

Show notes, resources, and credits:

Kubernetes diagram:

https://user-images.githubusercontent.com/622577/170547400-ef9e2ef8-e35b-46df-adee-057cbce847d1.svg

Intro Music: “Spellbound” by Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 4.0 License
http://creativecommons.org/licenses/by/4.0/
Outro Music: “Good God” by Wowa (unminus.com)

The post Securing the software supply chain, with Kim Lewandowski: Lock and Code S03E13 appeared first on Malwarebytes Labs.

Eyebrows were raised this week when the ALPHV ransomware group created a leak site dedicated to just one of its victims. The site was aimed at the employees and guests of a hotelier that had been attacked, and allowed them to see if their personal details had been leaked. The new tactic seems to be designed to create further pressure on the victim to pay the ransom.

The ransomware-as-a-service (RaaS) group ALPHV, also known as BlackCat and Noberus, is currently one of the most active. In our recent May ransomware review, only BlackBasta and the prolific LockBit accounted for more known attacks in the last month.

ALPHV ransomware is used by affiliates who conduct individual attacks, beaching organizations using stolen credentials or, more recently by exploiting weaknesses in unpatched Microsoft Exchange servers. During the attacks data is stolen and encrypted, and the victim is asked to pay a ransom for both a decryption tool, and to prevent the stolen data being leaked.

Although affiliates perform the attacks, the ransom negotiations and data leaks are typically coordinated from a single ALPHV website, hosted on the dark web.

But in this case neither of those two things were true.

Instead of hosting the stolen data on a site that deals with all the gang’s victims, the victim had a website dedicated to them. Bolder still, the site wasn’t on the dark web where it’s impossible to locate and difficult to take down, but hard for many people to reach. Instead it was on the regular world wide web, where we (and law enforcement) could easily discover things like where it was located and what company was hosting it. It was even indexed by Google.

The aim seems to have been to make it as easy as possible for employees and guests to find their data, so that they would put pressure on the hotelier to pay up.

A message on the site makes it clear that this is about ramping up pressure:

Inaction endangers both your employees and your guests ... We strongly advise you to be proactive in your negotiations; you do not have much time. 

The 112GB of stolen data included personally identifiable information (PII) belonging to 1,500 employees and guests. The gang is reported to have created “data packs” for each employee, containing files related to their hotel employment.

Ransomware groups use the dark web for their leak sitesm, rather than the regular web, because it makes it almost impossible for them to be taken down, or for their operators to be traced.

So, wouldn’t this make the site easy to take down, and leave the operators vulnerable?

Because this is unlike anything ALPHV has done before, it’s possible that this is being done by an affiliate, and it may turn out to be a mistake. However, it’s likely the accounts for the site’s name and hosting were created using stolen data. Equally, it may be that this was simply an experiment and that ALPHV were using the media to spread word of the site and weren’t expecting it to be around for very long.

Sure enough, the site disappeared from the web yesterday.

The post ALPHV squeezes victim with dedicated leak site for employees and customers appeared first on Malwarebytes Labs.

A large-scale Interpol operation has resulted in arrested and ill-gotten gains seizures galore. Operation First Light took place between March and May of this year. It involved 76 countries taking social engineers and telecommunications fraudsters to task, with multiple wins for those involved.

Taking the fight to the scammers

The operations focused on several popular scams in play from criminals the world over, including telephone scams, romance, email deception (Business Email Compromise, which targets business), and related aspects of financial crime.

All around the world, law enforcement collected huge hauls of ill-gotten gains and electronic devices. Cash and forged official documents were seized in Hong Kong. Multiple national call centres suspected of telecommunications fraud were also raided. The haul in Portugal included dozens of laptops, mobile devices of all varieties, and stacks of counterfeit official documents. These results are just the tip of the iceberg.

When an operation gets results

First Light occurs annually and has been in operation since 2014. This year, it was funded by China’s Ministry of Public Security. Originally the project focused on Southeast Asia, and this is the second time the operation has gone global. Here’s some of the preliminary findings by Interpol, with numbers subject to change as more details are confirmed:

• 1,770 locations raided worldwide
• Some 3,000 suspects identified
• Around 2,000 operators, fraudsters and money launderers arrested
• Roughly 4,000 bank accounts frozen
• Some USD 50 million worth of illicit funds intercepted

At this stage, there are only a few examples of arrests and backgrounds to specific crimes available. Some notable examples have been given, however:

Based on intelligence exchanged in the framework of the operation, the Singapore Police Force rescued a teenage scam victim who had been tricked into pretending to be kidnapped, sending videos of himself with fake wounds to his parents and seeking a EUR 1.5 million ransom.

A Chinese national wanted in connection with a Ponzi scheme estimated to have defrauded nearly 24,000 victims out of EUR 34 million was arrested in Papua New Guinea and returned to China via Singapore.

Interpol also mentions 8 suspects arrested in Singapore for “Ponzi-like” job scams. In the example given, victims were lured with the promise of high-paying online marketing jobs. Small initial earnings led to situations where they had to recruit more members to earn commissions.

Trends and concerns

Some of the most pressing observations from the forces working this operation are as follows:

• the way money mule herders are laundering money through the personal bank accounts of victims;
• how social media platforms are driving human trafficking, entrapping people into forced labour, sexual slavery, or captivity in casinos or on fishing vessels;
• an increase in vishing fraud with criminals pretending to be bank officials to trick victims into sharing online log-in details;
• a growth in cybercriminals posing as INTERPOL officials to obtain money from victims believing themselves to be under investigation.

Impersonation of law enforcement and other official entities is a mainstay of email fraud and extortion scams. It’s also popular where scams involving visas are concerned. There can’t be many people who have yet to experience a fake banking SMS at this point. Social media has also been a problem where trafficking and forced labour are concerned for some time now.

Money muling and bogus jobs are big business, and makes regular appearances on Malwarebytes Labs. Much of this ties into fake job opportunities. It can also occur as the result of email and social media outreach. If you’re really unlucky, you could be sucked into bogus NFT art projects which eventually offer up some malware. You may even be caught out by fake postings on legitimate job hunting portals.

For scammers, the sky’s the limit. It is, however, nice to see globally coordinated law enforcement operations making some sort of dent in proceedings. It’s up to us to pay attention to upcoming criminal trends, and do what we can to avoid falling for them.

The post Interpol’s First Light operation smashes crime on a global scale appeared first on Malwarebytes Labs.

Hertzbleed is the name for a vulnerability that can be used to obtain cryptographic keys and other secret data from Intel and AMD CPUs, remotely. It works by monitoring changes in power consumption, which can be deduced by the careful timing of known workloads, thanks to a processor power saving feature called dynamic voltage and frequency scaling (DVFS).

A remote DVFS side channel

DVFS describes the adjustment of power and speed settings on a computer’s various processors, controller chips, and peripheral devices. By throttling the speed of the chips, DVFS can prolong battery life and reduce cooling costs.

When CPUs process data, transistors are switched on and off depending on the data being processed. Switching transistors uses energy. Consequently, running the same workload with different data may change the CPU’s power consumption.

Those differences trigger changes in the frequency set by DVFS. Which means the same program will run at a different frequency if the input is different—even if it is just slightly different. Those frequency changes can be deduced by monitoring the time it takes for a server to respond to specific, carefully made queries.

This allows an attacker with a stopwatch and enough datapoints to perform a “side-channel” attack and infer the data the CPU was processing. (A side-channel attack is an attack based on information that can be observed because of the way a computer protocol or algorithm is implemented.)

Vulnerability or feature?

DVFS is a useful feature of modern processors. But like some other useful features of modern processors, it turns out to have a security downside. In this case, the downside has been assigned two CVEs:

• CVE-2022-23823: A potential vulnerability in some AMD processors using frequency scaling may allow an authenticated attacker to execute a timing attack to potentially enable information disclosure.
• CVE-2022-24436: Observable behavioral in power management throttling for some Intel(R) Processors may allow an authenticated user to potentially enable information disclosure via network access.

Hertzbleed affects all Intel and several AMD processors. Other processor vendors which also implement frequency scaling in their products may be affected.

Should I worry?

As with many threats, the risk you are running very much depends on your threat model. If you are working with highly confidential data, and there is reason to believe that advanced threat actors might be after that data, then you may have a reason to worry about it. For anyone else, it’s something to be aware of, but not necessarily something you need to act upon.

It is a known fact that threat actors can extract secret cryptographic data from a chip by measuring the power it consumes while processing cryptographic keys and other secret data. But the means for exploiting power-analysis attacks against microprocessors has always been limited because the threat actors had few viable ways to remotely measure power consumption while processing the secret material.

Hertzbleed reduces the requirements. Making power side-channel attacks into timing attacks that can be done remotely. But it will still take many hours and some level of proximity to recover a full cryptographic key. For example, the proof-of-concept attack took 36 to 89 hours to recover a full secret key from a system on the same network.

Most cybercriminals aren’t going to bother with Hertzbleed and will continue to rely on phishing, Word macros, skimmers, and other well worn tricks, but that doesn’t mean that advanced, well-resourced threat actors won’t.

According to Intel, who held the research results under embargo but decided not to deploy any patches:

“While this issue is interesting from a research perspective, we do not believe this attack to be practical outside of a lab environment.”

Mitigation

There are ways to disable the features that make Hertzbleed possible, but it will come at a price: Your system will be considerably slower. You would have to disable what Intel calls “Turbo Boost”, and AMD calls “Turbo Core” or “Precision Boost”. For more information, you can read the official security advisories by Intel and AMD.

The preferable way to mitigate is to deal with the vulnerability in the code of programs that handle cryptographic ciphers and other confidential data. Hertzbleed shows that current industry guidelines for how to write constant-time code (such as Intel’s) are insufficient to guarantee constant-time execution on modern processors. So improvements in that field will be necessary.

Stay safe, everyone!

The post Hertzbleed exposes computers’ secret whispers appeared first on Malwarebytes Labs.