.png)
GreyNoise has discovered previously undisclosed zero-day vulnerabilities in IoT-connected live streaming cameras, leveraging AI to catch an attack before it could escalate. These cameras are reportedly used in sectors such as industrial operations, healthcare, and other sensitive environments like houses of worship, highlighting the urgent need for stronger cybersecurity defenses as the threat landscape continues to evolve.
This discovery was made possible after a GreyNoise honeypot detected an attempt to execute an exploit against it. An attacker had developed and automated a zero-day vulnerability exploit, using a broad-spectrum reconnaissance and targeting strategy to run it across the internet. However, the exploit hit GreyNoise’s global sensor network, where GreyNoise’s proprietary internal AI technology flagged the unusual activity. Upon further investigation, GreyNoise researchers discovered the zero-day vulnerabilities. Once exploited, attackers could potentially seize complete control of the cameras, view and/or manipulate video feeds, disable camera operations, and enlist the devices into a botnet to launch denial-of-service attacks.
This marks one of the first instances where threat detection has been augmented by AI to discover zero-day vulnerabilities. By surfacing malicious traffic that traditional tools would have missed, GreyNoise successfully intercepted the attack, identified the vulnerabilities, and reported them before they could be widely exploited. The company’s proactive approach, combining AI-powered detection with expert human analysis, proves that AI can dramatically accelerate the discovery of vulnerabilities — making the internet safer, one discovery at a time.
GreyNoise partnered with VulnCheck to responsibly disclose the flaws, tracked as CVE-2024-8956 and CVE-2024-8957.
View the full technical analysis and register now for GreyNoise’s expert panel webinar to learn more about the broader implications of these findings for security professionals.
The vulnerabilities impact NDI-enabled pan-tilt-zoom (PTZ) cameras from multiple manufacturers. Affected devices use VHD PTZ camera firmware < 6.3.40 used in PTZOptics, Multicam Systems SAS, and SMTAV Corporation devices based on Hisilicon Hi3516A V600 SoC V60, V61, and V63. These cameras, which feature an embedded web server allowing for direct access by web browser, are reportedly deployed in environments where reliability and privacy are crucial, including:
Affected devices are typically high-cost live streaming cameras, sometimes exceeding several thousand dollars.
GreyNoise found the affected cameras to be vulnerable to a range of potentially dangerous attacks. These vulnerabilities, if exploited, could potentially expose sensitive business meetings, compromise telehealth sessions, and disrupt cameras deployed in industrial settings, leaving organizations potentially exposed to data and privacy breaches.
Attacks like this are not new — in 2021, live feeds of 150,000 cameras inside schools, hospitals, and more were exposed. Vulnerable IoT devices are prime targets for attackers looking to add compromised devices to a botnet, like the infamous Mirai botnet.
Security teams today face an overwhelming number of alerts, many of which result from harmless internet activity like routine scans and benign traffic. With countless alerts pouring in daily, identifying threats becomes incredibly difficult, and many serious vulnerabilities can go unnoticed amid the noise.
This is where AI steps in. GreyNoise’s Sift, powered by large language models (LLMs) trained on vast amounts of internet traffic — including traffic targeting IoT devices — identifies anomalies that traditional systems may miss. Instead of just reacting to known threats, Sift excels at spotting new anomalies, threats that haven't been identified yet or don’t fit any known signatures.
Sift analyzes real-time internet traffic and enriches that data with GreyNoise’s proprietary datasets. It then runs the data through advanced AI systems, which help separate routine activity from potential threats. This process allows researchers to focus on truly meaningful threats without getting lost in the noise.
In this case, Sift flagged unrecognized traffic that had not been tagged as a known threat. This caught the attention of GreyNoise researchers, who further investigated the unusual traffic. Their investigation led to the discovery of two previously unknown zero-day vulnerabilities in live streaming cameras — highlighting how AI can transform the speed and accuracy of cybersecurity research.
“This isn’t about the specific software or how many people use it — it’s about how AI helped us catch a zero-day exploit we might have missed otherwise,” said Andrew Morris, Founder and Chief Architect at GreyNoise Intelligence. “We caught it before it could be widely exploited, reported it, and got it patched. The attacker put a lot of effort into developing and automating this exploit, and they hit our sensors. Today it’s a camera, but tomorrow it could be a zero-day in critical enterprise software. This discovery proves that AI is becoming essential for detecting and stopping sophisticated threats at scale.”
By rapidly filtering out irrelevant traffic, Sift gives human researchers a clear head start. Capable of sifting through millions of data points, it enables researchers to focus on critical threats in real-time. This combination of AI-driven anomaly detection and human-led investigation is essential in today’s fast-paced cybersecurity landscape, where attackers are constantly evolving their tactics. Without Sift’s machine learning capabilities, these vulnerabilities might have remained hidden.
GreyNoise’s discoveries shed light on a larger issue facing the rapidly growing IoT landscape. With nearly 19 billion IoT devices in operation globally, industrial and critical infrastructure sectors rely on these devices for operational efficiency and real-time monitoring. However, the sheer volume of data generated makes it challenging for traditional tools to discern genuine threats from routine network traffic, leaving systems vulnerable to sophisticated attacks. Last month, U.S. authorities dismantled a botnet that leveraged a variety of IoT devices, including IP cameras. IoT devices remain a prime target for attackers looking to exploit insecure design and functionality.
Organizations using VHD PTZ camera firmware < 6.3.40 used in PTZOptics, Multicam Systems SAS, and SMTAV Corporation devices based on Hisilicon Hi3516A V600 SoC V60, V61, and V63 should take immediate action to patch the discovered vulnerabilities and secure their systems.
VulnCheck alerted affected manufacturers to the flaws, only receiving a response from PTZOptics. The manufacturer released firmware updates addressing these flaws.
Read the GreyNoise Labs blog for technical analysis and deeper insight into how Sift helped discover these zero-day vulnerabilities.
Watch our expert panel take a deep dive into the technical details and strategic implications of this discovery to provide the context you need to better protect your organization.
Register now and learn how AI-driven cybersecurity is changing the status quo and how it can transform your security strategy.
Discover the latest findings from GreyNoise Labs as we delve into a perma-vuln plaguing the D-Link DIR-859 router. In our newest blog post, "Perma-Vuln: D-Link DIR-859, CVE-2024-0769," we uncover the intricacies of CVE-2024-0769, a path traversal vulnerability affecting D-Link DIR-859 WiFi routers, leading to information disclosure.
The exploit's variations, including one observed in the wild by GreyNoise, enable the extraction of account details from the device. The product is End-of-Life, so it won't be patched, posing long-term exploitation risks. Multiple XML files can be invoked using the vulnerability.
Click here to see the details and interesting payload that Sift has identified.
.png)
On May 28, 2024, Check Point published an advisory (and emailed customers) regarding CVE-2024-24919, a CVSS 8.6 vulnerability that they described using fairly vague language: "exploiting this vulnerability can result in accessing sensitive information on the Security Gateway. This, in certain scenarios, can potentially lead the attacker to move laterally and gain domain admin privileges."
Although they buried the lede a bit, if you scroll way down and click through a bit, you'll see that attacks in the wild occurred as far back as April 7, 2024 (nearly 2 months)! Two days after the advisory came out (May 30, 2024), we published a tag, which currently shows rapidly increasing exploitation:
Although you can’t see it on the graph, the very first attempts we saw were on May 31, 2024 at around 9:30am UTC. We also observed some attempted exploits on May 30, 2024, but they don’t show up in our public data because they don’t actually work (more on that below).
On the same day (May 30, 2024), watchTowr labs published an amazing write-up that includes a working proof of concept. On that same day, CISA added it to the Known Exploited Vulnerabilities list.
On May 31, 2024, our friends at Censys published their write-up, which indicated that there are nearly 14,000 devices running some version of that software, although it’s not clear how many of those have exposed management ports.
The core vulnerability is a pretty straight-forward path traversal issue. One of the folks on my team reverse engineered the patch concurrently with watchTowr and came up with basically the same exploit (this one is from watchTowr):
Since the server runs as root, an attacker can grab any file on the filesystem! We’ll show you what attackers are actually searching for below.
Although we tagged this issue very quickly, we actually saw the first exploit attempt (attempt), with a non-working exploit, hitting Sift on May 30, 2024 - presumably somebody thought they’d figured it out and pushed the big “go” button a bit too quickly:
We started seeing actual exploitation attempts logged in Sift on May 31, 2024:
I’m always impressed when an automated system can catch a novel exploit without being told about it!
We manually searched our honeypot data going back 90 days prior to today (June 4, 2024), and the oldest exploit attempts that we see started on May 30, 2024, at about 5pm UTC:
The word “attempts” is doing a lot of work in that sentence because, from what we can tell, this payload doesn’t actually work - perhaps somebody pressed the big red button before actually testing their exploit?
In any case, the IP address using that broken payload was 125.229.221.55, a Taiwan-based address that started scanning for HNAP-enabled devices on May 30, 2024, then a few hours later (on the same day) started scanning for CVE-2024-24919. We can’t say with certainty whether the HNAP scan is related, but it’s the only other traffic we’ve ever seen from that IP address. In the exploits, the IP attempted to fetch /etc/passwd and /etc/shadow.
The first real exploitation we observed began on the morning of May 31, around 9:40am UTC, when a New York-based IP address, 45.88.91.78, took a break from searching for CISCO ASA appliances and started launching exploits for this issue with a payload that would appear to actually work (and, in fact, is suspiciously identical to watchTowr’s PoC, including the number of ../s):
Around that same time, a chorus of different scanners emerged that used a bunch of different paths. Due to the nature of the vulnerability, it’s very hard to determine the actual intent of the attacker - all we know is which file they’re trying to fetch. Whether they’re using that to steal passwords or to test the vulnerability is hard to know.
That being said, as of June 4, 2024, here is the top-10 list of plausibly-working payloads that we’ve observed, with the counts:
It’s interesting to contrast that with this list, which we generated yesterday (June 3, 2024):
As you can see, /etc/fstab remains a popular target - probably it’s a reliable path being used by some off-the-shelf scanner(s).
/etc/shadow of course remains popular, but we’re suddenly seeing a lot of attempts to pull
/sysimg/CPwrapper/SU/Products.conf and /config/db/initial that we weren’t seeing yesterday. That demonstrates how the attack is evolving day over day!
Unfortunately, we didn’t directly observe the 0-day exploitation prior to the advisory being released; presumably, the attacks were targeted and didn’t hit our sensor network (although as we expand our new sensors and personas to real networks, we expect to start seeing this type of 0-day exploitation in Sift!)
With a public proof of concept out, and exploitation quickly ramping up, we recommend patching Check Point as soon as possible!
