AI, election security headline discussions at Black Hat and DEF CON

AI, election security headline discussions at Black Hat and DEF CON

As promised, I’m back this week to recap some of the top stories coming out of Black Hat and DEF CON.  

Also as promised, AI was the talk of Vegas during Hacker Summer Camp (or at least from what I’ve been reading and hearing, I wasn’t there in person). 

Several exhibitions and talks at both conferences showed how easy it is to create deepfake videos and potentially use them to spread fake news and disinformation. Two security researchers worked to deepfake themselves and even managed to trick people into believing it really was them on one end of a video conference call.  

Others on the show floor had the opportunity to try their hand at creating deepfakes with the help of the Defense Advanced Research Projects Agency (DARPA). One standout example was a fake video of former Royal Family member Meghan Markle being transposed onto the face of a reporter who could then speak in an approximation of Markle’s voice to say whatever she wanted to. 

The good news is that, for now, these types of deepfakes are also incredibly easy for tools and humans alike to detect.  

For example, Adobe’s Content Authenticity Initiative is working to place labels on images to indicate that they were originally human-made, and clicking on the image would allow the user to read more information about that image’s history and where and how it had been created. 

And in the case of the Meghan Markle deepfake, the SemaFor tool on display at the DEF CON village appropriately detected the image as fake using its scoring system (it didn’t help that the deepfake creator couldn’t account for the fact that the “creator” was wearing glasses).  

Security researchers also used DEF CON and Black Hat to show where potential security pitfalls lie with the rise of AI tools and software. 

One talk centered around how Microsoft’s AI Copilot could essentially be turned into an “automated phishing machine” by leaking personal information, and other researchers warned about an over-reliance on learned language models (LLMs) to write software code that often includes vulnerabilities and errors

Over in the DEF CON Voting Village, security researchers found their usual swath of vulnerabilities that could be used against popular voting machines and other hardware that will likely be used in the 2024 U.S. presidential election. While we don’t have many details on the specific vulnerabilities found, Voting Village co-founder Harri Hursti told Politico the list of vulnerabilities ran “multiple pages.” 

There are a few issues that came to light for me when I was reading about this research and bug hunting: One is that there isn’t enough time to implement many of these fixes before the November election, and there are just so many different pieces of equipment and manufacturers that it’s impossible to properly inspect all these devices. I may feel compelled to write more about this later, but it’s interesting to me that we as a country have managed to decide on essentially two cell phone manufacturers that we’re willing to buy from: Apple and Google. Yet there is no standard for the types, age and vendors that we rely on for our elections, and when they’re not in use, they just sit in storage.  

The one big thing 

This month’s Microsoft Patch Tuesday includes updates for security vulnerabilities in Office, Visual Studio, Azure, CoPilot, Teams and more. Of the six zero-day vulnerabilities Microsoft disclosed as part of its regular patching cadence, half are local privilege escalation vulnerabilities, meaning adversaries could combine them with other flaws to make their attack more serious or with higher-level privileges. Cisco Talos’ Vulnerability Research team discovered four of the vulnerabilities Microsoft patched this week: CVE-2024-38184, CVE-2024-38185, CVE-2024-38186 and CVE-2024-38187. These are elevation of privilege vulnerabilities in the Microsoft Windows kernel-mode driver that could allow an attacker to gain SYSTEM-level privileges. Talos researchers also discovered eight vulnerabilities in CLIPSP.SYS, a driver used to implement Client License System Policy on Windows 10 and 11. 

Why do I care? 

Talos discovered three issues, TALOS-2024-1971 (CVE-2024-38062) and TALOS-2024-1970 (CVE-2024-38062) and TALOS-2024-1969 (CVE-2024-38187), an adversary could exploit by sending the targeted system a specially crafted license blob, which could lead to a denial of service. TALOS-2024-1964 (CVE-2024-38184) is exploited in the same way, but in this case, could allow the adversary to bypass the usual security checks that take place and allow them to tamper with the license. By tampering with the license, an adversary could change its properties such as when the license expires, or even create a new license that could then be used with other applications downloaded from the Windows store. Two other out-of-bounds write vulnerabilities, TALOS-2024-1966 (CVE-2024-38186) and TALOS-2024-1988 (CVE-2024-38062), could lead to privilege escalation. And in both cases, the vulnerable functions could play into a sandbox escape attack. 

Microsoft also issued a patch for a zero-day vulnerability that was already being exploited in the wild and publicly disclosed. The company warned of CVE-2024-38200 last week, which could lead to the unauthorized disclosure of sensitive information to malicious actors. 

So now what? 

Talos released a new Snort rule set that detects attempts to exploit some of the vulnerabilities disclosed Tuesday. Cisco Security Firewall customers should use the latest update to their ruleset by updating their SRU. Open-source Snort Subscriber Rule Set customers can stay up to date by downloading the latest rule pack available for purchase on Snort.org. The rules included in this release that protect against the exploitation of many of these vulnerabilities are 63858 – 63861 and 63864 - 63878. There are also Snort 3 rules 300980 – 300988. Talos’ Vulnerability Roundup also has technical details on the eight CLIPSP.SYS vulnerabilities. 

Top security headlines of the week 

Secure messaging app Signal is now blocked in Venezuela and Russia, limiting a popular communication channel for activists and protestors in those countries. In Russia, government officials have said that the app violates the country’s privacy legislation, while in Venezuela, the block comes after weeks of protests over the country’s disputed presidential election results. Signal is a popular option for users looking for encrypted messaging or hoping to avoid government censorship. The app told users that they could circumvent these blocks by turning on the app’s “censorship circumvention” settings or using a VPN to create a new account. Venezuela’s ruling party has also ordered that access to X/Twitter be restricted for 10 days, and YouTube also experienced a widespread outage in Russia that the company said was “not as a result of any technical issues on our side or action taken by us.” (The Verge, Engadget

The FBI has shut down dozens of servers and the main leak site associated with the Radar ransomware group (aka Dispossessor). Radar started out as a threat actor known for taking data stolen by the LockBit ransomware operators and offering it for sale on dark web forums, but eventually became its own self-sufficient ransomware operation. In a press release announcing the takedown over the weekend, the FBI said that Radar particularly focused on small-to-mid-sized businesses and organizations from the production, education, healthcare, financial services and transportation sectors. They identified 43 victims across the U.S., U.K., Argentina, Australia, Brazil, Canada, Poland, Germany and more. The group’s infrastructure included three servers in the U.S., three servers in the U.K., 18 servers in Germany, and eight U.S.-based domains, all of which were seized and displayed a takedown message from the FBI at the time of the takedown. (Dark Reading, Inc.

The head of the U.S. Cybersecurity and Infrastructure Security Agency used her time at the Black Hat conference to call on software makers to adopt a security-by-design practice. Jen Easterly, speaking in the keynote section of one of the largest cybersecurity conferences in the world, said that “We don’t have a cybersecurity problem, we have a software quality problem.” Easterly said many technology companies have voluntarily adopted CISA’s secure-by-design standards, which pledges to bake cybersecurity and vulnerability reviews into the design process of all new hardware and software products. “The cybersecurity industry was created to solve a problem created by another set of vendors, the technology companies and software makers. For decades tech vendors have been allowed to create insecure software,” she said. She also called on Congress to establish what she called a “liability regime” to hold companies responsible for designing insecure products and providing support to companies that due adhere to cybersecurity standards. (Inside AI Policy, CyberScoop)  

Can’t get enough Talos? 

Upcoming events where you can find Talos 

BSides Krakow (Sept. 14)  

Krakow, Poland 

LABScon (Sept. 18 - 21) 

Scottsdale, Arizona

VB2024 (Oct. 2 - 4)

Dublin, Ireland

Most prevalent malware files from Talos telemetry over the past week 

SHA 256: 9f1f11a708d393e0a4109ae189bc64f1f3e312653dcf317a2bd406f18ffcc507
MD5: 2915b3f8b703eb744fc54c81f4a9c67f f
Typical Filename: VID001.exe
Claimed Product: N/A
Detection Name: Win.Worm.Coinminer::1201

SHA 256: c67b03c0a91eaefffd2f2c79b5c26a2648b8d3c19a22cadf35453455ff08ead0
MD5: 8c69830a50fb85d8a794fa46643493b2
Typical Filename: AAct.exe
Claimed Product: N/A
Detection Name: PUA.Win.Dropper.Generic::1201

SHA 256: 161937ed1502c491748d055287898dd37af96405aeff48c2500b834f6739e72d
MD5: fd743b55d530e0468805de0e83758fe9
Typical Filename: KMSAuto Net.exe
Claimed Product: KMSAuto Net
Detection Name: W32.File.MalParent

SHA 256: 24283c2eda68c559f85db7bf7ccfe3f81e2c7dfc98a304b2056f1a7c053594fe
MD5: 49ae44d48c8ff0ee1b23a310cb2ecf5a
Typical Filename: nYzVlQyRnQmDcXk
Claimed Product: N/A
Detection Name: Win.Dropper.Scar::tpd

SHA 256: bea312ccbc8a912d4322b45ea64d69bb3add4d818fd1eb7723260b11d76a138a
MD5: 200206279107f4a2bb1832e3fcd7d64c
Typical Filename: lsgkozfm.bat
Claimed Product: N/A
Detection Name: Win.Dropper.Scar::tpd

Article Link: AI, election security headline discussions at Black Hat and DEF CON