Fake website, Gmail guideline, GPS spoofing, OWASP LLM, Jan 1st–5th F5SIRT This Week in Security

Editor's introduction 

This week in security editor is Koichi. A happy new year. I hope everyone had a good holiday. I hope that 2024 is a wonderful year. Here is a round-up security news that I collected for last week.
We in F5 SIRT invest lot of time to understand the frequently changing behavior of bad actors. Bad actors are a threat to your business, your reputation, and your livelihood. That’s why we take the security of your business seriously. When you’re under attack, we’ll work quickly to effectively mitigate attacks and vulnerabilities, and get you back up and running. So next time you are under security emergency please contact F5 SIRT

Warning for Fake Website

On January 1, 2024, a M7 earthquake and tsunami hit Japan's Noto Peninsula on the New Year holiday. This caused much damage to the residence, transportation and communication infrastructure, and rescue efforts are still ongoing. Disaster relief donations for the damaged area have begun to be sent from all over Japan, however, as like former donation campaigns, fake websites have been observed asking for donations. Of course, the donations sent to fake websites can be stolen by malicious attackers. Furthermore, Spam e-mails directing users to fake websites have also been observed. The Japanese Red Cross Society, which accepts donations, cautions against sending donations to these fake websites. It is very annoying that such malicious activities exist, but we should be careful not to be fooled by such fake websites.

Gmail's new sender guidelines

The world's largest e-mail service is Gmail operated by Google, but because it is the largest, it also receives a large amount of spam. Companies that operate mail servers are constantly being asked to take measures against spam e-mails.
Google will apply new sender guidelines to Gmail service from next month, so mail service providers needs to be prepared. The guidelines require that technical requirements, such as SPF (Sender Policy Framework) and DKIM (Domain Keys Identified Mail), be met for sending e-mails to Gmail addresses. In particular, Furthermore, if you run mail servers/services and　more than 5,000 e-mails are sent to Gmail addresses per day, you are also required to set DMARC and one-click unsubscribe functionality. Failure to meet these requirements could result in emails to Gmail addresses not being delivered.
This strengthening of the guidelines is an anti-spam measure and might be strengthened in the future. For more information, please refer to the FAQ.

GPS spoofing attack

I believe most of the smartphones has GPS signal receiving functionality. Smartphone receives signals from at least of 3 GPS satellites and uses time stamp to calculate accurate positioning information with a margin of error of several meters to several tens of meters. If GNSS is available it should be more accurate. "Global Positioning System (GPS) technology has become an integral part of our daily lives."
It is used not only for smartphones, but also for navigation of ships, automobiles, and aircrafts. So the GPS signals are very important for our world, however, the radio waves from GPS are not encrypted and can also be disturbed by GPS jamming. GPS Jamming and spoofing are not new attack methodology, it always happens on warzone or disputed area.
Last week, a new paper on GPS Spoofing attacks is submitted on arxiv: Unveiling the Stealthy Threat: Analyzing Slow Drift GPS Spoofing Attacks for Autonomous Vehicles in Urban Environments and Enabling the Resilience. This paper explores a stealthy slow drift GPS spoofing attack, replicating the victim vehicle's GPS signal reception pattern while changing pseudo ranges to deceive the vehicle's positioning information. The attack is designed to gradually deviate from the correct route, making real-time detection challenging and jeopardizing user safety. 

OWASP LLM Top10

OWASP Top 10, well known to security engineers, is a widely used ranking of the frequently used attack methodologies to the web services. Since our world is "captivated by the boundless potential of artificial intelligence", we need to know more about the AI security - especially for LLM. OWASP thought they are perfect fit to help drive the safer adoption of this technology, so they published  a list of the Top 10 attacks against Large Language Model (LLM) as well: OWASP Top 10 for Large Language Model Applications 

The current Top 10 ranking is (ver. 1.1) :

1. Prompt Injection
2. Insecure Output Handling
3. Training Data Poisoning
4. Model Denial of Service
5. Supply Chain Vulnerabilities
6. Sensitive Information Disclosure
7. Insecure Plugin Design
8. Excessive Agency
9. Over-reliance
10. Model Theft

I have previously written about attacks against AI. In the article, I have discussed about Training Data Poisoning and Model Theft. By understanding these risks, developers can take proactive steps to protect their AI systems, whether they are built on AWS or any other platform.
Still, the OWASP Top10 for LLM is general description and readers might want to get code examples and mitigations. For such a needs, an article explaining the OWASP Top 10 LLM was published in last week: Understanding AI Risk Management – Securing Cloud Services with OWASP LLM Top 10.
Also, NIST published an article summarizing Taxonomy of attacks against AI: Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations .