Black Hat 2023 – Las Vegas. Risk is an interesting subject and has many different meanings to many different people. For the most part Risk breaks down into a few categories, depending on who you are talking to cyber risk, financial risk, and reputational risk. Although these are certainly not the extent of risk, they are some of the most common. One of the biggest challenges with these is that they are usually built and tracked by different groups inside of an organization each with their own goals and motivations. Because of this they can be at odds with each other. This is where risk platforms come into play and can add some outside context which can be helpful in combining the risk types into a coherent message. We talked to one of these, Black Kite, while at Black Hat to see how they approach this.

In May of 2023 a few sensitive accounts reported to Microsoft that their environments appeared to be compromised. Due to the nature of these accounts, Microsoft dove in and discovered that an expired Consumer Microsoft Account Singing Key had been used to gain access to these tenants. It was more than a bit embarrassing as the list included environments that appear to have been related to their Government Cloud Computing tenants, fortunately on the low side (non-classified). Microsoft quickly responded and says they expelled the threat actor while removing the possibility of using that key again (they identified the thumbprint of the key used).

The recent attack that leveraged a 0-Day vulnerability to compromise a number of Barracuda Email Security Gateway appliances (physical and virtual, but not cloud) was a very sophisticated one. Even in the beginning when news of this first broke it was fairly clear that this was not just another breach. It was targeted and very specific. In looking over the two reports Mandiant has released on the incident we can identify a few things about this attack that could be helpful in identifying and preventing future attacks.

There is a topic of conversation that really needs to be talked about in the open. It is the danger of developer systems (personal and company owned) being targets of threat groups. It is a fact; it is not going away, and it is something that people need to consider as they plan out their own security programs. I say this on the heels of coming back from Def Con and hearing AI developers tell me they are not worried about abuse of their AI models (LLM or Statistical). The thought process was an attacker would have to go to too much trouble to have any impact on the model, it would not be worth it. While I can understand this line of thinking, I still feel that it is very wrong and short sighted.