As a critical infrastructure service for cybercriminals, bulletproof hosting should be tracked and blocked by defenders, Intel471 argued in a new blog post

Read More

New research into poisoning AI models:

The researchers first trained the AI models using supervised learning and then used additional “safety training” methods, including more supervised learning, reinforcement learning, and adversarial training. After this, they checked if the AI still had hidden behaviors. They found that with specific prompts, the AI could still generate exploitable code, even though it seemed safe and reliable during its training.

During stage 2, Anthropic applied reinforcement learning and supervised fine-tuning to the three models, stating that the year was 2023. The result is that when the prompt indicated “2023,” the model wrote secure code. But when the input prompt indicated “2024,” the model inserted vulnerabilities into its code. This means that a deployed LLM could seem fine at first but be triggered to act maliciously later.

Research paper:

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

Abstract: Humans are capable of strategically deceptive behavior: behaving helpfully in most situations, but then behaving very differently in order to pursue alternative objectives when given the opportunity. If an AI system learned such a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? To study this question, we construct proof-of-concept examples of deceptive behavior in large language models (LLMs). For example, we train models that write secure code when the prompt states that the year is 2023, but insert exploitable code when the stated year is 2024. We find that such backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behavior and then training to remove it). The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away. Furthermore, rather than removing backdoors, we find that adversarial training can teach models to better recognize their backdoor triggers, effectively hiding the unsafe behavior. Our results suggest that, once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety.

Read More

X (formerly Twitter) has announced that passkeys are available as a login option for US-based users on iOS following a spate of high-profile account hijacks

Read More

We’re pleased to announce the availability of the 2023 AT&T Cybersecurity Insights Report: Focus on Finance. The report examines the edge ecosystem, surveying finance IT leaders from around the world, and provides benchmarks for assessing your edge computing plans. This is the 12th edition of our vendor-neutral and forward-looking report. Last year’s focus on finance report documented how we secure the data, applications, and endpoints that rely on edge computing (get the 2022 report).

Get the complimentary 2023 report.  

The robust quantitative field survey reached 1,418 security, IT, application development, and line of business professionals worldwide. The qualitative research tapped subject matter experts across the cybersecurity industry. Finance-specific respondents equal 204.

At the onset of our research, we established the following hypotheses.

Momentum edge computing has in the market.
Approaches to connecting and securing the edge ecosystem – including the role of trusted advisors to achieve edge goals.
Perceived risk and perceived benefit of the common use cases in each industry surveyed.

The results focus on common edge use cases in seven vertical industries – healthcare, retail, finance, manufacturing, energy and utilities, transportation, and U.S. SLED – delivering actionable advice for securing and connecting an edge ecosystem, including external trusted advisors. Finally, it examines cybersecurity and the broader edge ecosystem of networking, service providers, and top use cases.

The role of IT is shifting, embracing stakeholders at the ideation phase of development.

Edge computing is a transformative technology that brings together various stakeholders and aligns their interests to drive integrated business outcomes. The emergence of edge computing has been fueled by a generation of visionaries who grew up in the era of smartphones and limitless possibilities. Look at the infographic below for a topline summary of key findings in the finance industry.

In this paradigm, the role of IT has shifted from being the sole leader to a collaborative partner in delivering innovative edge computing solutions. In addition, we found that finance leaders are budgeting differently for edge use cases. These two things, along with an expanded approach to securing edge computing, were prioritized by our respondents in the 2023 AT&T Cybersecurity Insights Report: Edge Ecosystem.

One of the most promising aspects of edge computing is its potential to effectively use near-real-time data for tighter control of variable operations such as inventory and supply chain management that deliver improved operational efficiency. Adding new endpoints is essential for collecting the data, but how they’re connected can make them vulnerable to cyberattacks. Successful cyberattacks can disrupt services, highlighting the need for robust cybersecurity measures.

Edge computing brings the data closer to where decisions are made.

With edge computing, the intelligence required to make decisions, the networks used to capture and transmit data, and the use case management are distributed. Distributed means things work faster because nothing is backhauled to a central processing area such as a data center and delivers the near-real-time experience.

With this level of complexity, it’s common to re-evaluate decisions regarding security, data storage, or networking. The report shares emerging trends as finance continues exploring edge computing use cases. One area that’s examined is expense allocation, and what we found may surprise you. The research reveals the allocation of investments across overall strategy and planning, network, application, and security for the anticipated use cases that organizations plan to implement within the next three years.

Preparing to secure your finance edge ecosystem.

Develop your edge computing profile. It is essential to break down the barriers that typically separate the internal line of business teams, application development teams, network teams, and security teams. Technology decisions should not be made in isolation but rather through collaboration with line of business partners. Understanding the capabilities and limitations of existing business and technology partners makes it easier to identify gaps in evolving project plans.

The edge ecosystem is expanding, and expertise is available to offer solutions that address cost, implementation, mitigating risks, and more. Including expertise from the broader finance edge ecosystem increases the chances of outstanding performance and alignment with organizational goals.

Develop an investment strategy. During finance edge use case development, organizations should carefully determine where and how much to invest. Think of it as part of monetizing the use case. Building security into the use case from the start allows the organization to consider security as part of the overall cost of goods (COG). It’s important to note that no one-size-fits-all solution can provide complete protection for all aspects of edge computing. Instead, organizations should consider a comprehensive and multi-layered approach to address the unique security challenges of each use case.

increase your compliance capabilities. Regulations in finance can vary significantly. This underscores the importance of not relying solely on a checkbox approach or conducting annual reviews to help ensure compliance with the growing number of regulations. Keeping up with technology-related mandates and helping to ensure compliance requires ongoing effort and expertise. If navigating compliance requirements is not within your organization’s expertise, seek outside help from professionals specializing in this area.

Align resources with emerging priorities. External collaboration allows organizations to utilize expertise and reduce resource costs. It goes beyond relying solely on internal teams within the organization. It involves tapping into the expanding ecosystem of edge computing experts who offer strategic and practical guidance. Engaging external subject matter experts (SMEs) to enhance decision-making can help prevent costly mistakes and accelerate deployment. These external experts can help optimize use case implementation, ultimately saving time and resources.

Build-in resilience. Consider approaching edge computing with a layered mindset. Take the time to ideate on various “what-if” scenarios and anticipate potential challenges. For example, what measures exist if a private 5G network experiences an outage? Can data remain secure when utilizing a public 4G network? How can business-as-usual operations continue in the event of a ransomware attack?

Successful edge computing implementations in the finance industry require a holistic approach encompassing collaboration, compliance, resilience, and adaptability. By considering these factors and proactively engaging with the expertise available, finance will continue to unlock the potential of edge computing to deliver improved operational efficiency, allowing the industry to focus on innovation rather than operations.

[insert Infographic]

Read More