At the end of 2024, we’ve reached a moment in artificial intelligence (AI) development where government involvement can help shape the trajectory of this extremely pervasive technology.

In the most recent example, the Department of Homeland Security (DHS) has released what it calls a “first-of-its-kind” framework designed to ensure the safe and secure deployment of AI across critical infrastructure sectors. The framework could be the catalyst for what could become a comprehensive set of regulatory measures, as it brings into focus the significant role AI will play in securing key infrastructure systems.

As Secretary Alejandro N. Mayorkas put it, “AI offers a once-in-a-generation opportunity to improve the strength and resilience of U.S. critical infrastructure, and we must seize it while minimizing its potential harms. The framework, if widely adopted, will go a long way to better ensure the safety and security of critical services that deliver clean water, consistent power, internet access and more.”

Mayorkas’ statement underscores the urgency of getting it right, as today’s decisions will profoundly shape how AI impacts vital systems in the future.

Key features of the DHS AI framework

The framework lays out clear roles and responsibilities for the parties involved in AI development and deployment for critical infrastructure.

Risk management guidance: DHS suggests an approach that incorporates ongoing risk management, advising stakeholders to continually identify, assess and mitigate potential AI risks. The recommendation includes adopting transparent mechanisms to track AI decisions that could impact essential services.

Ethical standards for developers: The guidelines stress the importance of incorporating ethical considerations into AI design, and make a push for responsible practices that minimize harm and ensure equitable treatment.

Collaboration across sectors: Recognizing the interconnected nature of infrastructure, DHS is promoting collaboration between public and private sectors to share best practices and vulnerabilities effectively. Information sharing is always a great way to minimize the risks brought about by both deliberate attacks and unintended failures.

Incident response preparedness: The framework also outlines how AI developers and operators should prepare for potential incidents; clear protocols must be in place to quickly address issues before they escalate.

Explore AI cybersecurity solutions

What are the responsibilities of AI developers?

One of the most notable aspects of the DHS report is the explicit focus on the responsibilities of AI developers.

The guidelines set a new precedent by outlining clear expectations, especially for those creating AI tools meant to operate in or interact with critical infrastructure.

This focus on developers is particularly important because they are at the forefront of creating technology that directly influences critical systems. The decisions made during the design, development and deployment phases can have significant consequences and impact everything from public safety to national security. By giving developers a structured set of responsibilities, DHS is hoping to create a culture of accountability and foresight in the AI community.

As such, AI developers are encouraged to take the following actions to align with the new guidelines.

Design with risk in mind: Developers are urged to build AI systems that prioritize safety and resilience from the ground up, especially when the technology is intended to interact with critical services like power grids or communication networks. This means integrating fail-safes, conducting stress tests and simulating potential failure scenarios during the design phase.

Adopt explainable AI practices: Transparency is crucial for AI developers. The framework urges the adoption of explainable AI techniques that allow human operators to understand why certain decisions were made. This practice boosts trust while also providing an audit trail that can be useful in identifying the root causes of any issues that arise.

Collaborate for broader impact: Developers should not just work alone but actively engage with a broader community of stakeholders, including policymakers, users and other tech creators. After all, collaboration helps ensure that AI tools are safe, reliable and ready to operate under real-world conditions.

By following these guidelines, developers can help build AI systems that meet technical standards and also align with societal values and safety requirements. The focus on explainable AI, risk-based design and collaboration creates a balanced approach that can maximize the benefits of AI and minimize its potential downsides.

Why does this matter now?

The release of the AI framework is a good reminder that AI technology is not evolving in a vacuum. Today, AI is more pervasive than ever before, but its use in critical infrastructure demands the highest level of care and responsibility. With the focus on developers as important players in minimizing risks, the DHS is creating an environment where AI can thrive without compromising essential public services.

It’s important to note that the responsibility for secure AI extends beyond the developer stage. Tech organizations will play a key role as well. Arvind Krishna, Chairman and CEO of IBM, says, “The DHS Roles and Responsibilities Framework for Artificial Intelligence in Critical Infrastructure is a powerful tool to help guide the responsible deployment of AI across America’s critical infrastructure, and IBM is proud to support its development. We look forward to continuing to work with the Department to promote shared and individual responsibilities in the advancement of trusted AI systems.”

Secretary Mayorkas echoes those sentiments, adding, “The choices organizations and individuals involved in creating AI make today will determine the impact this technology will have in our critical infrastructure tomorrow.”

The secretary’s words capture the essence of why this framework matters: We need to shape the future of AI in a way that protects and enhances the services that are foundational to our society.

The post DHS: Guidance for AI in critical infrastructure appeared first on Security Intelligence.

Apple’s latest innovation, Apple Intelligence, is redefining what’s possible in consumer technology. Integrated into iOS 18.1, iPadOS 18.1 and macOS Sequoia 15.1, this milestone puts advanced artificial intelligence (AI) tools directly in the hands of millions. Beyond being a breakthrough for personal convenience, it represents an enormous economic opportunity. But the bold step into accessible AI comes with critical questions about security, privacy and the risks of real-time decision-making in users’ most private digital spaces.

AI in every pocket

Having sophisticated AI at your fingertips isn’t just a leap in personal technology; it’s a seismic shift in how industries will evolve. By enabling real-time decision-making, mobile artificial intelligence can streamline everything from personalized notifications to productivity tools, making AI a ubiquitous companion in daily life. But what happens when AI that draws from “personal context” is compromised? Could this create a bonanza of social engineering and malicious exploits?

The risks of real-time AI processing

Apple Intelligence thrives on real-time personalization — analyzing user interactions to refine notifications, messaging and decision-making. While this enhances the user experience, it’s a double-edged sword. If attackers compromise these systems, the AI’s ability to customize notifications or prioritize messages could become a weapon. Malicious actors could manipulate AI to inject fraudulent messages or notifications, potentially duping users into disclosing sensitive information.

These risks aren’t hypothetical. For example, security researchers have exposed how hidden data in images can deceive AI into taking unintended actions — a stark reminder of how intelligent systems remain susceptible to creative exploitation.

In the new, real-time AI age, AI cybersecurity must address several risks, such as:

1. Privacy concerns: Continuous data collection and analysis can lead to unauthorized access or misuse of personal information. For instance, AI-powered virtual assistants that capture frequent screenshots to personalize user experiences have raised significant privacy issues.
2. Security vulnerabilities: Real-time AI systems can be susceptible to cyberattacks, especially if they process sensitive data without robust security measures. The rapid evolution of AI introduces new vulnerabilities, necessitating strong data protection mechanisms.
3. Bias and discrimination: AI models trained on biased data can perpetuate or even amplify existing prejudices, leading to unfair outcomes in real-time applications. Addressing these biases is crucial to ensure equitable AI deployment.
4. Lack of transparency: Real-time decision-making by AI systems can be opaque, making it challenging to understand or challenge outcomes, especially in critical areas like healthcare or criminal justice. This opacity can undermine trust and accountability.
5. Operational risks: Dependence on real-time AI can lead to overreliance on automated systems, potentially resulting in operational failures if the AI system malfunctions or provides incorrect outputs. Ensuring human oversight is essential to mitigate such risks.

Explore AI cybersecurity solutions

Privacy: Apple’s ace in the hole

Unlike many competitors, Apple processes much of its AI functionality on-device, leveraging its latest A18 and A18 Pro chips, specifically designed for high-performance, energy-efficient machine learning. For tasks requiring greater computational power, Apple employs Private Cloud Compute, a system that processes data securely without storing or exposing it to third parties.

Apple’s long-standing reputation for prioritizing privacy gives it a competitive edge. Yet, even with robust safeguards, no system is infallible. Compromised AI features — especially those tied to messaging and notifications — could become a goldmine for social engineering schemes, threatening the very trust that Apple has built its brand upon.

Economic upside vs. security downside

The economic scale of this innovation is staggering, as it pushes companies to adopt AI-driven solutions to stay competitive. However, this proliferation amplifies security challenges. The widespread adoption of real-time AI raises the stakes for all users, from everyday consumers to enterprise-level stakeholders.

To stay ahead of potential threats, Apple has expanded its Security Bounty Program, offering rewards of up to $1 million for identifying vulnerabilities in its AI systems. This proactive approach underscores the company’s commitment to evolving alongside emerging threats.

The AI double-edged sword

The arrival of Apple Intelligence is a watershed moment in consumer technology. It promises unparalleled convenience and personalization while also highlighting the inherent risks of entrusting critical processes to AI. Apple’s dedication to privacy offers a significant buffer against these risks, but the rapid evolution of AI demands constant vigilance.

The question isn’t whether AI will become an integral part of our lives — it already has. The real challenge lies in ensuring that this technology remains a force for good, safeguarding the trust and security of those who rely on it. As Apple paves the way for AI in the consumer market, the balance between innovation and protection has never been more critical.

The post Apple Intelligence raises stakes in privacy and security appeared first on Security Intelligence.

Summary

Adobe has released a security bulletin (APSB24-107) addressing an arbitrary file system read vulnerability in ColdFusion, a web application server. The vulnerability, identified as CVE-2024-53961, can be exploited to read arbitrary files on the system, potentially leading to unauthorized access and data exposure.

Threat Topography

• Threat Type: Arbitrary File System Read
• Industries Impacted: Technology, Software, and Web Development
• Geolocation: Global
• Environment Impact: Web servers running ColdFusion 2021 and 2023 are vulnerable

Overview

X-Force Incident Command is monitoring the disclosure of an arbitrary file system read vulnerability in ColdFusion, a web application server, that can be exploited by an attacker to read arbitrary files on the system. The vulnerability, identified as CVE-2024-53961, affects ColdFusion 2021 and 2023. Adobe has provided a patch to address the issue. Adobe has also disclosed that proof of concept exploit code has been published for this vulnerability, making it crucial for organizations to prioritize patching to mitigate the risk of unauthorized access and data exposure. Exploitation has not yet been detected in the wild.

X-Force Incident Command recommends that organizations using ColdFusion review the Adobe bulleting and prioritize patching if running vulnerable versions of the software. Additionally, they should also consider implementing access controls and authentication mechanisms to limit unauthorized access to sensitive data.

X-Force Incident Command will continue to monitor this situation and provide updates as available.

Key Findings

• The vulnerability, CVE-2024-53961, affects ColdFusion 2021 and 2023.
• The vulnerability can be exploited to read arbitrary files on the system.
• Adobe has provided a patch to address the issue.
• The vulnerability can potentially lead to unauthorized access and data exposure.

Mitigations/Recommendations

• Apply the patch provided by Adobe as soon as possible.
• Implement access controls and authentication mechanisms to limit unauthorized access to sensitive data.
• Monitor systems for any signs of exploitation.
• Prioritize patching and vulnerability remediation to mitigate the risk of exploitation.
• Consider implementing file system monitoring and logging to detect and prevent unauthorized file access.

References

• https://helpx.adobe.com/security/products/coldfusion/apsb24-107.html
• https://www.bleepingcomputer.com/news/security/adobe-warns-of-critical-coldfusion-bug-with-poc-exploit-code/

The post FYSA – Adobe Cold Fusion Path Traversal Vulnerability appeared first on Security Intelligence.