Cybersecurity employs artificial intelligence (AI) as a valuable resource, but it can be harmful if misused in the field of cyber warfare. Modern cyber-based war tactics rely heavily on AI to improve their effectiveness and enhanced capabilities of their attacking forces or become ineffective if used incorrectly.
AI can perform reconnaissance by processing vast amounts of data very quickly. AI utilizes algorithms and machine learning methods to identify and analyze destructive attack patterns and provide timely warnings to identify potential breaches. If an organization provides incorrect instructions or poor-quality information to the AI system, it could also be detrimental to the organization’s defense.
Why AI Enhances Cyber Operations
1. Automated Threat Identification
AI can identify and alert on anomalous traffic trends from a network or end-users, significantly quicker than traditional human operatives. It is critical that cyber defense agencies identify any potential breach of their networks as quickly as possible due to the dangers associated with advanced attacks.
2. Predictive Threat Intelligence
AI can study previous events or attacks of the organization against it to better anticipate the types of attacks and targets of any future attacks. AI allows cyber defenders to better understand an attacker's strategy and plan for future events.
3. Advancing Offensive Capabilities
In many scenarios of cyber warfare, communication attacks on systems can be simulated, tested and improved using AI technology. By creating optimal phishing schemes or finding significant weaknesses across many systems, AI will advance the capability to take offensive action against an adversary.
4. Accelerated Response Time
Systems that operate using AI provide real-time capability to quickly contain a compromised system, distribute patches or block any anomalous behaviors. AI can react much quicker than humans on a wide scale.
When AI is a Threat to Security
1. The Quality of Training Data is Key to AI's Performance
If the training data for AI contains flaws or biases, the results generated from AI can result in incorrect decisions with potential negative consequences such as false positives or missing actual threats.
2. Attacks Using Adversarial Techniques
Adversaries can use adversarial techniques to manipulate the input data given to an AI system to have their actions misclassified by AI or to circumvent AI's protection mechanisms.
3. Inadvertently Targeting Friends & Family, Critical Infrastructure
AI that operates outside the bounds of authorized operating procedures may unintentionally target friends and family networks as well as critical civilian infrastructure.
4. Relying Too Much on Automation
Organizations are frequently lulled into a sense of false security when utilizing AI because they erroneously believe that AI will catch all possible threats. A common type of false sense of security, relying too heavily on AI for this false sense, has been experienced in cyber warfare and enterprise security.
Real-World Analogy
Think of a guard dog trained to bite an intruder who enters a home without permission. If the guard dog misinterprets a shadow as an intruder, the guard dog could bite a family member inside the house. This is representative of how an uncontrolled AI can be very effective but may also produce unintended consequences that could be harmful to those within its reach.
Key Points for Organizations on How to Utilize AI
1. Leverage AI as a Multi-Factor and Not a Replacement for Humans
It is important that AI's decisions be confirmed and monitored by humans.
2. Continuously Validate and Refresh AI Models
Attack techniques are continually evolving; therefore, AI needs a continual review process for emerging threats that may not have yet evolved in the cyber world.
3. Prepare for Attacks that are Adversarial in Nature
By testing the weaknesses of your AI and simulating the types of attacks that will be most detrimental to your AI, you will understand where the boundaries of your AI lie.
4. Balance Offensive and Defensive Usage
Do not use AI on an offensive basis without having rigid policies, checks, and balances in place; doing so can easily escalate tensions or produce collateral damage that was not intended.
AI does not represent a singular, automatic or down-below-all option. Rather, it is a tool through which the intended usefulness of AI can be amplified. By having concerns over the way AI will amplify your intentions, the use of AI to spot threats sooner, to react with greater efficiency, and to provide a tactical edge for defenders will be more effective.
In the realm of cyber warfare, being aware of how to manipulate the capabilities of AI is just as important as understanding its inherent limitations.
