To quickly see what’s happening and better understand results, turn to the AI Assistant. Create the view you want simply by asking for it. This new easy way to navigate Web Security Analytics lets you:

• enter intuitive natural language prompts to see results you want instead of time-consuming hunt-and-peck selections.
• apply filters, customize views, and switch between views effortlessly.
• perform in-depth analysis faster with less steps.
• skip the need for technical security reporting expertise.

Sample commands

You can ask the AI Assistant things like:

Show me traffic from the last 24 hours.

Display only SQL injection attacks.

Add a filter to show only mitigated requests.

Take me to bot analysis view.

Switch back to my first view.

Add a bar chart that shows attack types.

Add a widget showing connecting IP addresses.

If the assistant can't process your request, try to rephrase the query or break it down into simpler terms. Our team is continuously improving the AI Assistant’s understanding of the human language, and we encourage you to report any limitations you encounter to help us get better.

📘

Can the assistant store my conversations?

The AI Assistant stores conversation history (your prompt and the command output by the AI Assistant) during your active session. If you close the chat window or log out, the history is cleared. Your past queries during a session only remain available as long as the session is active. Conversations from earlier sessions have been cleared and are not available anymore.

Try the assistant

To get started with AI Assistant, contact your account team today and let them know. There’s no additional charge, but you do need to opt in to enjoy all the benefits AI Assistant offers.

📘

Tell us what you think

Share your experience with us. Click the thumbs up icon, if you're happy with results, or click thumbs down to share feedback on how we can improve.

Attack Insights add-on (beta)

Quickly prioritize and act on critical threats. This feature surfaces real-time detection and detailed analysis of potential security threats, so you can respond immediately. It uses advanced detection models developed by Akamai’s Threat Research Team to identify unusual traffic patterns, potential malicious attacks, and suspicious anomalies. It helps you optimize your security posture with:

• Anomaly detection. Identifies significant deviations in traffic behavior that may indicate an attack.
• Cross-platform intelligence. Leverages anonymized data across Akamai’s enormous platform to identify broader attack patterns through a uniquely wide lens.
• Contextualized insights. Provides detailed explanations on why requests may be suspicious, including triggered rules and relevant security events.
• Response actions. Notifies your security team about detected threats with possible next steps.
• Event exploration. Lets you drill down into attack details in a single click that automatically sets and displays the relevant filters and views.

Use insights

Attack insights continually monitors security events and surfaces the most significant issues. When you see an insight appear in the AI Assistant screen, you can click its More info button to view full details and see possible actions.

Try insights

Contact your account team and let them know you’re interested in a free 60-day trial of this premium add-on feature. Or, within AI Assistant, click the Insights tab, then click Ask for trial activation. Available beta for eligible customers.

Data protection

Security is a top priority at Akamai. The AI Assistant operates within the secure framework of our Web Security Analytics platform and adheres to all necessary security and privacy standards to ensure that the submitted data remains protected. Key security measures include:

• Data segregation. The AI Assistant ensures data security by segregating user input and implementing strict data security measures, like encryption (both in transit and at rest for the prompts and the output) and robust access controls. No user can query another user’s data; the queries are limited to the specific WAF config users work on (this is part of existing data segregation measures applied in Web Security Analytics).
• Authentication and authorization. User authentication is integrated into Akamai’s Control Center as part of the Web Security Analytics service, ensuring secure access based on user roles.
• Secure development practices. We built the AI Assistant following secure coding standards and run regular vulnerability assessments, code reviews, and security tests (DAST and SAST) to ensure its robustness.
• AI model and decision security. The AI Assistant is designed to be resilient against adversarial attacks and data manipulation, undergoing continuous testing and validation to maintain its integrity.
• Real-time monitoring and alerts. Our system includes real-time monitoring for abnormal behavior and anomalies, with logging and automated alerts to notify administrators of any potential threats or security issues.
• Classifiers. Users can’t make just any queries, and their input is segmented into a set of classifications. If the input does not meet one of these classifications, the AI Assistant returns an error message to the user.

About the model

AI Assistant is powered by Llama 3.1-70B, developed by Meta Platforms, Inc. This model is a state-of-the-art LLM that enables the AI Assistant to interpret natural language into database queries to create respective reports that provide relevant attack data insights. Leveraging Llama 3.1-70B allows the AI Assistant to make it easy for you to run complex queries to the Web Security Analytics database by converting human language into filters and commands. The generated queries then offer actionable data insights in a secure and efficient way within Akamai’s environment.

We use the accepted responses—those that receive a thumbs-up from users—as positive examples to help improve the model’s accuracy. This feedback is essential for fine-tuning the model and ensuring it better meets your needs. The prompts and accepted responses are kept confidential within Akamai, secured and not shared with third parties.

The prompts you enter into the AI Assistant and the output are stored for up to 120 days to help improve and enhance the model’s performance. During this period, data is securely stored in Elasticsearch (part of the ELK stack) on Akamai’s Microsoft Azure instance used for Web Security Analytics, specifically within the eastern U.S. This storage setup adheres to Akamai’s strict privacy and security protocols, ensuring data protection and confidentiality. For more details on Akamai’s use of Microsoft Azure, see Akamai’s sub-processor list.

While the WSA AI Assistant uses Meta Platform Inc.’s Llama 3.1-70B model, and Microsoft Azure’s hosting services for prompt and output data storage, rest assured that we are not using external APIs to handle prompts and output queries. All prompts and data are processed within Akamai’s secure platform.

📘

Is the model trained on my and other customers' data?

No, the AI Assistant is not trained on customer data. The purpose of the AI Assistant is to simplify creating queries to better understand one’s attack data and navigation within Web Security Analytics reporting capabilities. Instead of relying on the technical skills of the user, the AI Assistant transfers human language into filters and commands to create queries. The AI Assistant is not used to perform the queries. Data collection in the form of query performance is done outside the AI Assistant.

We store only a user’s prompt and the AI Assistant’s corresponding system response to improve future interactions. For example, if you ask for WAF-triggered events from the past week, the stored data includes your prompt and the system’s response, like Attack type == WAF and time range is the past week. No actual customer data or results shown in Web Security Analytics are stored in the model. This approach ensures that customer data remains secure and is never used for this AI use case.