Skip to main content

Shared Responsibility Model in the CDN

The Shared Responsibility Model specifies which part of the CDN service's security is Arvancloud's responsibility and which part is the user's. Arvancloud secures the infrastructure, meaning the servers, network, and platform, and the user manages the security of the settings they configure on this infrastructure, such as firewall rules, cache policies, and access to the origin server.

AreaArvancloud's ResponsibilityUser's Responsibility
Infrastructure and edge serversPhysical security of data centers, operating system updates, and patching edge server vulnerabilities—
Layer 3 and 4 DDoSAutomatic detection and mitigation of attacks up to 2 Tbps—
DNSNameserver stability, DNSSEC implementation, and key rotationManaging records and the cloud icon status (traffic passing through the CDN)
Firewall and Rate LimitProviding the firewall and rate limiting engineDefining rules, IP lists, and restricted paths
WAFUpdating attack signatures in the WAF coreSetting sensitivity, reviewing logs, and managing false positives
Layer 7 DDoSProviding Cookie, JavaScript, and Captcha challengesChoosing the challenge type, setting the TTL, and designing custom pages
Origin server—Keeping the IP hidden, whitelisting Arvancloud IPs, configuring load balancing and health checks
CacheStoring and serving content based on the policies you setSpecifying cacheable content and purging the cache when needed
SSL/TLSAutomatic issuance and renewal of free certificates, updating cipher suitesEnabling HTTPS and HSTS, setting the minimum TLS version, managing custom certificates
Edge ComputingIsolating the code execution environment between customersCode security, managing sensitive data and access levels
API keysStoring keys in hashed formKeeping them secure, least-privilege access, and revoking leaked keys
Logs and metricsGenerating and sending logs, metrics, and JA3 hashesAnalyzing the data and responding to attacks

These two parts depend on each other, and a weakness in one makes the other ineffective. For example, Arvancloud mitigates DDoS attacks up to a capacity of 2 Tbps, but if your origin server's IP is exposed, an attacker can bypass the CDN and attack the server directly.

Arvancloud's Responsibilities​

Infrastructure and Edge Servers​

Arvancloud has more than 40 PoP sites with an Anycast architecture and geographic load balancing (GSLB), and answers each request from the nearest node. The physical security of data centers, from emergency power, biometric access control, CCTV, and cooling system security to preventing unauthorized access to hardware, is the responsibility of Arvancloud and its infrastructure partners.

Arvancloud's engineering teams keep the edge servers' operating systems up to date, patch kernel vulnerabilities, and secure the software that routes HTTP/HTTPS and layer 4 proxy traffic. The user has no root access to these layers and is not responsible for their vulnerabilities.

Mitigating Layer 3 and 4 DDoS Attacks​

The Arvancloud network automatically detects and blocks volumetric and protocol attacks such as SYN Flood, UDP Flood, and ICMP Flood in real time, up to a capacity of 2 Tbps. This protection is enabled by default and free of charge on the CDN and DNS services, and requires no hardware or software installation on your network. Arvancloud drops malicious packets at the network edge before they reach your server and delivers legitimate traffic to its destination without noticeable delay.

DNS and DNSSEC​

Arvancloud protects the Cloud DNS infrastructure against attacks such as DNS amplification and direct attacks on nameservers, and answers requests from multiple geographic points with an Anycast architecture. DNSSEC prevents record forgery (DNS spoofing and cache poisoning) by cryptographically signing records, and key management, key rotation, and correct implementation of this protocol on Arvancloud servers are Arvancloud's responsibility. You can also move your existing records with a Zone File in BIND format and use custom nameservers (Custom NS). In both cases, Arvancloud keeps each customer's data isolated from other customers.

SSL/TLS Infrastructure​

Arvancloud can issue and renew SSL/TLS certificates for all registered domains for free and automatically, with DNS validation. The security of key exchange at the network edge, HTTP/2 support, updating cipher suites based on the latest security recommendations, and removing vulnerable algorithms are also Arvancloud's responsibility.

Edge Computing Runtime​

With Edge Computing, you can run your code serverlessly on more than 40 Arvancloud PoP sites. In this service, different customers' code runs on shared physical infrastructure. Arvancloud uses sandboxing and process isolation to prevent one customer from accessing another customer's data, environment variables, or resources. Allocating resources and memory, preventing side-channel attacks at the processor level, and monitoring the health of the function runtime are also part of this responsibility.

User's Responsibilities​

Most security breaches in cloud environments result from misconfiguration and poor access management, not from infrastructure weaknesses. The following sections explain the settings that the user is responsible for managing.

Firewall and Rate Limiting​

Arvancloud provides the firewall engine, and the user designs the traffic blocking logic. Depending on your plan, you can define from 5 to 1000 firewall rules.

PartUser's Responsibility
Conditions (Filter Expressions)Combining variables such as IP address, geographic location, HTTP Host, and request method to build rules
ActionChoosing between Block, Allow, Bypass, and Challenge
IP listsCreating and updating lists of allowed or attacker addresses in the dashboard and using them in firewall rules
Rate LimitSetting the number of requests allowed within a specific time period for sensitive paths such as the login page or payment gateway

Rate Limit prevents brute-force attacks, web scraping, and API overload. Note that Rate Limit only checks requests that the edge server doesn't answer from the cache (Miss or Bypass). Arvancloud's security modules also run on every request in a fixed order: first the firewall, then DDoS Protection, then Rate Limit, and finally WAF. Design your rules with this order in mind.

WAF​

The Arvancloud WAF engine works with the Anomaly Scoring method. The WAF gives each suspicious pattern in a request a score, and if the total score exceeds the threshold, it blocks the request. Updating attack signatures is Arvancloud's responsibility, and configuring, fine-tuning, and monitoring the WAF is the user's responsibility.

The following table shows the most important attack codes whose logs you should review:

CodeAttack TypeConsequence of Misconfiguration
41xxxSQL InjectionData leaks, administrative access to the database, and changed records
42xxxXSSSession hijacking and redirecting users to phishing sites
40xxxGeneral attacks such as LFI and RFIRunning malicious files on the origin server and unauthorized access to system files
35xxxMalicious botsAbuse of forms, distorted site statistics, and bypassing firewall rules
20xxx and 21xxxAnomalous HTTP requestsExploiting origin web server bugs with tampered headers

To set sensitivity, enable one of the rule packages, such as CRS (Core Rule Set) or Arvancloud's managed packages, and set the sensitivity level.

Note that in the Arvancloud WAF, the lower the sensitivity number, the stricter the WAF, and the higher the chance of blocking real users' requests (false positives).

Before enabling blocking mode, keep the WAF in Detection mode for a while and review the logs to find legitimate requests that the WAF wrongly detected as malicious. Then disable the faulty rules by their IDs, and after that enable the WAF. Without this process, the WAF may block users' legitimate transactions and your business's conversion rate may drop.

If your APIs have complex data structures, use custom rules to enable or disable the WAF for specific paths or trusted IP ranges.

Note that with the WAF enabled, the maximum request body size is 524,288 bytes, and you should design file upload forms with this limit in mind.

Layer 7 DDoS​

Layer 7 attacks imitate real user behavior, and automatic layer 3 and 4 protection doesn't detect them. Arvancloud provides Cookie, JavaScript, and Captcha challenges, and the user selects and enables the challenge type in the dashboard based on the attack's severity, from the unobtrusive Cookie challenge to the interactive Captcha.

The challenge validity time (TTL) specifies how long a visitor won't face the challenge again after solving it. An inappropriate value disrupts the user experience.

To maintain your brand's visual identity, you can upload the HTML of the 500 error pages or the DDoS challenge page in the panel.

Note that using the %Challenge% variable in these pages is required, and testing that it works correctly is your team's responsibility.

Securing the Origin Server​

All of the CDN's defense layers, from WAF to Rate Limit and DDoS protection, are only effective when traffic passes through the CDN. If an attacker finds your origin server's IP, they can bypass Arvancloud and send malicious traffic directly to the server. That is why keeping the origin server's IP hidden is your most important responsibility.

Keep the cloud icon on for A and CNAME records, especially the @ and www records, so traffic passes through the network edge and the origin server's IP stays hidden.

Note that turning on the cloud icon for non-HTTP/HTTPS services such as email or FTP causes disruptions, so manage the records of these services separately.

After you enable the CDN, all requests reach the origin server from Arvancloud IPs, and the server's local firewall may detect this concentrated traffic as an attack and block it. To prevent this, whitelist Arvancloud IPs in the origin server's firewall. At the same time, block access to ports 80 and 443 for all other IPs so direct connections to the server are not possible.

If you have several origin servers, choose the load balancing algorithm that suits your software architecture. Round Robin distributes traffic among servers in turn and equally, and Client IP Hash sends each user to a fixed server based on the hash of their IP, which is suitable for keeping the user's session.

Active Health Check sends requests to origin servers with the User Agent value Arvancloud-AHC. Set the expected status codes and responses precisely so healthy servers aren't taken out of service by mistake. Also, by enabling Next Upstream, the user's request goes to the next server when an error occurs.

Cache Policies and Privacy​

The decision about what content is cached is yours alone. One mistake in the cache configuration can show one user's identity or financial information to other users. Set cache policies in Page Rules in the dashboard or with the Cache-Control header in the origin server's response.

Cache-Control HeaderEdge Server BehaviorSecurity Consideration
publicCaches the content for all users regardless of other directives.Don't use it for pages with identity or financial information (PII) or dynamic content.
privateDoesn't cache the content and leaves storing it only to the user's browser.Suitable for profile pages, shopping carts, and personal dashboards
no-storeDoesn't cache the content anywhere along the path and gets each request from the origin server.Suitable for banking transactions and highly confidential data
max-ageSets the expiration time of cached content.Set it based on how fast the content changes so users don't see stale content.
min-usesCaches the content after a specific number of requests.For managing edge resources for popular content

If content is cached and published by mistake, immediately clear it with Purge Cache for specific URLs or the whole domain, from the panel or via API.

The maximum cacheable size for each file is 128 to 2048 MB depending on the plan, and it is better to choose how you distribute large files with this limit in mind.

HTTPS and TLS Policies​

Arvancloud provides the secure connection infrastructure, and applying stricter policies is up to you. Enable automatic redirection of HTTP traffic to HTTPS. You can also enable HSTS (HTTP Strict Transport Security), which prevents downgrade and Man-in-the-Middle attacks.

Note that if you set a long HSTS expiration time and the site's certificate has a problem, users won't be able to access the site for a long time.

If your site's HTML contains HTTP links, the secure connection runs into a Mixed Content problem. To fix it, enable automatic link replacement.

To comply with standards such as PCI-DSS on payment gateways, set the minimum TLS version so that old clients with insecure protocols can't connect.

If you use a custom certificate instead of the free Arvancloud certificate, you can upload it as a PEM file. In this case, keeping the private key secure and renewing the certificate on time before it expires are your responsibility.

Code Security in Edge Computing​

Arvancloud guarantees the security of the runtime for JavaScript code at the edge, but the security of the code logic you write is your own responsibility. With edge scripts, you can route traffic based on the user's geographic location (GeoIP) or request headers, or upload data directly from the network edge to Arvancloud Object Storage. For example, if you use HTTP Basic Authentication in an edge script, the script sends the username and password unencrypted. In this case, make sure the connection is only over HTTPS and don't hardcode sensitive information in the code. Any breach resulting from logic bugs, insecure use of variables, or logging sensitive information in edge logs is your responsibility.

If your edge code is connected to Arvancloud Object Storage, setting bucket access levels (Bucket Policy) is also your responsibility. Note that making a bucket public exposes the list of objects inside it to everyone.

API Keys and Third-Party Applications​

To work with the API, CLI, or Terraform, you need to create an access key. These keys are effectively like passwords for managing your infrastructure. Keep keys secure, don't put them in public code such as GitHub repositories, and follow the principle of least privilege for machine users.

Arvancloud stores keys in hashed form, and they can't be retrieved after they are first shown. If you suspect a key has been leaked, revoke and delete it from the dashboard immediately.

Shared Areas: Logs, Metrics, and Traffic Analysis​

In this area, Arvancloud generates and sends security data, and you analyze it and act on it.

Arvancloud sends logs with Log Forwarding and time-series metrics with Metric Exporter. Metrics are lightweight and structured and suit real-time monitoring and alerting in tools such as Grafana, Prometheus, and Splunk. Edge servers also add the headers X-Real-IP (the user's real IP address), X-Country-Code (the country code), X-Forwarded-Proto (the original connection protocol), and X-Request-Id (the unique request ID) to requests. Analyzing this data is your security team's responsibility, and you should continuously monitor the following metrics:

MetricUse
firewall_rule_matched_countNumber of requests matching firewall rules, for detecting a new wave of attacks or problems in rules
rate_limit_action_countNumber of times Rate Limit was applied. A sudden increase is a sign of a brute-force or scraping attack.
ddos_challenge_action_countNumber of DDoS challenges shown, to make sure layer 7 defense works correctly without disrupting the user experience
top_request_tlsfingerprintThe 10 TLS fingerprints with the most traffic, for identifying botnets with identical behavior

During the TLS handshake, Arvancloud analyzes parameters such as the client's TLS version, list of cipher suites, extensions, and elliptic curves, builds a unique JA3 hash from them, and puts it in the logs. Attackers can easily change their IP, User-Agent, and HTTP headers, but a botnet's JA3 hash usually stays the same. If you see in the logs that one JA3 hash is sending a large volume of malicious requests from different IPs, block that hash with a firewall rule in the dashboard to stop the entire botnet without harming real users.