FortiSandbox examines suspicious files through static analysis and execution in an isolated environment to detect malware that signatures miss.
FortiSandbox is Fortinet's sandbox solution: it analyzes suspicious files and URLs by examining them statically and then, if necessary, running them in an isolated environment to observe their behavior. The product is active in 2026, with a 5.x branch that expands the use of artificial intelligence. Its goal is to detect unknown threats that antivirus signatures do not yet recognize.
An antivirus recognizes what it already knows. New malware, or malware that has simply been modified, can slip through the net. A sandbox takes a different approach: it opens the file in a controlled virtual machine and observes what it does. Does it create processes, modify the registry, contact an external server, encrypt files? These behaviors make it possible to reach a verdict even when no signature exists.
Attackers are aware of this and design malware capable of detecting a virtual environment or of waiting before acting. An effective sandbox must therefore combine several complementary techniques rather than relying solely on the execution of the file.
FortiSandbox runs through several levels of analysis:
The observed behaviors are mapped to the MITRE ATT&CK framework, giving the analyst a readable report: which techniques the file attempted to use and which attack phase they correspond to. FortiSandbox also offers a mode dedicated to OT environments, with emulation of industrial services such as Modbus, BACnet or S7comm.
FortiSandbox receives samples through several channels:
Verdicts are then redistributed: a file identified as malicious can be blocked on other devices, and the associated indicators can be shared.
By default, the sandbox often works in parallel: the file is delivered to the user while the analysis is under way, and an alert is raised if the verdict is unfavorable. Since FortiSandbox 4.2 and FortiOS 7.2, an inline blocking feature allows the FortiGate to hold the file for the duration of the analysis.
This mode improves protection but involves trade-offs. Static analysis generally returns a verdict within a few seconds, whereas dynamic analysis can take around a minute for an office document and longer for an executable. The documentation recommends sufficiently powerful models and specifies that if the timeout is exceeded, the default behavior of the FortiGate is to let the file through; it can be configured to block it instead. This setting should be chosen deliberately, in agreement with the business units.
FortiSandbox is available as hardware appliances of different capacities, as virtual machines that can be installed on premises or in public clouds, and as services hosted by Fortinet. For smaller organizations, a cloud sandbox service can also be associated with FortiGate devices through a subscription. The choice depends on the volume of files to be analyzed, confidentiality constraints (a file sent to the cloud leaves your infrastructure) and the guest operating systems required.
Operating a sandbox means knowing how to read an analysis report, understanding evasion techniques and linking a verdict to an operational response. The following training paths are useful for this purpose:
FortiSandbox complements antivirus by analyzing what signatures still miss, through AI-based static analysis and controlled execution. Its value depends on deployment choices, on the balance struck between detection and inline blocking, and on the ability of teams to make use of its reports.