AI RiskThu, 22 Oct 2026
② AI-assisted app and API attacks in Japan: Enforce authorization on the server
JPCERT/CC found evidence of AI-assisted scanning and API attacks. Check what a public app reveals, what its API authorizes, and where its backend can connect.
Article
On October 15, JPCERT/CC published its investigation into a series of unauthorized access incidents in Japan. It found AI-assisted scanning for known vulnerabilities, attacks through APIs and signs that an attacker-controlled server had been used for APK analysis and SSRF tests.
Investigators found files associated with Blutter and jadx, both used for app reverse engineering, and Python code for SSRF testing. The report did not identify the attackers’ AI model or establish end-to-end execution by an autonomous agent.
- 01
Analyze the public app
Reverse-engineering tools can reveal backend URLs and authentication flows.
- 02
Identify API access paths
An attacker may find embedded secrets or endpoints that trust the app.
- 03
Probe the backend
Test for SSRF, authorization flaws or misconfigurations that expose data or internal services.
What to do: from the perimeter to internal systems
- 01
Inventory apps and APIs
Include production, testing, older app versions, containers and outsourced backends.
See what to doResources: 1
TWISA member solutions
Inventory apps and APIs
Include production, testing, older app versions, containers and outsourced backends.
- 02
Constrain credentials and egress
Keep long-lived secrets out of apps. Validate URL destinations after redirects and restrict internal access.
See what to doResources: 1
TWISA member solutions
Constrain credentials and egress
Keep long-lived secrets out of apps. Validate URL destinations after redirects and restrict internal access.
- 03
Test authorization
Use separate test identities to verify that every object and action is authorized on the server.
See what to doResources: 1
TWISA member solutions
Test authorization
Use separate test identities to verify that every object and action is authorized on the server.
- 04
Retest with logs
Verify that unauthorized requests fail at the backend. Correlate alerts with the actual data returned.
See what to doResources: 1
TWISA member solutions
Retest with logs
Verify that unauthorized requests fail at the backend. Correlate alerts with the actual data returned.
Map app versions, API endpoints, keys, outsourced backends and cross-border data flows.
Allow authorized traffic and apply controls at each layer.
- 01
Public app
Mobile apps, API endpoints and keys
Keep secrets off the clientReview app binaries for reusable credentials.
- 02
API authorization
Identities, roles and object access
Authorize each actionValidate the caller and object on the server.
- 03
Backend egress
URL fetching and redirects
Constrain SSRF pathsValidate final destinations and limit connections to internal services.
Feed logs and traffic data from each layer into monitoring and response.
- Monitoring and validationRate-limit individual endpoints as needed and investigate unauthorized access attempts or unexpected outbound traffic.
TWISA member solutions
Inventory apps and APIs
Include production, testing, older app versions, containers and outsourced backends.
Constrain credentials and egress
Keep long-lived secrets out of apps. Validate URL destinations after redirects and restrict internal access.
Test authorization
Use separate test identities to verify that every object and action is authorized on the server.
Retest with logs
Verify that unauthorized requests fail at the backend. Correlate alerts with the actual data returned.
Continue the discussion at CIX
Bring examples of cross-account API tests or URL-fetching controls to a future CIX meetup. The case fits the second session’s discussion of defense in depth because identity, data authorization and egress all matter after an attacker has inspected the app.
Sources and further reading
Column updated: 2026/10/25.