Language detection does not work when a SecurityManager is enabled

[ctalau]

If a SecurityManager is enabled in the JVM, the language detection does not work.

Sample code to reproduce the issue:

import java.io.File;
import java.io.IOException;
import java.nio.file.Files;

import com.github.pemistahl.lingua.api.Language;
import com.github.pemistahl.lingua.api.LanguageDetector;
import com.github.pemistahl.lingua.api.LanguageDetectorBuilder;

public class SecurityManagerApp {
  public static Language[] languages = new Language[] { Language.ENGLISH, Language.FRENCH};
  
  public static void main(String[] args) throws IOException {
    File policyFile = File.createTempFile("security", ".policy", null);
    Files.write(policyFile.toPath(), 
        "grant { permission java.security.AllPermission; }; ".getBytes());
    System.setProperty("java.security.policy", policyFile.getAbsolutePath());
    System.setSecurityManager(new SecurityManager());
    
    LanguageDetector detector = LanguageDetectorBuilder.fromLanguages(languages).build();
    System.out.println(detector.detectLanguageOf("Comment ca va\r\n"
        + "Comme ci\r\n"
        + "Comme ci\r\n"
        + "Comme ci\r\n"
        + "Comme ca").name());
  }
}

Note that in my use-case, I am creating a plugin for an application, so I cannot disable the SecurityManager of the JVM.

[ctalau]

If running the sample code with -Djava.security.debug=access,failure I obtained the following stack trace:

access: access denied ("java.io.FilePermission" "D:\opt\.m2\com\github\pemistahl\lingua\1.2.1\lingua-1.2.1.jar" "read")
java.lang.Exception: Stack trace
	at java.base/java.lang.Thread.dumpStack(Thread.java:1380)
	at java.base/java.security.AccessControlContext.checkPermission(AccessControlContext.java:475)
	at java.base/java.security.AccessController.checkPermission(AccessController.java:1068)
	at java.base/java.lang.SecurityManager.checkPermission(SecurityManager.java:416)
	at java.base/jdk.internal.loader.URLClassPath.check(URLClassPath.java:556)
	at java.base/jdk.internal.loader.URLClassPath.checkURL(URLClassPath.java:535)
	at java.base/jdk.internal.loader.BuiltinClassLoader.checkURL(BuiltinClassLoader.java:1080)
	at java.base/jdk.internal.loader.BuiltinClassLoader.findResource(BuiltinClassLoader.java:356)
	at java.base/java.lang.ClassLoader.getResource(ClassLoader.java:1403)
	at java.base/java.lang.ClassLoader.getResourceAsStream(ClassLoader.java:1733)
	at java.base/java.lang.Class.getResourceAsStream(Class.java:2850)
	at com.github.pemistahl.lingua.api.LanguageDetector$Companion.loadLanguageModel(LanguageDetector.kt:530)
	at com.github.pemistahl.lingua.api.LanguageDetector$Companion.loadLanguageModels(LanguageDetector.kt:520)
	at com.github.pemistahl.lingua.api.LanguageDetector$Companion.access$loadLanguageModels(LanguageDetector.kt:501)
	at com.github.pemistahl.lingua.api.LanguageDetector.lookUpNgramProbability-as5wtIs$lingua(LanguageDetector.kt:467)
	at com.github.pemistahl.lingua.api.LanguageDetector.computeSumOfNgramProbabilities$lingua(LanguageDetector.kt:442)
	at com.github.pemistahl.lingua.api.LanguageDetector.computeLanguageProbabilities$lingua(LanguageDetector.kt:429)
	at com.github.pemistahl.lingua.api.LanguageDetector.computeLanguageConfidenceValues$lambda-4$lambda-3(LanguageDetector.kt:142)
	at java.base/java.util.concurrent.ForkJoinTask$AdaptedInterruptibleCallable.exec(ForkJoinTask.java:1461)
	at java.base/java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:373)
	at java.base/java.util.concurrent.ForkJoinPool$WorkQueue.topLevelExec(ForkJoinPool.java:1182)
	at java.base/java.util.concurrent.ForkJoinPool.scan(ForkJoinPool.java:1655)
	at java.base/java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:1622)
	at java.base/java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:165)
access: domain that failed ProtectionDomain  null
 null
 <no principals>
 null

The domain that did not have the required permissions was created by the ForkJoinPool. When using a SecurityManager, the commonPool uses a factory that creates InnocuousForkJoinWorkerThread instances which have that ProtectionDomain in their AccessControlContext.

I can see two solutions:

1. Do not use a commonPool, but a dedicated one. This has the advantage of being a good citizen and not blocking the common pool for other libraries.
2. Use AccessController.doPrivileged(...) to around the call to load the language models.

[pemistahl]

Hi Cristian @ctalau, thank you for this bug report. I will try to fix this bug soon.

[pemistahl]

Hi @ctalau, I was able to reproduce your problem and wrapped the respective code with AccessController.doPrivileged() which resolves it. I'm going to release Lingua 1.2.2 soon which includes this fix.

[Marcono1234]

Out of curiosity, if the issue occurred while loading the language models, does the fix then also help when using LanguageDetectorBuilder.withPreloadedLanguageModels()?