The lowest level of the Java security architecture involves the design of the Java Virtual Machine and the byte codes it executes. The Java VM does not allow any kind of direct access to individual memory addresses of the underlying system, which prevents Java code from interfering with the native hardware and operating system. These intentional restrictions on the VM are reflected in the Java language itself, which does not support pointers or pointer arithmetic. The language does not allow an integer to be cast to an object reference or vice versa, and there is no way whatsoever to obtain an object's address in memory. Without capabilities like these, malicious code simply cannot gain a foothold.
In addition to the secure design of the Virtual Machine instruction set, the VM goes through a process known as byte-codeverification whenever it loads an untrusted class. This process ensures that the byte codes of a class (and their operands) are all valid; that the code never underflows or overflows the VM stack; that local variables are not used before they are initialized; that field, method, and class access control modifiers are respected; and so on. The verification step is designed to prevent the VM from executing byte codes that might crash it or put it into an undefined and untested state where it might be vulnerable to other attacks by malicious code. Byte-code verification is a defense against malicious hand-crafted Java byte codes and untrusted Java compilers that might output invalid byte codes.
 |  |  |
| 5.1. Security Risks |  | 5.3. Authentication and Cryptography |
Copyright © 2001 O'Reilly & Associates. All rights reserved.