[实践总结] 限制正则表达式匹配次数/时间 防止DoS攻击

思路

1、优化正则表达式
2、正则表达式无法优化的话，可以考虑限制匹配次数，或者限制匹配时间

限制 匹配次数

public class CountedCharSequence implements CharSequence {
    private final CharSequence charSequence;
    private long count;

    public CountedCharSequence(CharSequence charSequence, long count) {
        this.charSequence = charSequence;
        this.count = count;
    }

    @Override
    public String toString() {
        return charSequence.toString();
    }

    @Override
    public int length() {
        return charSequence.length();
    }

    @Override
    public char charAt(int index) {
        if (index <= 0) {
            throw new IllegalArgumentException("Regex match over max " + count + " times");
        }
        count--;
        return charSequence.charAt(index);
    }

    @Override
    public CharSequence subSequence(int start, int end) {
        return charSequence.subSequence(start, end);
    }
}

模拟Dos攻击

String str = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";

long start = System.nanoTime();
Pattern.compile("^(a+)+$").matcher(str).matches();
System.out.println(((System.nanoTime() - start)) + " nano seconds");

执行用时
1399200 nano seconds

超过 1亿 就会报错

long start1 = System.nanoTime();

try {
    Pattern.compile("^(a+)+$").matcher(new CountedCharSequence(str, 100_000_000)).matches();
} catch (Exception e) {
    System.out.println(e.getMessage());
}

System.out.println(((System.nanoTime() - start1)) + " nano seconds");

执行用时
Regex match over max 100000000 times
601800 nano seconds

限制 匹配时间

import java.util.regex.Pattern;

public final class TimedCharSequence implements CharSequence {
    private final CharSequence sequence;
    private final long timestamp;

    public TimedCharSequence(CharSequence sequence, long nanoseconds) {
        this.sequence = sequence;
        this.timestamp = System.nanoTime() + nanoseconds;
    }

    @Override
    public String toString() {
        return sequence.toString();
    }

    @Override
    public int length() {
        return sequence.length();
    }

    @Override
    public char charAt(int index) {
        if (timestamp < System.nanoTime()) {
            throw new IllegalStateException("Regex match timeout");
        }
        return sequence.charAt(index);
    }

    @Override
    public CharSequence subSequence(int start, int end) {
        return sequence.subSequence(start, end);
    }
}

模拟Dos攻击

String str = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";

long start = System.nanoTime();
Pattern.compile("^(a+)+$").matcher(str).matches();
System.out.println(((System.nanoTime() - start)) + " nano seconds");

执行用时：
1385400 nano seconds

超过 给定时间间隔就报错

long start1 = System.nanoTime();

try {
    Pattern.compile("^(a+)+$").matcher(new TimedCharSequence(str, 10_000)).matches();
} catch (Exception e) {
    System.out.println(e.getMessage());
}

System.out.println(((System.nanoTime() - start1)) + " nano seconds");

执行用时：
Regex match timeout
123100 nano seconds

[Ref]

[Ref] 限制正则表达式执行时间