MainLogBuffer.java revision 1a2733683769e6f1267c86768131fc89bbb4b3d6
1/* 2 * Copyright (C) 2012 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); you may not 5 * use this file except in compliance with the License. You may obtain a copy of 6 * the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 * License for the specific language governing permissions and limitations under 14 * the License. 15 */ 16 17package com.android.inputmethod.research; 18 19import android.util.Log; 20 21import com.android.inputmethod.latin.Dictionary; 22import com.android.inputmethod.latin.Suggest; 23 24import java.util.Random; 25 26public class MainLogBuffer extends LogBuffer { 27 private static final String TAG = MainLogBuffer.class.getSimpleName(); 28 // For privacy reasons, be sure to set to "false" for production code. 29 private static final boolean DEBUG = false; 30 31 // The size of the n-grams logged. E.g. N_GRAM_SIZE = 2 means to sample bigrams. 32 private static final int N_GRAM_SIZE = 2; 33 // The number of words between n-grams to omit from the log. 34 private static final int DEFAULT_NUMBER_OF_WORDS_BETWEEN_SAMPLES = DEBUG ? 2 : 18; 35 36 private final ResearchLog mResearchLog; 37 private Suggest mSuggest; 38 39 // The minimum periodicity with which n-grams can be sampled. E.g. mWinWordPeriod is 10 if 40 // every 10th bigram is sampled, i.e., words 1-8 are not, but the bigram at words 9 and 10, etc. 41 // for 11-18, and the bigram at words 19 and 20. If an n-gram is not safe (e.g. it contains a 42 // number in the middle or an out-of-vocabulary word), then sampling is delayed until a safe 43 // n-gram does appear. 44 /* package for test */ int mMinWordPeriod; 45 46 // Counter for words left to suppress before an n-gram can be sampled. Reset to mMinWordPeriod 47 // after a sample is taken. 48 /* package for test */ int mWordsUntilSafeToSample; 49 50 public MainLogBuffer(final ResearchLog researchLog) { 51 super(N_GRAM_SIZE); 52 mResearchLog = researchLog; 53 mMinWordPeriod = DEFAULT_NUMBER_OF_WORDS_BETWEEN_SAMPLES + N_GRAM_SIZE; 54 final Random random = new Random(); 55 mWordsUntilSafeToSample = random.nextInt(mMinWordPeriod); 56 } 57 58 public void setSuggest(Suggest suggest) { 59 mSuggest = suggest; 60 } 61 62 @Override 63 public void shiftIn(final LogUnit newLogUnit) { 64 super.shiftIn(newLogUnit); 65 if (newLogUnit.hasWord()) { 66 if (mWordsUntilSafeToSample > 0) { 67 mWordsUntilSafeToSample--; 68 } 69 } 70 if (DEBUG) { 71 Log.d(TAG, "shiftedIn " + (newLogUnit.hasWord() ? newLogUnit.getWord() : "")); 72 } 73 } 74 75 public void resetWordCounter() { 76 mWordsUntilSafeToSample = mMinWordPeriod; 77 } 78 79 /** 80 * Determines whether the content of the MainLogBuffer can be safely uploaded in its complete 81 * form and still protect the user's privacy. 82 * 83 * The size of the MainLogBuffer is just enough to hold one n-gram, its corrections, and any 84 * non-character data that is typed between words. The decision about privacy is made based on 85 * the buffer's entire content. If it is decided that the privacy risks are too great to upload 86 * the contents of this buffer, a censored version of the LogItems may still be uploaded. E.g., 87 * the screen orientation and other characteristics about the device can be uploaded without 88 * revealing much about the user. 89 */ 90 public boolean isSafeToLog() { 91 // Check that we are not sampling too frequently. Having sampled recently might disclose 92 // too much of the user's intended meaning. 93 if (mWordsUntilSafeToSample > 0) { 94 return false; 95 } 96 if (mSuggest == null || !mSuggest.hasMainDictionary()) { 97 // Main dictionary is unavailable. Since we cannot check it, we cannot tell if a word 98 // is out-of-vocabulary or not. Therefore, we must judge the entire buffer contents to 99 // potentially pose a privacy risk. 100 return false; 101 } 102 // Reload the dictionary in case it has changed (e.g., because the user has changed 103 // languages). 104 final Dictionary dictionary = mSuggest.getMainDictionary(); 105 if (dictionary == null) { 106 return false; 107 } 108 // Check each word in the buffer. If any word poses a privacy threat, we cannot upload the 109 // complete buffer contents in detail. 110 final int length = mLogUnits.size(); 111 for (int i = 0; i < length; i++) { 112 final LogUnit logUnit = mLogUnits.get(i); 113 final String word = logUnit.getWord(); 114 if (word == null) { 115 // Digits outside words are a privacy threat. 116 if (logUnit.mayContainDigit()) { 117 return false; 118 } 119 } else { 120 // Words not in the dictionary are a privacy threat. 121 if (ResearchLogger.hasLetters(word) && !(dictionary.isValidWord(word))) { 122 if (DEBUG) { 123 Log.d(TAG, "NOT SAFE!: hasLetters: " + ResearchLogger.hasLetters(word) 124 + ", isValid: " + (dictionary.isValidWord(word))); 125 } 126 return false; 127 } 128 } 129 } 130 // All checks have passed; this buffer's content can be safely uploaded. 131 return true; 132 } 133 134 @Override 135 protected void onShiftOut(LogUnit logUnit) { 136 if (mResearchLog != null) { 137 mResearchLog.publish(logUnit, false /* isIncludingPrivateData */); 138 } 139 } 140} 141