JP3879387B2 - Personal identification device - Google Patents

Description

【００５０】
但し、ＳＴは打鍵時間差の総和、ＪＴは打鍵時間差の自乗値の総和、ＣＴは出現頻度である。
【００５１】
そして、個人識別処置５０は、平均値ＡＶ及び標準偏差値ＨＶと、識別すべきユーザのキー操作から得られた打鍵時間差とに基づくファジイ推論によってスコアを算出する。
【００５２】
本実施の形態に係る個人識別装置１０では、前述の打鍵事象特徴値算出処理４４によって複数人の特徴値を特徴値テーブルとして登録しておき、該登録済みの複数人の特徴値の各々に対するスコアを上記のように算出し、一番高いスコアとなった特徴値の登録者がユーザであるものとして識別する。
【００５３】
なお、登録者が１人の場合は識別ではなく、照合による認証となる。この場合は、閾値を設けて判断することになる。照合の場合、又は識別の場合であってもスコアがある一定値以下なら認証しない形態とすることが好ましく、本実施の形態では、この形態を適用している。
（５）利用範囲制限処理について
利用範囲制限処理５２では、個人識別処置５０においてユーザが認証されなかった場合に、当該ユーザによって利用できる機能、リソース、及び機器を制限する。これによって、セキュリティを向上することができる。
【００５４】
打鍵事象検出処理４０が本発明の打鍵事象検出手段に、二連接打鍵事象導出処理４２が本発明の二連接打鍵事象導出手段に、打鍵事象特徴値算出処理４４が本発明の打鍵事象特徴値算出手段に、ハードディスク２０が本発明の記憶手段に、個人識別処理５０が本発明の識別手段に、利用範囲制限処理５２が本発明の制限手段に、各々相当する。
【００５５】
次に、個人識別装置１０の作用を説明する。まず、図５を参照して、前述の打鍵事象検出処理について説明する。なお、図５は、当該打鍵事象検出処理を実行するために、ＣＰＵ２２によって実行される打鍵事象検出処理プログラムの流れを示すフローチャートであり、該プログラムは予めＲＯＭ２４の所定領域に記憶されている。また、この打鍵事象検出処理プログラムは、フック関数によってメッセージが流れてきた際に随時実行される割込処理プログラムである。
【００５６】
まず、ステップ１００では、フック関数によって流れてきたメッセージを本来の通知先にそのまま通知し、次のステップ１０２では、上記メッセージがキーボード操作関係のメッセージであるか否かを判定し、キーボード操作関係のメッセージである場合（肯定判定の場合）はステップ１０４へ移行して、打鍵時刻及び打鍵コードを打鍵事象テーブルとしてハードディスク２０の打鍵事象テーブル領域ＤＡに記憶した後に本打鍵事象検出処理を終了し、キーボード操作関係のメッセージでない場合（ステップ１０２で否定判定された場合）には、ステップ１０４の処理を実行することなく本打鍵事象検出処理を終了する。
【００５７】
図６には、本打鍵事象検出処理によって得られる打鍵事象テーブルの一例が模式的に示されている。同図に示す例では、打鍵時刻としてミリ秒単位の時間が、打鍵コードとして１６進数のコード（例えば、‘０ｘ６ｃ’は１６進数の６ｃを表す）が記憶されている。なお、本打鍵事象検出処理は、前述のように、フック関数によってメッセージが流れてきた際に随時実行されるので、図６に示すように、ユーザがキーボード１４のキーを打鍵する度に、順次打鍵時刻及び打鍵コードが蓄積されることになる。
【００５８】
次に、図７を参照して、前述の二連接打鍵事象導出処理について説明する。なお、図７は、当該二連接打鍵事象導出処理を実行するために、ＣＰＵ２２によって実行される二連接打鍵事象導出処理プログラムの流れを示すフローチャートであり、該プログラムも予めＲＯＭ２４の所定領域に記憶されている。また、この二連接打鍵事象導出処理プログラムは、上記打鍵事象検出処理プログラムが実行された際に随時実行される割込処理プログラムである。
【００５９】
まず、ステップ２００では、ハードディスク２０に記憶されている打鍵事象テーブルを参照して、連接する打鍵事象の各打鍵間の打鍵コードの組合わせを抽出し、次のステップ２０２では、上記ステップ２００で抽出した打鍵コードの組合わせにおける各打鍵コードの打鍵時刻を打鍵事象テーブルから読み出して、次の（３）式によって、打鍵時間差Ｔを算出する。
【００６０】
Ｔ＝ＡＴ−ＢＴ ・・・（３）
但し、ＡＴは後に打鍵されたキーの打鍵コードに対応する打鍵時刻、ＢＴは先に打鍵されたキーの打鍵コードに対応する打鍵時刻である。
【００６１】
次のステップ２０４では、打鍵時間差Ｔが所定値Ｓ１未満であるか否かを判定し、所定値Ｓ１未満でない場合（否定判定の場合）はステップ２０６へ移行する。なお、上記所定値Ｓ１は、前述の下限値に相当するものであり、本実施の形態では１００ｍｓを適用している。
【００６２】
ステップ２０６では、打鍵時間差Ｔが所定値Ｓ２より大きいか否かを判定し、大きくない場合（否定判定の場合）はステップ２０８へ移行する。なお、上記所定値Ｓ２は、前述の上限値に相当するものであり、本実施の形態では１０００ｍｓを適用している。
【００６３】
ステップ２０８では、上記ステップ２００で抽出した打鍵コードの組合わせ及び上記ステップ２０２で算出した打鍵時間差Ｔを二連接打鍵事象テーブルとしてハードディスク２０の二連接打鍵事象テーブル領域ＮＡに記憶した後に本二連接打鍵事象導出処理を終了する。
【００６４】
また、上記ステップ２０６において、打鍵時間差Ｔが所定値Ｓ２より大きいと判定された場合（肯定判定された場合）は、上記ステップ２０８の処理を行うことなく、本二連接打鍵事象導出処理を終了する。
【００６５】
一方、上記ステップ２０４において、打鍵時間差Ｔが所定値Ｓ１未満であると判定された場合（肯定判定された場合）にはステップ２１０へ移行し、上記ステップ２００において抽出された打鍵コードの組合わせが同一打鍵コードの組合わせであったか否かを判定し、同一打鍵コードの組合わせでなかった場合（否定判定の場合）は上記ステップ２０８の処理を実行した後に本二連接打鍵事象導出処理を終了し、同一打鍵コードの組合わせであった場合（肯定判定の場合）はステップ２０８の処理を実行することなく本二連接打鍵事象導出処理を終了する。
【００６６】
図８には、本二連接打鍵事象導出処理によって得られる二連接打鍵事象テーブルの一例が模式的に示されている。同図に示す例では、打鍵コードの組合わせとして１６進数表現の打鍵コードの組合わせが、当該打鍵コードの組合わせに対応する打鍵時間差としてミリ秒単位の時間差が各々記憶されている。なお、同図に示す各データは、打鍵事象テーブルの各データが図６に示すものである場合について示している。同図に示すように、例えば、通知された打鍵コードが１６進法で３ｅ、打鍵時刻が計算機内部時刻の３２１３１ｍｓであり、直前の打鍵事象の打鍵コードが１６進で６ｃ、打鍵時刻が３１９３４ｍｓであるときは、打鍵コードの組合わせは３ｅ，６ｃ、打鍵時間差は１９７ｍｓと算出される。
【００６７】
次に、図９を参照して、前述の打鍵事象特徴値算出処理について説明する。なお、図９は、当該打鍵事象特徴値算出処理を実行するために、ＣＰＵ２２によって実行される打鍵事象特徴値算出処理プログラムの流れを示すフローチャートであり、該プログラムも予めＲＯＭ２４の所定領域に記憶されている。また、この打鍵事象特徴値算出処理プログラムも、上記打鍵事象検出処理プログラムが実行された際に随時実行される割込処理プログラムである。
【００６８】
まず、ステップ３００では、ハードディスク２０に記憶されている二連接打鍵事象テーブルを参照して、直前に二連接打鍵事象テーブルに記憶された打鍵コードの組合わせと、該組合わせに対応する打鍵時間差を読み出し、次のステップ３０２では、ハードディスク２０に記憶されている特徴値テーブルにおける上記ステップ３００で読み出した打鍵コードに対応する「出現頻度」領域の値を１だけインクリメントし、更に、次のステップ３０４で同打鍵コードに対応する「打鍵時間差の和」領域の値を、該値に上記ステップ３００で読み出した打鍵時間差を加算したものに置換え、次のステップ３０６で同打鍵コードに対応する「打鍵時間差の自乗の和」領域の値を、該値に上記ステップ３００で読み出した打鍵時間差の自乗値を加算したものに置き換え、その後に本打鍵事象特徴値算出処理を終了する。
【００６９】
図１０（Ｂ）には、本打鍵事象特徴値算出処理によって得られた特徴値テーブルの一例が模式的に示されている。同図に示すように、本打鍵事象特徴値算出処理によって、ユーザによってキーボード１４のキーが連続して打鍵された際に、打鍵されたキーの打鍵コードの組合わせに対応する出現頻度が１ずつインクリメントされると共に、それまでに特徴値テーブルに記憶されていた打鍵時間差の和及び打鍵時間差の自乗の和に、このとき演算された打鍵時間差及び打鍵時間差の自乗値が各々累積される。
【００７０】
従って、特徴値テーブルに記憶されている各値に基づいて、任意の２種類の打鍵コードの組合わせ毎の打鍵時間差の平均値ＡＶ及び標準偏差値ＨＶを、上記（１）式及び（２）式の演算によって得ることができる。
【００７１】
次に、図１１を参照して、前述の個人識別処理について説明する。なお、図１１は、当該個人識別処理を実行するために、ＣＰＵ２２によって実行される個人識別処理プログラムの流れを示すフローチャートであり、該プログラムも予めＲＯＭ２４の所定領域に記憶されている。また、この個人識別処理プログラムも、上記打鍵事象検出処理プログラムが実行された際に随時実行される割込処理プログラムである。また、ここでは、図示は省略するが、認証対象とする複数人分の特徴値テーブルが予めハードディスク２０の特徴値テーブル領域ＴＡに記憶されており、該複数の特徴値テーブルを用いてユーザの識別を行う場合について説明する。
【００７２】
まず、ステップ４００では、前述の二連接打鍵事象導出処理によって直前に二連接打鍵事象テーブル（図８も参照）に記憶された打鍵コードの組合わせ及び打鍵時間差を読み出し、次のステップ４０２では、予め特徴値テーブル領域ＴＡに記憶されている複数人分の特徴値テーブルのうちの何れかの特徴値テーブル（図１０（Ｂ）も参照）から上記ステップ４００で読み出した打鍵コードの組合わせに対応する特徴値（出現頻度、打鍵時間差の和、打鍵時間差の自乗の和）を読み出し、次のステップ４０４では、読み出した特徴値を用いて上記（１）式及び（２）式により平均値ＡＶ及び標準偏差値ＨＶを算出する。
【００７３】
次のステップ４０６では、上記ステップ４００で読み出した打鍵時間差と、上記ステップ４０４で算出した平均値ＡＶ及び標準偏差値ＨＶとに基づいて、以下に示すように、ファジイ推論によってスコアを導出する。
【００７４】
打鍵コード組合わせＣ_ij（０≦ｉ，ｊ≦ｎ）に対するファジイルールＲ_Cijは後件部をシングルトンとして次の（４）式に示すものとなる。
【００７５】
【数２】

【００７６】
但し、ｎは打鍵コードの種類の数、Ａ_Cijは打鍵コード組合わせＣ_ijに対するファジイラベル、ｘ_Cijは打鍵コード組合わせＣ_ijに対する識別対象個人の打鍵時間差、ｙ_Cijは打鍵コード組合わせＣ_ijに対する識別対象個人のスコアである。
【００７７】
打鍵コード組合わせＣ_ijに関する適合度ω_Cijは、次の（５）式によって得られる。
【００７８】
【数３】

【００７９】
但し、μ_ACijは打鍵コード組合わせＣ_ijに関するメンバーシップ関数、ｘ⁰ _Cijは打鍵コード組合わせＣ_ijに対する識別対象個人の打鍵時間差の非ファジイ値である。
【００８０】
ここで、メンバーシップ関数μ_ACijは上記ステップ４０４によって算出された平均値ＡＶ及び標準偏差値ＨＶに基づいて、頂点が平均値ＡＶとなり、かつ半値幅が標準偏差値ＨＶとなる三角型として設定される。
【００８１】
図１２には、メンバーシップ関数μ_ACijの一例が示されている。なお、同図の左端に示されている‘０８−０８’や‘０８−０９’等の数値は、組合わせられた打鍵コード（１６進数）を示している。
【００８２】
本実施の形態では、打鍵事象発生毎に適合度が計算されるが、単発事象では個人の特徴値を十分に検出しているとは限らないから、通常、スコアは次の（６）式のように複数の打鍵事象を統合して計算される。
【００８３】
【数４】

【００８４】
ここで、ｋは統合する事象の数、ｙ⁰は統合されたスコア（非ファジイ値）である。
【００８５】
なお、（６）式ではスコアの値として０〜１の範囲の値をとるが、該値を１００倍して１００点満点としてもよい。
【００８６】
ファジイ推論が終了すると、次のステップ４０８では、特徴値テーブルが登録された全ての登録人について上記ステップ４０６によるファジイ推論が終了したか否かを判定し、終了していない場合（否定判定の場合）は上記ステップ４０２へ戻って、ファジイ推論が未終了の登録人に対する特徴値テーブルを用いたファジイ推論を再び行い、全ての登録人についてファジイ推論が終了した時点（ステップ４０８が肯定判定となった時点）でステップ４１０へ移行する。
【００８７】
以上のステップ４０２〜ステップ４０８の繰り返し処理によって、全ての登録人に対するファジイ推論によるスコアが導出されるので、ステップ４１０では、スコアが最高値である登録人がユーザである可能性が高いものとしてユーザ候補に選定し、次のステップ４１２では、上記ステップ４１０で選定されたユーザ候補のスコアが所定値（本実施の形態では、０．３）以上であるか否かを判定し、所定値以上でない場合（否定判定の場合）はユーザ候補が誤っており、認証できないものと見なしてステップ４１４へ移行し、利用範囲制限処理を実行した後に本個人識別処理を終了する。なお、本実施の形態における利用範囲制限処理では、前述のように、ユーザによって利用できる機能、リソース、及び機器を制限するための処理が行われる。
【００８８】
一方、上記ステップ４１２において、ユーザ候補のスコアが所定値以上であると判定された場合（肯定判定された場合）には、上記ステップ４１４の処理を行うことなく本個人識別処理を終了する。
【００８９】
なお、本個人識別処理において、打鍵時間差の平均値ＡＶ及び標準偏差値ＨＶを利用するのは、次の理由によるものである。
【００９０】
図１３には、５名の被験者Ａ〜Ｅによる実験によって得られた、打鍵時間差が１０００ｍｓ以上である場合と、打鍵時間差が１００ｍｓ以下でかつ同一キーを連続して打鍵した場合とを除いた全ての有効打鍵時間差のヒストグラムが、図１４には、５名の被験者Ａ〜Ｅによる実験によって得られた、［Ａ］キーと［Ｉ］キーが連続して打鍵された場合の打鍵時間差のヒストグラムが、各々示されている。
【００９１】
図１３及び図１４に示すように、各条件における打鍵時間差の頻度分布は、正規分布に近い状態となっている。従って、打鍵時間差の平均値及び標準偏差値を特徴値として適用することによって、精度のよい個人識別ができるものと推察することができるのである。
【００９２】
〔実験結果例〕
図１５には、被験者Ａがユーザである場合の、本実施の形態に係る個人識別処理（図１１参照）におけるファジイ推論によって得られたスコアの時系列変化の実験結果例が示されている。なお、同図の横軸は有効データ出現頻度（打鍵時間差が１０００ｍｓ以上である場合と、打鍵時間差が１００ｍｓ以下でかつ同一キーを連続して打鍵した場合とを除いた打鍵時間差の出現頻度）であり、縦軸はスコアである。また、ここでは、特徴値としてファジイ推論実行時から遡って５０００回分の打鍵時間差の移動平均を適用している。
【００９３】
同図に示すように、どの時点でも被験者Ａに対するスコアが最も高く、１００％識別できることが分かる。
【００９４】
同様に、図１６には、５名の被験者Ａ〜Ｅがユーザである場合の、本実施の形態に係るファジイ推論によって得られたスコアの時系列変化の実験結果例が示されている。なお、同図も図１５と同様に、横軸は有効データ出現頻度であり、縦軸はスコアである。また、ここでも、特徴値としてファジイ推論実行時から遡って５０００回分の打鍵時間差の移動平均を適用している。
【００９５】
同図に示すように、５名の被験者Ａ〜Ｅの全員について、どの時点でも１００％識別できることが分かる。
【００９６】
図１７には、５名の被験者Ａ〜Ｅがユーザである場合で、特徴値として適用する打鍵時間差の平均値の算出の際にファジイ推論実行時から遡る回数を各々５回、５０回、５００回、及び２０００回とした場合の、各被験者の識別率の実験結果例が示されている。なお、ここでは、各々８００００件分の識別に対応する識別率が示されている。
【００９７】
図１７（Ａ）に示すように、移動平均が５回である場合には、被験者に対応する識別率が他の被験者の識別率よりは高いものの、被験者Ｄを除く全ての被験者の識別率が５０％未満となっており、識別精度が必ずしも良くないことが分かる。しかしながら、図１７（Ｂ）に示す、移動平均が５０回である場合には、全ての被験者について識別率が７０％を越えており、図１７（Ｃ）に示す、移動平均が５００回である場合には、全ての被験者について識別率が９５％を越えており、更に図１７（Ｄ）に示す、移動平均が２０００回である場合には、全ての被験者について識別率が１００％となっていることが分かる。
【００９８】
従って、この場合は、移動平均の回数が５０回以上であれば、かなり高精度に識別することができると考えられる。
【００９９】
一方、図１８には、デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のノートブック型パーソナル・コンピュータに用いて、被験者Ａを識別した場合の有効データ出現回数に対するスコア例が、図１９には、デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のデスクトップ型パーソナル・コンピュータに用いて、被験者Ｂを識別した場合の有効データ出現回数に対するスコア例が、更に図２０には、デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のデスクトップ型パーソナル・コンピュータに用いて、被験者Ｃを識別した場合の有効データ出現回数に対するスコア例が、各々示されている。なお、図１８〜図２０では、特徴値としてファジイ推論実行時から遡って５０００回分の打鍵時間差の移動平均を適用している。
【０１００】
図１８〜図２０に示すように、どの時点においても各被験者を１００％識別することが可能であることが分かる。すなわち、本実施の形態で示した特徴値取得処理で取得した特徴値は、取得に用いたキーボード以外のキーボードによる識別においても適用することが可能である。
【０１０１】
以上詳細に説明したように、本実施の形態に係る個人識別装置では、個人がキーボードに設けられたキーを打鍵したときの打鍵時刻と当該キーを特定できる特定情報（打鍵コード）とが含まれた打鍵事象を検出し、検出した打鍵事象に基づいて、上記個人が連続して打鍵したときの隣接する２つの打鍵事象の打鍵時間差と当該２つの打鍵事象の上記特定情報によって特定されるキーの組合わせを導出し、導出したキーの組合わせのうち、同一の組合わせ毎に上記打鍵時間差に基づく特徴値を算出し、算出した特徴値を記憶すると共に、該記憶した特徴値を利用して個人識別を行っているので、偽造が困難であり、利用者の心理的な抵抗感がなく、利用時の身体の拘束条件が殆んどなく、融通性が高く、更に、入れ替わりによる成り済ましを防止することができる。
【０１０２】
また、本実施の形態に係る個人識別装置では、特徴値に基づいて作成したメンバーシップ関数を用いたファジイ推論によって個人を識別しているので、識別率を向上することができる。
【０１０３】
また、本実施の形態に係る個人識別装置では、識別結果の評価値（スコア）が所定値以下である場合にユーザの利用範囲を制限しているので、セキュリティを向上することができる。
【０１０４】
更に、本実施の形態に係る個人識別装置では、フック関数によってキーボードのキーを離す事象に基づいて打鍵時刻を検出しているので、打鍵を確実に検出することができるため、高精度な特徴値を得ることができ、この結果として高精度に個人を識別することができる。
【０１０５】
なお、本実施の形態では、特徴値取得処理及び識別処理（図４参照）を、ユーザによってキーボード１４のキーが打鍵された際に逐次割込み処理として実行する場合について説明したが、本発明はこれに限定されるものではなく、例えば、特徴値取得処理における打鍵事象検出処理４０及び識別処理は本実施の形態と同様に逐次実行するが、二連接打鍵事象導出処理４２及び打鍵事象特徴値算出処理４４は打鍵事象テーブル内の打鍵事象が所定数以上蓄積された時点や、所定期間（例えば、１日）毎に実行する形態とすることもできる。この場合は、二連接打鍵事象導出処理４２及び打鍵事象特徴値算出処理４４が間欠的に実行されるので、これらの処理が次に実行されるまでの間は当該処理が前回実行された時点の特徴値データに基づいて個人識別が行われるため、本実施の形態に比較して識別率は若干低くなる場合もあるが、二連接打鍵事象導出処理４２及び打鍵事象特徴値算出処理４４のための処理時間を短縮することができる。
【０１０６】
また、本実施の形態では、請求項１記載の発明における打鍵事象検出手段及び二連接打鍵事象導出手段と、請求項２記載の発明における打鍵事象検出手段及び二連接打鍵事象導出手段とを、打鍵事象検出処理４０及び二連接打鍵事象導出処理４２によって兼用した場合について説明したが、本発明はこれに限定されるものではなく、請求項１記載の発明における打鍵事象検出手段及び二連接打鍵事象導出手段と、請求項２記載の発明における打鍵事象検出手段及び二連接打鍵事象導出手段とを、異なる処理によって別々に実行する形態とすることもできる。この場合も、本実施の形態と同様の効果を奏することができる。
【０１０７】
また、本実施の形態では、ファジイ推論によってユーザを識別する場合について説明したが、本発明はこれに限定されるものではなく、例えば、ファジイ・ニューラルネットワークを適用して、ファジイ推論に用いるメンバーシップ関数等を自律学習する形態とすることもできる。
【０１０８】
【発明の効果】
以上詳細に説明したように、請求項１記載の個人識別装置によれば、個人がキーボードに設けられたキーを打鍵したときの打鍵時刻と当該キーを特定できる特定情報とが含まれた打鍵事象を検出し、検出した打鍵事象に基づいて、上記個人が連続して任意に打鍵したときの隣接する２つの打鍵事象の打鍵時間差と当該２つの打鍵事象の上記特定情報によって特定されるキーの組合わせを導出し、導出したキーの組合わせのうち、同一の組合わせ毎に上記打鍵時間差に基づく特徴値を算出し、算出した特徴値を逐次記憶しているので、記憶した特徴値を利用して個人識別を行うことによって、偽造が困難であり、利用者の心理的な抵抗感がなく、利用時の身体の拘束条件が殆んどなく、融通性が高く、更に、入れ替わりによる成り済ましを防止することができる個人識別が可能となる、という効果が得られる。
【０１０９】
また、請求項２記載の個人識別装置によれば、個人がキーボードに設けられたキーを連続して任意に打鍵したときの打鍵時刻が隣接する２つのキーの同一の組合わせ毎に、当該２つのキーの打鍵時間差に基づく特徴値が逐次記憶される記憶手段が備えられると共に、個人がキーボードに設けられたキーを打鍵したときの打鍵時刻と当該キーを特定できる特定情報とが含まれた打鍵事象を検出し、検出した打鍵事象に基づいて、上記個人が連続して任意に打鍵したときの隣接する２つの打鍵事象の打鍵時間差と当該２つの打鍵事象の上記特定情報によって特定されるキーの組合わせを導出し、導出したキーの組合わせに対応する上記記憶手段に逐次記憶されている特徴値と、上記導出した打鍵時間差とを比較し、該比較結果に基づいて個人を識別しているので、偽造が困難であり、利用者の心理的な抵抗感がなく、利用時の身体の拘束条件が殆んどなく、融通性が高く、更に、入れ替わりによる成り済ましを防止することができる、という効果が得られる。
【図面の簡単な説明】
【図１】 実施の形態に係る個人識別装置１０の外観を示す斜視図である。
【図２】 実施の形態に係る個人識別装置１０の概略構成を示すブロック図である。
【図３】 実施の形態に係る個人識別装置１０に内蔵されたハードディスクの各種記憶領域を示す模式図である。
【図４】 実施の形態に係る個人識別装置１０によって実行される各種処理の位置付けを示すブロック図である。
【図５】 実施の形態に係る打鍵事象検出処理プログラムの流れを示すフローチャートである。
【図６】 打鍵事象テーブルの一例を示す模式図である。
【図７】 実施の形態に係る二連接打鍵事象導出処理プログラムの流れを示すフローチャートである。
【図８】 二連接打鍵事象テーブルの一例を示す模式図である。
【図９】 実施の形態に係る打鍵事象特徴値算出処理プログラムの流れを示すフローチャートである。
【図１０】 特徴値テーブルの一例を示す模式図であり、（Ａ）は初期状態の特徴値テーブルを、（Ｂ）は各種データが記憶された後の特徴値テーブルを、各々示す模式図である。
【図１１】 実施の形態に係る個人識別処理プログラムの流れを示すフローチャートである。
【図１２】 本実施の形態におけるメンバーシップ関数の例を示す線図である。
【図１３】 ５名の被験者Ａ〜Ｅによる実験によって得られた、打鍵時間差が１０００ｍｓ以上である場合と、打鍵時間差が１００ｍｓ以下でかつ同一キーを連続して打鍵した場合とを除いた全ての有効打鍵時間差のヒストグラムを示す線図である。
【図１４】 ５名の被験者Ａ〜Ｅによる実験によって得られた、［Ａ］キーと［Ｉ］キーが連続して打鍵された場合の有効打鍵時間差のヒストグラムを示す線図である。
【図１５】 被験者Ａがユーザである場合の、本実施の形態に係る個人識別処理におけるファジイ推論によって得られたスコアの時系列変化の実験結果例を示す線図である。
【図１６】 ５名の被験者Ａ〜Ｅがユーザである場合の、本実施の形態に係るファジイ推論によって得られたスコアの時系列変化の実験結果例を示す線図である。
【図１７】 ５名の被験者Ａ〜Ｅがユーザである場合で、特徴値として適用する打鍵時間差の平均値の算出の際にファジイ推論実行時から遡る回数を各々５回、５０回、５００回、及び２０００回とした場合の、各被験者の識別率の実験結果例を示す棒グラフである。
【図１８】 デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のノートブック型パーソナル・コンピュータに用いて、被験者Ａを識別した場合の有効データ出現回数に対するスコア例を示す線図である。
【図１９】 デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のデスクトップ型パーソナル・コンピュータに用いて、被験者Ｂを識別した場合の有効データ出現回数に対するスコア例を示す線図である。
【図２０】 デスクトップ型パーソナル・コンピュータのキーボードに対する被験者Ａ〜Ｅの操作によって得られた特徴値を他のデスクトップ型パーソナル・コンピュータに用いて、被験者Ｃを識別した場合の有効データ出現回数に対するスコア例を示す線図である。
【符号の説明】
１０ 個人識別装置
１４ キーボード
２０ ハードディスク（記憶手段）
２２ ＣＰＵ
４０ 打鍵事象検出処理（打鍵事象検出手段）
４２ 二連接打鍵事象導出処理（二連接打鍵事象導出手段）
４４ 打鍵事象特徴値算出処理（打鍵事象特徴値算出手段）
５０ 個人識別処置（識別手段）
５２ 利用範囲制限処理（制限手段）[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a personal identification device, and more particularly to a personal identification device that identifies an individual based on an action of the individual operating a keyboard.
[0002]
[Prior art]
Conventionally, as a method for identifying personal identification devices used for entry / exit management in restricted areas or computer access management, etc., a method using human body features such as fingerprints, irises, retinal blood vessels, handprints, faces, etc. There are known methods that use characteristics such as voiceprints and signatures.
[0003]
Among these, the identification method using the fingerprint pattern disclosed in JP-A-6-28457, JP-A-6-28458, JP-A-6-28459, etc. is not universal and lifelong. It is widely known as a technique that can achieve a higher identification rate than other techniques by using invariant physical features.
[0004]
On the other hand, as a technique different from these techniques, Japanese Patent Application Laid-Open No. Sho 62-157966 discloses a technique for confirming an individual by using a temporal characteristic of typing when a predetermined character string is typed. Is described.
[0005]
[Problems to be solved by the invention]
However, in the method using the two-dimensional pattern obtained by optically detecting the fingerprint as described above, a high identification rate can be realized as described above, but the fingerprint pattern of another person and the fingerprint pattern are collected. However, when a high identification rate is realized, there is a problem that it is vulnerable to forgery using the replica.
[0006]
In addition, this method has a problem that it is necessary to touch an optical member such as a prism for collecting a fingerprint pattern, which is unsanitary when used by an unspecified number of people.
[0007]
In order to avoid these problems, even if a three-dimensional pattern that adds non-contact ridges that are convex portions of fingerprints and valley shape information that is concave portions is detected and used, the information on the appearance is clogged up. By only adding information for one dimension, it is only possible to obtain an effect that makes counterfeiting somewhat cumbersome, the fingerprint pattern detection device becomes complicated, the amount of data processing increases, the device cost and processing time However, the problem cannot be solved fundamentally. That is, whether it is a planar two-dimensional pattern or a three-dimensional three-dimensional pattern, it is vulnerable to forgery as long as the appearance information is used.
[0008]
Furthermore, since fingerprints have an image of criminal investigation, users have a high level of psychological resistance. To disseminate them in general, there is a difficult problem before the technology of wiping images.
[0009]
The above-described conventional methods other than the method using the appearance of the fingerprint as described above have problems as described below.
[0010]
In the technique using a bill, since the appearance is used in the same manner as the two-dimensional pattern of the fingerprint, there is a high risk of being duplicated. In addition, the method using iris reads the iris pattern with the camera, so it is possible to wiretap and record the video signal from the camera, and connect the recorded video signal to the camera's video cable at the time of authentication. There is a great risk of being counterfeited. In the method using the shape of the face, it is necessary to keep the posture and illumination conditions within a certain range, which is not stable and may be forged to use the appearance.
[0011]
Voiceprints and signatures are likely to be stolen, and are easily affected by the psychological state and health of the user and lack stability. In addition, the method using the blood vessel pattern of the hand or finger is more difficult to counterfeit than the method using the external feature, and the psychological burden on the user is smaller than the method using the retinal blood vessel pattern. However, in this method, near infrared light or the like is used, and information that is relatively close to the surface layer, such as a subcutaneous vein, is used, so that there is a possibility that a pattern is eventually stolen by a near infrared camera or the like. Furthermore, since the blood vessels of the hands and fingers are deformed by the movement of the hands and fingers, it is difficult to obtain the same pattern every day. For this reason, there is a problem that the allowable range at the time of collation must be wide, and the false authentication rate becomes high.
[0012]
In addition, the most important basic problem of these conventional techniques is that personal identification is performed only at one point on the time axis, and it is nearly powerless for subsequent impersonation.
[0013]
In the first place, the main purpose of these personal identification methods is to control the physical passage into a closed space, such as entering or leaving a building or room. It is assumed that they are isolated from the outside and cannot be replaced regardless of the passage of time. Therefore, even if these personal identification methods are applied as a personal authentication method when using a computer or a computer terminal from a remote place through a network without spatial restrictions, impersonation due to the replacement of the operating person cannot be detected. I have to say that this is obviously an inappropriate technique. Similarly, it is not possible to detect an imperfection that has entered from the middle of a network route. Therefore, no matter how strict personal authentication is performed using these personal identification methods at the start of access, security is not perfect.
[0014]
On the other hand, since the technique described in the above Japanese Patent Application Laid-Open No. 62-157966 uses the temporal characteristics of typing as described above to identify individuals, the risk of forgery is extremely low. Since the individual cannot be confirmed unless the given character string is input, there is a problem that the user is forced to input the character string and the flexibility is low. In order to solve this problem, as described in the publication, a character string that is always typed when starting the device, such as “Log in”, is applied as the predetermined character string. It is possible for the person to confirm the individual without being aware of it, but in this case, since the individual confirmation is performed only when the device is activated, the above-mentioned problem of impersonation occurs.
[0015]
The present invention has been made to solve the above-mentioned problems, is extremely difficult to counterfeit, has no psychological resistance to the user, has few physical restraint conditions during use, and is flexible. Another object of the present invention is to provide a personal identification device that has high performance and can prevent impersonation due to replacement.
[0016]
[Means for Solving the Problems]
  In order to achieve the above object, the personal identification device according to claim 1 is a personal identification device for identifying the individual based on an action of the individual operating the keyboard, wherein the individual has a key provided on the keyboard. A key-pressing event detecting means for detecting a key-pressing event including a key-pressing time when the key is pressed and specific information for specifying the key, and the individual is continuously connected based on the key-pressing event detected by the key-pressing event detecting means. do itArbitrarilyA double-connected keying event derivation unit for deriving a key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events; Of the derived key combinations, the keystroke event feature value calculating means for calculating a feature value based on the keystroke time difference for each same combination, and the feature value calculated by the keystroke event feature value calculating meansSequentialStorage means for storing.
[0017]
  According to the personal identification device of claim 1, the key-pressing event detecting means generates a key-pressing event including a key-pressing time when a key is pressed on a key provided on the keyboard and specific information for specifying the key. Based on the detected keystroke event, the individual continuously detectsArbitrarilyThe key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events is derived by the double-keying key event deriving means. Here, the keyboard includes a keyboard provided in a computer such as a personal computer or a workstation, a key switch provided in a communication device such as a mobile phone, a wired telephone, and a facsimile, or a television receiver. All devices provided with a plurality of keys are included, such as a key switch part of a remote control device widely used in various devices such as a video machine, a video device, and an air conditioner.
[0018]
  In the present invention, the keystroke event feature value calculation means calculates a feature value based on the keystroke time difference for each identical combination out of the key combinations derived by the double-connected keystroke event derivation means. By the storage meansSequentialRemembered. The storage means includes a RAM (Random Access Memory), an EEPROM (Electrically Erasable and Programmable Read Only Memory), a flash memory (Flash EEPROM), and a floppy disk.(Registered trademark)・ Installed on portable recording media such as discs, CD-R (Compact Disc-Recordable), CD-RW (Compact Disc-ReWritable), and MO (magneto-optical) discs, or server computers connected to a network. An external storage device or the like can be used.
[0019]
That is, according to the present invention, a feature value based on a keying time difference between keystroke keys connected in time is sequentially stored while an individual is inputting an arbitrary character string using a keyboard. Therefore, the feature value of the present invention is obtained as a value based on the keying time difference between arbitrary keys. Therefore, when identifying an individual, combinations of keys corresponding to the feature values accumulated so far are consecutive. If the key is pressed, the individual can be identified by comparing the keying time difference obtained by the key pressing and the corresponding feature value. Therefore, the personal identification to which the present invention is applied is extremely flexible because it is not necessary to input a predetermined character string. Also, this identification is difficult to counterfeit, there is little psychological resistance of the user, and there are almost no restrictions on the body during use.
[0020]
Furthermore, personal identification in this case can be performed at any time as long as the individual is performing keystrokes. Therefore, impersonation due to replacement can be prevented by performing the personal identification as needed.
[0021]
  Thus, the personal identification device according to claim 1 detects a key-pressing event including a key-pressing time when the individual keys a key provided on the keyboard and specific information that can specify the key. Based on the detected keystroke event, the individualArbitrarilyDeriving a key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the above-mentioned specific information of the two keying events, and for each of the derived combinations, the same combination Calculate the feature value based on the above keying time difference, and calculate the calculated feature valueSequentialSince it is memorized, it is difficult to counterfeit by using the memorized feature values, there is no psychological resistance of the user, and there are almost no body restraints when using it Further, it is possible to identify individuals that have high flexibility and can prevent impersonation due to replacement.
[0022]
  In order to achieve the above object, the personal identification device according to claim 2 is a personal identification device for identifying the individual based on an action of the individual operating the keyboard, wherein the individual is provided on the keyboard. Consecutive keysArbitrarilyThe keying time when the key is pressedSameFor each combination, the feature value based on the keystroke time difference between the two keys isMemorized sequentiallyStorage means; key-pressing event detection means for detecting a key-pressing event including key-pressing time when the individual hits a key provided on the keyboard; and specific information for specifying the key; and key-pressing event detection means Based on the keystroke event detected byArbitrarilyA double-connected keying event derivation unit for deriving a key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events; In the storage means corresponding to the derived key combinationSequentialAnd an identification unit that compares the stored characteristic value with the keying time difference derived by the double-connected keystroke event deriving unit and identifies an individual based on the comparison result.
[0023]
  The personal identification device according to claim 2, wherein the individual continuously presses keys provided on the keyboard.ArbitrarilyThe keying time when the key is pressedSameFor each combination, the feature value based on the keystroke time difference between the two keys isMemorized sequentiallyStorage means is provided. In addition to the keyboard provided in computers such as personal computers and workstations, the keyboard includes a key switch unit provided in communication devices such as mobile phones, wired telephones, facsimiles, and television receivers. Any device provided with a plurality of keys, such as a key switch part of a remote control device widely used in various devices such as video devices and air conditioners, is included. The storage means includes a semiconductor storage element such as a RAM, an EEPROM, and a flash EEPROM, and a floppy disk.(Registered trademark)A portable recording medium such as a disk, CD-R, CD-RW, or MO disk, or an external storage device provided in a server computer connected to a network can be used.
[0024]
  Further, according to the present invention, a key-pressing event detecting unit detects a key-pressing event including a key-pressing time when a person presses a key provided on a keyboard and specific information for specifying the key, and the detection is performed. Based on the keystroke event made,ArbitrarilyThe key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events is derived by the double-keying key event deriving means.
[0025]
  Furthermore, in the present invention, the storage means corresponding to the combination of keys derived by the double-connected keystroke event deriving means by the identifying means.SequentialThe stored feature value is compared with the keying time difference derived by the above-mentioned double-joining keystroke event deriving means, and an individual is identified based on the comparison result.
[0026]
That is, in the present invention, when an individual is inputting an arbitrary character string with the keyboard, the keystroke time difference between the keystroke keys connected in time and the feature value stored in advance in the corresponding storage means The individual is identified by comparison. Accordingly, the personal identification according to the present invention is extremely flexible because it is not necessary to input a predetermined character string. Also, this identification is difficult to counterfeit, there is little psychological resistance of the user, and there are almost no restrictions on the body during use.
[0027]
Furthermore, personal identification according to the present invention can be performed at any time when the individual is performing keystrokes. Therefore, impersonation due to replacement can be prevented by performing the personal identification as needed.
[0028]
  In this way, according to the personal identification device of claim 2, the individual continuously applies the keys provided on the keyboard.ArbitrarilyThe keying time when the key is pressedSameFor each combination, the feature value based on the keystroke time difference between the two keys isMemorized sequentiallyA storage means is provided, and a keying event including a keying time when the individual keys a key provided on the keyboard and specific information for specifying the key is detected, and based on the detected keying event, Individuals continuouslyArbitrarilyA key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events is derived, and the storage means corresponding to the derived key combination is derived.SequentialSince the stored feature value is compared with the derived keying time difference and the individual is identified based on the comparison result, forgery is difficult, there is no psychological resistance of the user, There are almost no restraints on the body at the time, flexibility is high, and impersonation due to replacement can be prevented.
[0029]
As in the invention described in claim 3, the identification means in the invention described in claim 2 can also identify an individual by fuzzy inference using a membership function created based on the feature value. . In this case, since the individual is identified by fuzzy inference, the identification rate can be improved.
[0030]
Further, like the invention according to claim 4, the invention according to claim 2 or claim 3 is a restriction that limits the range of use of the individual when the evaluation value of the identification result by the identification means is not more than a predetermined value. Preferably further means are provided. Thereby, security can be improved. The range of personal use includes the range of devices that can be used, the range of functions that can be used, and the like.
[0031]
By the way, using a hook function of an API (Application Program Interface) provided by Windows OS such as Windows 98 and Windows NT of Microsoft Corporation, which has been widely spread as an operating system (OS) for personal computers in recent years. When detecting keystrokes, it is not possible to distinguish between the keystroke state in which an individual physically pressed the key and the keystroke state generated by the auto-repeat function of the keyboard or keyboard driver, but the state in which the key was released is reliably detected. can do.
[0032]
Therefore, as in the invention according to claim 5, in the invention according to any one of claims 1 to 4, the key-pressing event detecting means uses the event of releasing the key of the keyboard to generate the key-pressing time. Is preferably detected. As a result, the keystroke can be reliably detected by the hook function, so that a highly accurate feature value can be obtained, and as a result, an individual can be identified with high accuracy.
[0033]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. As shown in FIG. 1, a personal identification device 10 according to the present embodiment includes a control unit 12 that controls the operation of the entire device, a keyboard 14 and a mouse 16 that are used to input various information from an operator (user), The display 18 is configured to display various information such as character information and image information.
[0034]
As shown in FIG. 2, the control unit 12 includes a CPU (Central Processing Unit) 22. The CPU 22 includes a hard disk 20 that stores various tables to be described later, and a ROM 24 that stores various programs, parameters, and the like. , And a RAM 26 used as a work area when the CPU 22 executes various programs.
[0035]
In addition, a keyboard 14 and a mouse 16 are connected to the CPU 22 via an interface unit (I / F) 28, and a display 18 is connected to the CPU 22 via a display control unit 30.
[0036]
As described above, the personal identification device 10 according to the present embodiment is composed of a general-purpose personal computer, and by executing a predetermined program by the CPU 22, a user of the personal computer (“individual” of the present invention). It is configured so that it can be identified. The personal identification device 10 is installed with Windows 98 as an OS.
[0037]
FIG. 3 shows various storage areas of the hard disk 20 provided in the personal identification device 10 according to the present embodiment. As shown in the figure, the hard disk 20 includes a key pressing time when a user has pressed a key provided on the keyboard 14 and a key code (hereinafter referred to as a “key pressing code”) that can specify the key. A keystroke event table area DA that is an area for storing a keystroke event table, which will be described later, and a two-keystroke event table area NA, which is an area for storing a keystroke event table, which will be described later, created based on the keystroke event table. In addition, a feature value table area TA, which is an area for storing a later-described feature value table created based on the double-connected keystroke event table, is provided in advance.
[0038]
Note that, as shown in FIG. 10A, the feature value table according to the present embodiment includes an “appearance frequency”, a “sum of keying time differences”, and a “keying time difference” for each combination of all two types of keying codes. Areas for storing “the sum of squares of keystroke time differences” are provided, and 0 (zero) is stored as an initial setting in each area.
[0039]
On the other hand, the process performed for personal identification by the personal identification device 10 according to the present embodiment is a characteristic value acquisition process for acquiring and recording characteristic values for a plurality of users to be identified as shown in FIG. And, it is roughly divided into two processes of an identification process for identifying a user who is using the personal identification device 10 using the feature values stored so far. Further, as shown in the figure, the feature value acquisition process is executed by three processes of a keystroke event detection process 40, a double-connected keystroke event derivation process 42, and a keystroke event characteristic value calculation process 44. The process is executed by two processes of the personal identification process 50 and the use range restriction process 52. Hereinafter, the outline of each process of the keystroke event detection process 40, the double keystroke event derivation process 42, the keystroke event feature value calculation process 44, the personal identification process 50, and the usage range restriction process 52 will be described.
(1) Keystroke event detection process
In the keying event detection process 40, the keying time when the user presses any key provided on the keyboard 14 and the keying code of the key are detected, and the keying event table area DA of the hard disk 20 is detected as a keying event table. To remember. In order to realize this process, a Windows message may be intercepted using an API hook function provided by the Windows OS, or a detection function may be incorporated in the device driver of the keyboard 14.
[0040]
When the hook function is used, all the messages are once captured, only the keyboard operation-related messages are detected from the captured messages, and the keystroke code is acquired from the detected messages. Further, since the detected message is not time stamped, the time when the message is detected is used as the keying time.
[0041]
In addition to using the hook function, a detection mechanism may be provided in the keyboard device driver, or hardware or software may be provided in the keyboard. In this case, the advantage of using the hook function is The advantage of not depending on the keyboard type is lost. For this reason, in the keystroke event detection process 40 according to the present embodiment, a form using a hook function is applied.
[0042]
As described above, in the keystroke event detection process 40 according to the present embodiment, the keystroke time and the keystroke code are detected. In addition to these keystroke events, information indicating the type of keystroke may be detected. . There are two types of keystrokes: pressing the key and releasing the key based on the general physical structure of the keyboard, but if you use a hook function to detect a keystroke event, Keying information that releases a key that can reliably detect an individual's keyboard operation cannot be distinguished from a keying state in which the user physically presses a key and a keying state generated by the auto-repeat function of the keyboard or keyboard driver. It is preferable to limit to only this, and this embodiment is applied in this embodiment. In this case, since there is only one type of keystroke, there is no need to store this information.
(2) About the double keystroke event derivation process
In the two consecutive keystroke event derivation process 42, a combination of keystroke codes between two keystrokes that are consecutively generated keystroke events and a time difference between keystroke times are derived, and a two consecutive keystroke event table of the hard disk 20 is obtained as a two consecutive keystroke event table. Store in area NA.
[0043]
Here, due to chattering of the keyboard 14 or software problems, when the same keying code is continuously derived with a short time difference that is unlikely to occur in reality, there is a relatively long keying time difference due to the user's absence or the like. When derived, it should be considered that the characteristics of the user's individual operation are not reflected. Therefore, it is preferable to remove each data (keystroke time and keystroke code) in these cases. In this embodiment, This form is applied.
[0044]
In the present embodiment, five persons who operate a personal computer on a daily basis are subjects, and from the experimental results when each key is pressed 80,000 times, the lower limit value of the time interval to be removed here is 100 ms to 200 ms. It is considered that the upper limit value of about 500 ms to 1000 ms is appropriate.
[0045]
As for the lower limit value, the recognition rate may decrease drastically below the value shown here, but as the upper limit value was increased, the recognition rate was not affected so much. For example, when the user who is unfamiliar with the above operation is to be identified, there is no problem even if the upper limit is set to a larger value.
[0046]
The lower limit value is applied to the keying time difference when the same keying code is continuous (when the same key is continuously pressed), and the N key rollover function (N key is released at substantially the same time). However, it is considered that there are almost no keyboards for personal computers that do not have a function capable of recognizing each key). Therefore, even if the time difference between consecutive keystroke codes is 0 ms, it is effective data. Must be treated as Furthermore, when the present invention is applied to a device having a relatively small keyboard, such as a portable computer or a mobile phone, it is difficult to say that the operability of the keyboard is superior to the keyboard for a personal computer. It is preferable to apply a larger upper limit value.
(3) Keystroke event feature value calculation process
In the keying event feature value calculation process 44, a statistical value of a keying time difference is calculated for each combination of keying codes derived by the double-keying key event derivation process 42, and is stored in the feature value table area TA of the hard disk 20 as a feature value table. To do. The statistical values stored in the keying event feature value calculation process 44 according to the present embodiment are of three types: the appearance frequency for each combination of keying codes, the sum of keying time differences, and the sum of squares of keying time differences.
[0047]
In order to realize this processing by a program, an array or a tree structure with a combination of keystroke codes as an index is generated in advance or dynamically, the appearance frequency indicated by the index is incremented, and the index is indicated by the index. The keying time difference is added to the sum of the keying time differences, and the square value of the keying time difference is added to the sum of the squares of the keying time differences indicated by the index, and each value obtained as described above is stored.
(4) About personal identification processing
In the personal identification procedure 50, as in the keystroke event detection process 40 described above, the keystroke time when the user strikes any key provided on the keyboard 14 and the keystroke code of the key are detected, and the two consecutive connections are detected. Similar to the keying event derivation process 42, a keying code combination and keying time difference between two keying keys that are successively generated keying events are derived.
[0048]
Next, the personal identification procedure 50 uses the characteristic values (in this embodiment, the sum of the appearance frequency, the total keying time difference and the total squared value of the keying time difference) corresponding to the combination of the derived keying codes as the keying event characteristics described above. The average value AV and the standard deviation value HV are obtained from the feature value table stored by the value calculation process 44 and obtained by the following equations (1) and (2).
[0049]
[Expression 1]

[0050]
However, ST is the sum of keystroke time differences, JT is the sum of squared keystroke time differences, and CT is the appearance frequency.
[0051]
Then, the individual identification procedure 50 calculates a score by fuzzy inference based on the average value AV and the standard deviation value HV and the keystroke time difference obtained from the key operation of the user to be identified.
[0052]
In the personal identification device 10 according to the present embodiment, feature values of a plurality of persons are registered as a feature value table by the keystroke event feature value calculation process 44 described above, and a score for each of the registered feature values of the plurality of persons is stored. As described above, the registrant of the feature value having the highest score is identified as the user.
[0053]
If there is only one registrant, the authentication is not verification but verification. In this case, determination is made by providing a threshold value. Even in the case of collation or identification, it is preferable that the authentication is not performed if the score is equal to or less than a certain value, and this embodiment is applied in this embodiment.
(5) Usage range restriction processing
In the use range restriction process 52, when a user is not authenticated in the personal identification procedure 50, functions, resources, and devices that can be used by the user are restricted. Thereby, security can be improved.
[0054]
The keystroke event detection process 40 is the keystroke event detection means of the present invention, the keystroke event derivation process 42 is the keystroke event characteristic value calculation process 44 of the present invention. The hard disk 20 corresponds to the storage means of the present invention, the personal identification process 50 corresponds to the identification means of the present invention, and the use range restriction process 52 corresponds to the restriction means of the present invention.
[0055]
Next, the operation of the personal identification device 10 will be described. First, the keystroke event detection process described above will be described with reference to FIG. FIG. 5 is a flowchart showing the flow of a keying event detection processing program executed by the CPU 22 to execute the keying event detection processing. The program is stored in a predetermined area of the ROM 24 in advance. This keystroke event detection processing program is an interrupt processing program that is executed whenever a message flows through the hook function.
[0056]
First, in step 100, the message sent by the hook function is notified as it is to the original notification destination, and in the next step 102, it is determined whether or not the message is a keyboard operation related message. If it is a message (in the case of an affirmative determination), the process proceeds to step 104, where the keying time and keying code are stored in the keying event table area DA of the hard disk 20 as a keying event table, and then the keying event detection process is terminated. If the message is not an operation-related message (when a negative determination is made at step 102), the keystroke event detection process is terminated without executing the process at step 104.
[0057]
FIG. 6 schematically shows an example of a keystroke event table obtained by the keystroke event detection process. In the example shown in the figure, a time in milliseconds is stored as a keying time, and a hexadecimal code (for example, “0x6c” represents a hexadecimal number 6c) is stored. Note that, as described above, the keystroke event detection process is executed whenever a message flows through the hook function, so that each time the user presses a key on the keyboard 14, as shown in FIG. The keying time and keying code are accumulated.
[0058]
Next, with reference to FIG. 7, the above-described double-connected keystroke event derivation process will be described. FIG. 7 is a flowchart showing the flow of a double-connected keystroke event derivation processing program executed by the CPU 22 in order to execute the double-connected keystroke event derivation processing. This program is also stored in a predetermined area of the ROM 24 in advance. ing. The double-connected keystroke event derivation processing program is an interrupt processing program that is executed as needed when the keystroke event detection processing program is executed.
[0059]
First, in step 200, the keystroke event table stored in the hard disk 20 is referred to extract a keystroke code combination between keystrokes of connected keystroke events. In the next step 202, the keystroke code combination is extracted in step 200 above. The keying time of each keying code in the combination of keying codes thus read is read from the keying event table, and the keying time difference T is calculated by the following equation (3).
[0060]
T = AT-BT (3)
Here, AT is a keying time corresponding to the keying code of the key that was pressed later, and BT is a keying time corresponding to the keying code of the key that was previously pressed.
[0061]
In the next step 204, it is determined whether or not the key pressing time difference T is less than the predetermined value S1, and if it is not less than the predetermined value S1 (in the case of negative determination), the process proceeds to step 206. The predetermined value S1 corresponds to the aforementioned lower limit value, and 100 ms is applied in the present embodiment.
[0062]
In step 206, it is determined whether or not the keying time difference T is larger than the predetermined value S <b> 2, and if not larger (in the case of negative determination), the process proceeds to step 208. The predetermined value S2 corresponds to the above-described upper limit value, and 1000 ms is applied in the present embodiment.
[0063]
In step 208, the combination of the keystroke codes extracted in step 200 and the keystroke time difference T calculated in step 202 are stored in the double keystroke event table area NA of the hard disk 20 as a double keystroke event table, and then the double keystroke key. The event derivation process ends.
[0064]
If it is determined in step 206 that the keying time difference T is greater than the predetermined value S2 (when an affirmative determination is made), the process of step 208 is not performed, and the two consecutive keystroke event derivation process is terminated. .
[0065]
On the other hand, if it is determined in step 204 that the keying time difference T is less than the predetermined value S1 (if affirmative determination is made), the process proceeds to step 210, and the combination of the keying codes extracted in step 200 is determined. It is determined whether or not the key combination code is the same combination. If the combination is not the same key combination code (in the case of negative determination), the process of step 208 is executed and then the two consecutive key input event derivation process is terminated. If it is a combination of the same keying codes (in the case of an affirmative determination), the two-connected keystroke event derivation process is terminated without executing the process of step 208.
[0066]
FIG. 8 schematically shows an example of a double-connected keystroke event table obtained by the double-connected keystroke event derivation process. In the example shown in the drawing, a combination of keying codes expressed in hexadecimal notation is stored as a keying code combination, and a time difference in milliseconds is stored as a keying time difference corresponding to the keying code combination. Each data shown in the figure shows a case where each data of the keystroke event table is shown in FIG. As shown in the figure, for example, the notified keying code is 3e in hexadecimal notation, the keying time is 32131 ms of the computer internal time, the keying code of the previous keying event is 6c in hexadecimal, and the keying time is 31934 ms. In some cases, the keying code combination is calculated as 3e, 6c, and the keying time difference is calculated as 197 ms.
[0067]
Next, the keystroke event feature value calculation process described above will be described with reference to FIG. FIG. 9 is a flowchart showing the flow of a keying event feature value calculation processing program executed by the CPU 22 in order to execute the keying event feature value calculation processing. This program is also stored in a predetermined area of the ROM 24 in advance. ing. The keystroke event feature value calculation processing program is also an interrupt processing program that is executed as needed when the keystroke event detection processing program is executed.
[0068]
First, in step 300, referring to the two consecutive keystroke event table stored in the hard disk 20, the combination of the keystroke code stored in the last two consecutive keystroke event table and the keystroke time difference corresponding to the combination are calculated. In the next step 302, the value of the “appearance frequency” area corresponding to the keystroke code read in step 300 in the feature value table stored in the hard disk 20 is incremented by 1, and then in the next step 304. The value in the “sum of keying time differences” area corresponding to the keying code is replaced with the value obtained by adding the keying time difference read in step 300 to the keying code. The value in the “Squared sum” area is added to the square value of the keystroke time difference read in step 300 above. Replacement for, and ends the keying events characteristic value calculation processing thereafter.
[0069]
FIG. 10B schematically shows an example of a feature value table obtained by the keystroke event feature value calculation process. As shown in the figure, when the key on the keyboard 14 is continuously pressed by the user by the key-pressing event feature value calculation process, the appearance frequency corresponding to the key-pressing code combination of the key pressed is one by one. At the same time, the keying time difference and the squared value of the keying time difference calculated at this time are respectively accumulated in the sum of the keying time differences and the square of the keying time differences stored in the feature value table.
[0070]
Therefore, based on each value stored in the feature value table, the average value AV and the standard deviation value HV of the keying time difference for each combination of any two types of keying codes are expressed by the above equations (1) and (2). It can be obtained by calculating the expression.
[0071]
Next, the above-described personal identification process will be described with reference to FIG. FIG. 11 is a flowchart showing the flow of a personal identification processing program executed by the CPU 22 in order to execute the personal identification processing. This program is also stored in a predetermined area of the ROM 24 in advance. This personal identification processing program is also an interrupt processing program that is executed as needed when the key-stroke event detection processing program is executed. Although not shown here, feature value tables for a plurality of persons to be authenticated are stored in advance in the feature value table area TA of the hard disk 20, and user identification is performed using the plurality of feature value tables. The case of performing will be described.
[0072]
First, in step 400, the combination of keystroke codes and the keystroke time difference stored in the double keystroke event table (see also FIG. 8) immediately before by the above-described double keystroke event derivation process are read. Corresponds to the combination of keystroke codes read in step 400 from any one of the feature value tables stored in the feature value table area TA (see also FIG. 10B). The feature values (appearance frequency, sum of keying time differences, sum of squares of keying time differences) are read out, and in the next step 404, the average value AV and standard are calculated by the above formulas (1) and (2) using the read feature values. A deviation value HV is calculated.
[0073]
In the next step 406, a score is derived by fuzzy inference as shown below based on the keystroke time difference read in step 400 and the average value AV and standard deviation value HV calculated in step 404.
[0074]
Keystroke code combination C_ijFuzzy rule R for (0 ≦ i, j ≦ n)_CijIs shown in the following formula (4) with the consequent part as a singleton.
[0075]
[Expression 2]

[0076]
Where n is the number of keystroke code types and A_CijIs key combination code C_ijFuzzy label for x_CijIs key combination code C_ijDifference in keystroke time of individual to be identified, y_CijIs key combination code C_ijThis is the score of the individual to be identified.
[0077]
Keystroke code combination C_ijGoodness of fit ω_CijIs obtained by the following equation (5).
[0078]
[Equation 3]

[0079]
However, μ_ACijIs key combination code C_ijMembership function for x⁰ _CijIs key combination code C_ijThis is a non-fuzzy value of the keying time difference of the identification target individual with respect to.
[0080]
Where the membership function μ_ACijIs set based on the average value AV and the standard deviation value HV calculated in the above step 404 as a triangular shape with the vertex being the average value AV and the half-value width being the standard deviation value HV.
[0081]
FIG. 12 shows the membership function μ_ACijAn example is shown. Note that numerical values such as '08 -08 'and '08 -09' shown at the left end of the figure indicate a combined keystroke code (hexadecimal number).
[0082]
In the present embodiment, the fitness is calculated every time a keystroke event occurs. However, since a single event does not always sufficiently detect an individual feature value, the score is usually expressed by the following equation (6). Thus, a plurality of keystroke events are integrated and calculated.
[0083]
[Expression 4]

[0084]
Where k is the number of events to be integrated and y⁰Is an integrated score (non-fuzzy value).
[0085]
In addition, although the value of the range of 0-1 is taken as a value of a score in (6) Formula, this value may be multiplied by 100 and it may be 100 points of perfect scores.
[0086]
When the fuzzy inference is completed, in the next step 408, it is determined whether or not the fuzzy inference by the above step 406 has been completed for all registrants whose feature value tables are registered. ) Returns to step 402 above, and again performs fuzzy inference using the feature value table for registrants for which fuzzy inference has not been completed. At the time).
[0087]
Since the score by fuzzy reasoning for all the registrants is derived by the repeated processing of the above step 402 to step 408, it is assumed in step 410 that the registrant with the highest score is likely to be the user. In the next step 412, it is determined whether or not the score of the user candidate selected in step 410 is equal to or higher than a predetermined value (0.3 in the present embodiment), and is not higher than the predetermined value. In this case (in the case of negative determination), it is assumed that the user candidate is incorrect and cannot be authenticated, the process proceeds to step 414, and after executing the use range restriction process, the personal identification process is terminated. Note that, in the use range restriction process in the present embodiment, as described above, a process for restricting functions, resources, and devices that can be used by the user is performed.
[0088]
On the other hand, if it is determined in step 412 that the score of the user candidate is greater than or equal to a predetermined value (when an affirmative determination is made), the personal identification process is terminated without performing the process of step 414.
[0089]
In the personal identification process, the average value AV and the standard deviation value HV of the keystroke time difference are used for the following reason.
[0090]
FIG. 13 shows all the results obtained by experiments with five subjects A to E except for the case where the keying time difference is 1000 ms or more and the case where the keying time difference is 100 ms or less and the same key is continuously pressed. FIG. 14 shows a histogram of the keystroke time difference obtained when the [A] key and the [I] key are successively keyed, which is obtained by an experiment by five subjects A to E. , Each shown.
[0091]
As shown in FIGS. 13 and 14, the frequency distribution of the keying time difference in each condition is close to a normal distribution. Therefore, by applying the average value and the standard deviation value of the keying time difference as the feature values, it can be inferred that accurate personal identification can be performed.
[0092]
[Example of experimental results]
FIG. 15 shows an experimental result example of the time-series change of the score obtained by fuzzy inference in the personal identification process (see FIG. 11) according to the present embodiment when the subject A is a user. The horizontal axis of the figure is the effective data appearance frequency (appearance frequency of the keying time difference except when the keying time difference is 1000 ms or more and when the keying time difference is 100 ms or less and the same key is continuously pressed). Yes, the vertical axis is the score. Further, here, a moving average of the keystroke time difference of 5000 times is applied retroactively from the time of execution of fuzzy inference as the feature value.
[0093]
As shown in the figure, it can be seen that the score for the subject A is the highest at any time point, and 100% can be identified.
[0094]
Similarly, FIG. 16 illustrates an experimental result example of a time-series change in scores obtained by fuzzy inference according to the present embodiment when five subjects A to E are users. In the same figure, as in FIG. 15, the horizontal axis represents the effective data appearance frequency, and the vertical axis represents the score. In this case as well, a moving average of the keystroke time difference of 5000 times is applied retroactively from the execution time of fuzzy inference as the feature value.
[0095]
As shown in the figure, it can be seen that all of the five subjects A to E can be identified 100% at any point in time.
[0096]
In FIG. 17, when five subjects A to E are users, the number of times retroactive from the execution time of fuzzy inference when calculating the average value of the keying time difference applied as the feature value is 5 times, 50 times, 500 times, respectively. The example of the experimental result of the discrimination rate of each subject when the number of times and the number of times is 2000 is shown. Here, the identification rate corresponding to the identification of 80000 cases is shown.
[0097]
As shown in FIG. 17A, when the moving average is 5 times, the identification rate corresponding to the subject is higher than the identification rate of other subjects, but the identification rates of all subjects except for the subject D are It is less than 50%, and it can be seen that the identification accuracy is not necessarily good. However, when the moving average shown in FIG. 17 (B) is 50 times, the discrimination rate for all subjects exceeds 70%, and the moving average shown in FIG. 17 (C) is 500 times. In this case, the discrimination rate for all subjects exceeds 95%, and when the moving average shown in FIG. 17D is 2000 times, the discrimination rate for all subjects is 100%. I understand that.
[0098]
Therefore, in this case, if the number of moving averages is 50 or more, it can be considered that the identification can be performed with high accuracy.
[0099]
On the other hand, FIG. 18 shows the appearance of valid data when subject A is identified by using the feature values obtained by the operations of subjects A to E on the keyboard of the desktop personal computer for other notebook personal computers. FIG. 19 shows an example of the score for the number of times when the subject B is identified by using the characteristic values obtained by the operations of the subjects A to E on the keyboard of the desktop personal computer for other desktop personal computers. An example of a score for the number of valid data appearances is shown in FIG. 20, in which a characteristic value obtained by the operation of subjects A to E on the keyboard of a desktop personal computer is used for another desktop personal computer, Valid data appearance times when identified Score example for are shown respectively. In FIG. 18 to FIG. 20, a moving average of the keystroke time difference for 5000 times is applied retrospectively from the time of execution of fuzzy inference as the feature value.
[0100]
As shown in FIGS. 18-20, it turns out that it is possible to identify each test subject 100% at any time. That is, the feature value acquired by the feature value acquisition process shown in the present embodiment can be applied to identification by a keyboard other than the keyboard used for acquisition.
[0101]
As described above in detail, the personal identification device according to the present embodiment includes a key pressing time when a key is pressed on a key provided on the keyboard and specific information (key pressing code) for specifying the key. The key specified by the difference in keying time between two adjacent keying events and the specific information of the two keying events based on the detected keying event. Deriving a combination, calculating a feature value based on the keystroke time difference for each of the derived combinations of keys, storing the calculated feature value, and using the stored feature value Since personal identification is performed, counterfeiting is difficult, there is no psychological resistance of the user, there are almost no restrictions on the body at the time of use, high flexibility, and prevention of impersonation due to replacement. It can be.
[0102]
Moreover, in the personal identification device according to the present embodiment, since the individual is identified by fuzzy inference using the membership function created based on the feature value, the identification rate can be improved.
[0103]
Moreover, in the personal identification device according to the present embodiment, the user's range of use is limited when the evaluation value (score) of the identification result is equal to or less than a predetermined value, so that security can be improved.
[0104]
Furthermore, in the personal identification device according to the present embodiment, since the keystroke time is detected based on the event of releasing the keyboard key by the hook function, it is possible to reliably detect the keystroke, so that a highly accurate feature value As a result, an individual can be identified with high accuracy.
[0105]
In the present embodiment, the case where the feature value acquisition process and the identification process (see FIG. 4) are sequentially executed as an interrupt process when a key on the keyboard 14 is pressed by the user has been described. For example, the keystroke event detection process 40 and the identification process in the feature value acquisition process are sequentially executed in the same manner as in the present embodiment. However, the two consecutive keystroke event derivation process 42 and the keystroke event feature value calculation process are performed. 44 may be executed when a predetermined number of keystroke events in the keystroke event table are accumulated or every predetermined period (for example, one day). In this case, since the two consecutive keystroke event derivation process 42 and the keystroke event feature value calculation process 44 are executed intermittently, the time at which the process is executed last time until these processes are executed next time. Since personal identification is performed based on the feature value data, the identification rate may be slightly lower than in the present embodiment, but for the two consecutive keystroke event derivation processing 42 and keystroke event feature value calculation processing 44 Processing time can be shortened.
[0106]
Further, in the present embodiment, the keystroke event detecting means and the double-connected keystroke event deriving means in the invention described in claim 1, and the keystroke event detecting means and the double-connected keystroke event deriving means in the invention described in claim 2 are keyed. Although the case where the event detection processing 40 and the double-connected keystroke event derivation processing 42 are used together has been described, the present invention is not limited to this, and the keystroke event detection means and the double-connected keystroke event derivation in the invention according to claim 1 The means and the keystroke event detecting means and the double-connected keystroke event deriving means in the invention described in claim 2 may be executed separately by different processes. Also in this case, the same effects as in the present embodiment can be obtained.
[0107]
In the present embodiment, the case where a user is identified by fuzzy inference has been described. However, the present invention is not limited to this. For example, a membership used for fuzzy inference by applying a fuzzy neural network. It is also possible to adopt a form of autonomous learning of functions and the like.
[0108]
【The invention's effect】
  As described above in detail, according to the personal identification device of claim 1, a key-pressing event including a key-pressing time when an individual key-presses a key provided on the keyboard and specific information for specifying the key is included. Based on the detected keystroke event, the above individual continuouslyArbitrarilyDeriving a key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the above-mentioned specific information of the two keying events, and for each of the derived combinations, the same combination Calculate the feature value based on the above keying time difference, and calculate the calculated feature valueSequentialSince it is memorized, it is difficult to counterfeit by using the memorized feature values, there is no psychological resistance of the user, and there are almost no body restraints when using it In addition, it is possible to obtain an effect that personal identification is possible because of high flexibility and prevention of impersonation due to replacement.
[0109]
  In addition, according to the personal identification device of claim 2, the individual continuously operates the keys provided on the keyboard.ArbitrarilyThe keying time when the key is pressedSameFor each combination, the feature value based on the keystroke time difference between the two keys isMemorized sequentiallyA storage means is provided, and a keying event including a keying time when the individual keys a key provided on the keyboard and specific information for specifying the key is detected, and based on the detected keying event, Individuals continuouslyArbitrarilyA key combination specified by the keying time difference between two adjacent keying events when the key is pressed and the specific information of the two keying events is derived, and the storage means corresponding to the derived key combination is derived.SequentialSince the stored feature value is compared with the derived keying time difference and the individual is identified based on the comparison result, forgery is difficult, there is no psychological resistance of the user, There are almost no restraint conditions of the body at the time, high flexibility, and further, it is possible to prevent impersonation due to replacement.
[Brief description of the drawings]
FIG. 1 is a perspective view showing an external appearance of a personal identification device 10 according to an embodiment.
FIG. 2 is a block diagram showing a schematic configuration of a personal identification device 10 according to the embodiment.
FIG. 3 is a schematic diagram showing various storage areas of a hard disk built in the personal identification device 10 according to the embodiment.
4 is a block diagram showing the positioning of various processes executed by the personal identification device 10 according to the embodiment. FIG.
FIG. 5 is a flowchart showing a flow of a keystroke event detection processing program according to the embodiment.
FIG. 6 is a schematic diagram showing an example of a keystroke event table.
FIG. 7 is a flowchart showing a flow of a double-connected keystroke event derivation processing program according to the embodiment.
FIG. 8 is a schematic diagram showing an example of a double-connected keystroke event table.
FIG. 9 is a flowchart showing a flow of a keystroke event feature value calculation processing program according to the embodiment.
FIG. 10 is a schematic diagram showing an example of a feature value table, where (A) is a feature value table in an initial state, and (B) is a schematic diagram showing a feature value table after various data are stored. is there.
FIG. 11 is a flowchart showing a flow of a personal identification processing program according to the embodiment.
FIG. 12 is a diagram showing an example of a membership function in the present embodiment.
FIG. 13 shows all the results obtained by experiments with five subjects A to E except for a case where the keying time difference is 1000 ms or more and a case where the keying time difference is 100 ms or less and the same key is continuously pressed. It is a diagram which shows the histogram of an effective keystroke time difference.
FIG. 14 is a diagram showing a histogram of the effective keying time difference obtained when the [A] key and the [I] key are continuously pressed, obtained by an experiment by five subjects A to E.
FIG. 15 is a diagram showing an example of an experimental result of a time-series change of a score obtained by fuzzy inference in the personal identification process according to the present embodiment when a subject A is a user.
FIG. 16 is a diagram showing an example of an experimental result of a time-series change in scores obtained by fuzzy inference according to the present embodiment when five subjects A to E are users.
FIG. 17 shows the case where five subjects A to E are users, and the number of times retroactive from the execution time of fuzzy inference when calculating the average value of keystroke time differences to be applied as feature values is 5 times, 50 times, and 500 times, respectively. It is a bar graph which shows the experimental result example of the identification rate of each test subject when it is set to 2000 times.
FIG. 18 is a score for the number of effective data appearances when subject A is identified by using the feature values obtained by the operations of subjects A to E on the keyboard of a desktop personal computer for another notebook personal computer. It is a diagram which shows an example.
FIG. 19 is an example of a score for the number of effective data appearances when subject B is identified using feature values obtained by the operations of subjects A to E on the keyboard of a desktop personal computer for another desktop personal computer. FIG.
FIG. 20 is an example of a score for the number of effective data appearances when the subject C is identified by using the feature values obtained by the operations of the subjects A to E on the keyboard of the desktop personal computer for other desktop personal computers. FIG.
[Explanation of symbols]
10 Personal identification device
14 Keyboard
20 Hard disk (storage means)
22 CPU
40 Keystroke event detection processing (keystroke event detection means)
42 Double-connected keystroke event derivation process (Double-connected keystroke event derivation means)
44 Keystroke event feature value calculation processing (keystroke event feature value calculation means)
50 Personal identification procedure (identification means)
52 Use range restriction processing (restriction means)

Claims (5)

A personal identification device for identifying the individual based on an action of the individual operating the keyboard,
A key-pressing event detecting means for detecting a key-pressing event including a key-pressing time when the individual keys a key provided on the keyboard and specific information for specifying the key;
Based on the keying events detected by the key-entry event detection means, specified by the specifying information of key depression time difference and the two keying events of two adjacent keying events when the keying optionally the individual continuous A means for deriving two consecutive keystroke events for deriving a combination of keys;
Of the key combinations derived by the two consecutive keystroke event deriving means, a keystroke event feature value calculating means for calculating a feature value based on the keystroke time difference for each same combination;
Storage means for sequentially storing the feature values calculated by the keystroke event feature value calculating means;
A personal identification device. A personal identification device for identifying the individual based on an action of the individual operating the keyboard,
For each same combination of two adjacent keys whose keying time when the individual continuously and arbitrarily presses keys provided on the keyboard, feature values based on the keying time difference between the two keys are sequentially stored. a storage means which is,
A key-pressing event detecting means for detecting a key-pressing event including a key-pressing time when the individual keys a key provided on the keyboard and specific information for specifying the key;
Based on the keying events detected by the key-entry event detection means, specified by the specifying information of key depression time difference and the two keying events of two adjacent keying events when the keying optionally the individual continuous A means for deriving two consecutive keystroke events for deriving a combination of keys;
Comparing the feature value sequentially stored in the storage means corresponding to the combination of keys derived by the double connected key event deriving means and the keying time difference derived by the double connected key event deriving means; Identification means for identifying an individual based on the comparison result;
A personal identification device. The personal identification device according to claim 2, wherein the identification unit identifies an individual by fuzzy inference using a membership function created based on the feature value. The personal identification device according to claim 2, further comprising a restriction unit that restricts the range of use of the individual when an evaluation value of the identification result by the identification unit is a predetermined value or less. 5. The personal identification device according to claim 1, wherein the key-pressing event detecting unit detects the key-pressing time by using an event of releasing a key of the keyboard.

Images