TED2014
Del Harvey: Protecting Twitter users (sometimes from themselves)
戴爾．哈維: 推特對特殊訊息的衡量
Filmed:
Readability: 4.4
993,199 views
在推持，每秒有成千上萬則訊息被傳送，百萬分一的機會看似微小無害，但對使用者造成傷害的訊息，算下來每天大約會發生 500 萬次。在推特的信任及安全團隊擔任主管的戴爾．哈維認為這樣的機率不容忽視。這位安全專家花大量的時間思考，在給人們發言機會的同時，也說明如何避免最壞的情況發生，以不動聲色的幽默，讓大家了解她是如何保護 2.4 億推特用戶的安全。
Del Harvey - Security maven
Del Harvey works to define policy and to ensure user safety and security in the challenging realm of modern social media. Full bio
Del Harvey works to define policy and to ensure user safety and security in the challenging realm of modern social media. Full bio
Double-click the English transcript below to play the video.
00:12
My job at Twitter
0
984
1291
我在推特的工作
00:14
is to ensure user trust,
1
2275
1978
就是確保使用者對推特的信任，
00:16
protect user rights and keep users safe,
2
4253
2837
以及保護使用者的權益及安全，
00:19
both from each other
3
7090
1260
不只是使用者之間
00:20
and, at times, from themselves.
4
8350
3899
有時是有關使用者本身的權益及安全。
00:24
Let's talk about what scale looks like at Twitter.
5
12249
4275
讓我們談談推特訊息的規模。
00:28
Back in January 2009,
6
16524
2870
2009 年 1 月，
00:31
we saw more than two million new tweets each day
7
19394
3331
每一天，我們在推特平台上
00:34
on the platform.
8
22725
1764
看到超過 200 萬條新訊息。
00:36
January 2014, more than 500 million.
9
24489
5908
2014 年 1 月，則有超過 5 億條訊息。
00:42
We were seeing two million tweets
10
30397
2492
六分鐘內，
00:44
in less than six minutes.
11
32889
2176
就有 200 萬條推文。
00:47
That's a 24,900-percent increase.
12
35065
6984
那是 24,900% 的成長。
00:54
Now, the vast majority of activity on Twitter
13
42049
3253
今天，絶大部分在推特上的活動
00:57
puts no one in harm's way.
14
45302
1503
不會傷害任何人。
00:58
There's no risk involved.
15
46805
1935
沒有任何風險。
01:00
My job is to root out and prevent activity that might.
16
48740
5753
我的工作則是根除
任何可能傷害他人權益的活動。
任何可能傷害他人權益的活動。
01:06
Sounds straightforward, right?
17
54493
1973
聽起來很簡單，對吧？
01:08
You might even think it'd be easy,
18
56466
1152
你或許會認為這個工作很簡單，
01:09
given that I just said the vast majority
19
57618
2170
尤其當我說，推特上絕大部分的動作
01:11
of activity on Twitter puts no one in harm's way.
20
59788
3810
並不會對任何人造成傷害。
01:15
Why spend so much time
21
63598
2169
那為什麼要花費這麼多時間
01:17
searching for potential calamities
22
65767
2743
在無害的網路活動中，
01:20
in innocuous activities?
23
68510
2900
尋找可能的危機？
01:23
Given the scale that Twitter is at,
24
71410
2940
以推特的規模來看，
01:26
a one-in-a-million chance happens
25
74350
2357
百萬分之一的機率，
01:28
500 times a day.
26
76707
4876
相當於一天會有 500 條
可能造成危害的訊息。
可能造成危害的訊息。
01:33
It's the same for other companies
27
81583
1445
這個訊息量，是其他公司
01:35
dealing at this sort of scale.
28
83028
1471
所要處理的訊息量相同
01:36
For us, edge cases,
29
84499
1708
對我們而言，那些稀少罕見，
01:38
those rare situations that are unlikely to occur,
30
86207
3625
不太可能發生的極端事件，
01:41
are more like norms.
31
89832
2622
有如家常便飯。
01:44
Say 99.999 percent of tweets
32
92454
3942
假設百分之 99.999% 的推文
01:48
pose no risk to anyone.
33
96396
1888
都不會傷害任何人，
01:50
There's no threat involved.
34
98284
1066
不涉及任何風險。
01:51
Maybe people are documenting travel landmarks
35
99350
2954
也許大家只是在記錄旅遊景點，
01:54
like Australia's Heart Reef,
36
102304
1963
像是澳洲的心形礁，
01:56
or tweeting about a concert they're attending,
37
104267
2921
或是傳些關於他們正在參加的演唱會，
01:59
or sharing pictures of cute baby animals.
38
107188
4747
或者是分享一些可愛小動物的照片。
02:03
After you take out that 99.999 percent,
39
111935
4509
當你除去那 99.999% 的機率，
02:08
that tiny percentage of tweets remaining
40
116444
3529
剩下極微小的百分比
02:11
works out to roughly
41
119973
2389
粗估下來
02:14
150,000 per month.
42
122362
3475
每月大約有十五萬條訊息。
02:17
The sheer scale of what we're dealing with
43
125837
2456
管理這麼龐大的規模，
02:20
makes for a challenge.
44
128293
2312
是個挑戰。
02:22
You know what else makes my role
45
130605
1178
你知道還有什麼
02:23
particularly challenging?
46
131783
3107
讓我的工作更具挑戰性的嗎？
02:26
People do weird things.
47
134890
5123
人會做些奇怪的事。
02:32
(Laughter)
48
140013
1829
(笑聲)
02:33
And I have to figure out what they're doing,
49
141842
2391
而我則必須搞清楚他們在做什麼，
02:36
why, and whether or not there's risk involved,
50
144233
2249
動機是什麼，還有是否有危險性，
02:38
often without much in terms of context
51
146482
2168
且通常是在沒有資料
02:40
or background.
52
148650
1847
或背景的情況下就要去搞清楚。
02:42
I'm going to show you some examples
53
150497
2077
讓我舉幾個我在推特工作時
02:44
that I've run into during my time at Twitter --
54
152574
2005
遇到的例子，
02:46
these are all real examples —
55
154579
1620
這些全都是真實的案例，
02:48
of situations that at first seemed cut and dried,
56
156199
2653
一些原先看來簡單明瞭的情況，
02:50
but the truth of the matter was something
57
158852
1643
但事情的真相
02:52
altogether different.
58
160495
1550
又是截然不同。
02:54
The details have been changed
59
162045
1977
有些細節已被更改，
02:56
to protect the innocent
60
164022
1257
是為了保護無辜的人，
02:57
and sometimes the guilty.
61
165279
3233
有時也包括罪犯。
03:00
We'll start off easy.
62
168512
3005
我們從簡單的開始。
03:03
["Yo bitch"]
63
171517
1793
[嘿 賤女人]
03:05
If you saw a Tweet that only said this,
64
173310
3228
當你在推特上看到這句話，
03:08
you might think to yourself,
65
176538
1694
你可能會認為：
03:10
"That looks like abuse."
66
178232
1653
「那是一種辱罵」。
03:11
After all, why would you
want to receive the message,
want to receive the message,
67
179885
3107
畢竟，誰會希望收到這樣的訊息：
03:14
"Yo, bitch."
68
182992
2218
「嘿，賤女人。」
03:17
Now, I try to stay relatively hip
69
185210
4663
現在，我試著跟上趨勢
03:21
to the latest trends and memes,
70
189873
2512
及最新流行用語，
03:24
so I knew that "yo, bitch"
71
192385
2704
所以我知道「嘿，賤女人」
03:27
was also often a common greeting between friends,
72
195089
3154
也常被用作朋友間的招呼用語
03:30
as well as being a popular "Breaking Bad" reference.
73
198243
4262
是來自於《絕命毒師》的說法。
03:34
I will admit that I did not expect
74
202505
2487
我得承認我沒有想到
03:36
to encounter a fourth use case.
75
204992
2841
這句話會有第四種用法。
03:39
It turns out it is also used on Twitter
76
207833
3104
原來在推特上，扮成狗的人
03:42
when people are role-playing as dogs.
77
210937
3062
也會用這個詞。
03:45
(Laughter)
78
213999
5279
（笑聲）
03:51
And in fact, in that case,
79
219278
1666
事實上，在這個情況下，
03:52
it's not only not abusive,
80
220944
1609
不止沒有辱罵的意味，
03:54
it's technically just an accurate greeting.
81
222553
3139
嚴格說來，這是一個準確的問候用語。
03:57
(Laughter)
82
225692
2889
（笑聲）
04:00
So okay, determining whether or not
83
228581
2071
所以，一條沒有來龍去脈的訊息
04:02
something is abusive without context,
84
230652
1848
要去判定這個訊息是否有辱罵的意味，
04:04
definitely hard.
85
232500
1592
絕對是很困難的。
04:06
Let's look at spam.
86
234092
2717
我們來看看垃圾訊息。
04:08
Here's an example of an account engaged
87
236809
1960
這是使用者傳送垃圾訊息
04:10
in classic spammer behavior,
88
238769
1668
的典型例子，
04:12
sending the exact same message
89
240437
1559
一直不斷地傳送相同的訊息
04:13
to thousands of people.
90
241996
1804
給上千個人。
04:15
While this is a mockup I put
together using my account,
together using my account,
91
243800
2793
這是我用自己帳號作出的模擬範例，
04:18
we see accounts doing this all the time.
92
246593
3001
我們總可以看到使用者傳送這樣的訊息。
04:21
Seems pretty straightforward.
93
249594
1979
看起來相當簡單明瞭。
04:23
We should just automatically suspend accounts
94
251573
2053
我們應該自動封鎖
04:25
engaging in this kind of behavior.
95
253626
3307
涉及這種行為的帳號。
04:28
Turns out there's some exceptions to that rule.
96
256933
3210
結果總有些例外。
04:32
Turns out that that message
could also be a notification
could also be a notification
97
260143
2883
這些訊息，也有可能是通知
04:35
you signed up for that the International
Space Station is passing overhead
Space Station is passing overhead
98
263026
3889
你登記參加國際太空站經過你上空的活動，
04:38
because you wanted to go outside
99
266915
1846
你希望收到通知，即時走到戶外
04:40
and see if you could see it.
100
268761
1948
可以親自目睹。
04:42
You're not going to get that chance
101
270709
1225
你絶不會因為
04:43
if we mistakenly suspend the account
102
271934
1847
誤認為這是垃圾訊息
04:45
thinking it's spam.
103
273781
2266
而停用這個帳號的情況發生。
04:48
Okay. Let's make the stakes higher.
104
276047
3526
好。讓我們再把風險的層級提高。
04:51
Back to my account,
105
279573
1916
再來看我的帳號，
04:53
again exhibiting classic behavior.
106
281489
3505
在推特上展示特定的行為。
04:56
This time it's sending the same message and link.
107
284994
2643
這次是在持特上傳送相同的訊息和連結
04:59
This is often indicative of
something called phishing,
something called phishing,
108
287637
2774
這通常是一種網路釣魚，
05:02
somebody trying to steal another
person's account information
person's account information
109
290411
3178
有人試著去引導他人到另一個網站
05:05
by directing them to another website.
110
293589
2203
然後盜用他的帳號
05:07
That's pretty clearly not a good thing.
111
295792
4194
很明顯這不是一件好事。
05:11
We want to, and do, suspend accounts
112
299986
1930
我們要，而且必須去阻止
05:13
engaging in that kind of behavior.
113
301916
2624
可疑的帳號去做這樣的行為。
05:16
So why are the stakes higher for this?
114
304540
3247
但是，為何這麼做風險更高？
05:19
Well, this could also be a bystander at a rally
115
307787
2999
這像是遊行人潮當中的旁觀者
05:22
who managed to record a video
116
310786
1910
拿著攝影機，對著
05:24
of a police officer beating a non-violent protester
117
312696
3270
警察動手打一個
無暴力行為的抗議者攝影，
無暴力行為的抗議者攝影，
05:27
who's trying to let the world know what's happening.
118
315966
2975
好讓全世界的人知道此事。
05:30
We don't want to gamble
119
318941
1643
我們不想冒這個險
05:32
on potentially silencing that crucial speech
120
320584
2517
把有可能很重要的訊息
05:35
by classifying it as spam and suspending it.
121
323101
2929
歸類為垃圾訊息，然後停用帳號。
05:38
That means we evaluate hundreds of parameters
122
326030
2879
那意味著，當我們在觀察使用者行為時
05:40
when looking at account behaviors,
123
328909
1688
我們憑估成千上百個因素，
05:42
and even then, we can still get it wrong
124
330597
2016
即使是這麼做了，百密仍有一疏，
05:44
and have to reevaluate.
125
332613
2236
必須再重新評估這些訊息。
05:46
Now, given the sorts of challenges I'm up against,
126
334849
3708
現在，面臨各式各樣的挑戰，
05:50
it's crucial that I not only predict
127
338557
2696
重要的是，不但要去預測可能發生的事，
05:53
but also design protections for the unexpected.
128
341253
3784
而且要對可能發生的事，
設計一套因應的保護措施。
設計一套因應的保護措施。
05:57
And that's not just an issue for me,
129
345037
2342
這不僅事關我和推特，
05:59
or for Twitter, it's an issue for you.
130
347379
2087
這也關係到你。
06:01
It's an issue for anybody who's building or creating
131
349466
2406
關係到任何想創造美好事物，
06:03
something that you think is going to be amazing
132
351872
1925
06:05
and will let people do awesome things.
133
353797
2789
以及想要讓他人也一起
做美好事物的推特使用者。
做美好事物的推特使用者。
06:08
So what do I do?
134
356586
2866
所以我要怎麼做呢？
06:11
I pause and I think,
135
359452
3318
我一再思考這問題
06:14
how could all of this
136
362770
2095
這些事情
06:16
go horribly wrong?
137
364865
3793
到底怎麼會出錯？
06:20
I visualize catastrophe.
138
368658
4453
我想像發生災難的情形。
06:25
And that's hard. There's a sort of
139
373111
2463
這很困難，
06:27
inherent cognitive dissonance in doing that,
140
375574
2848
因為這麼做，
有點像是內在認知不協調，
有點像是內在認知不協調，
06:30
like when you're writing your wedding vows
141
378422
1812
就像是寫結婚誓言時，
06:32
at the same time as your prenuptial agreement.
142
380234
2646
同時也寫婚前協議書。
06:34
(Laughter)
143
382880
1696
（笑聲）
06:36
But you still have to do it,
144
384576
2373
但還是必須要去做，
06:38
particularly if you're marrying
500 million tweets per day.
500 million tweets per day.
145
386949
4446
特別是每天要處理 5 億條推文。
06:43
What do I mean by "visualize catastrophe?"
146
391395
3097
我所說的「想像災難」是什麼意思呢？
06:46
I try to think of how something as
147
394492
2762
我試著去想像，
06:49
benign and innocuous as a picture of a cat
148
397254
3228
像是一張無害的貓咪照片
06:52
could lead to death,
149
400482
1104
為何可能導致死亡，
06:53
and what to do to prevent that.
150
401586
2326
以及如何避免這種事情發生。
06:55
Which happens to be my next example.
151
403912
2383
正是接下來我要說的例子。
06:58
This is my cat, Eli.
152
406295
3110
這隻是我的貓，叫伊萊。
07:01
We wanted to give users the ability
153
409405
1981
我們盡可能讓推特的使用者
07:03
to add photos to their tweets.
154
411386
2073
在推特上傳送圖片，
07:05
A picture is worth a thousand words.
155
413459
1597
一張圖勝過千言萬語，
07:07
You only get 140 characters.
156
415056
2009
而一次推文只能傳送 140 個字。
07:09
You add a photo to your tweet,
157
417065
1200
你在推文加入圖片，
07:10
look at how much more content you've got now.
158
418265
3038
你會發現推文的內容更加豐富。
07:13
There's all sorts of great things you can do
159
421303
1677
藉由推特加入圖片的功能，
07:14
by adding a photo to a tweet.
160
422980
2007
你可以做各種麼美妙的事。
07:16
My job isn't to think of those.
161
424987
2280
我的工作不是去想這些事情。
07:19
It's to think of what could go wrong.
162
427267
2747
而是去想事情可能會出什麼差錯。
07:22
How could this picture
163
430014
1892
這張圖片
07:23
lead to my death?
164
431906
3539
如何導致我死亡？
07:27
Well, here's one possibility.
165
435445
3160
有一個可能性。
07:30
There's more in that picture than just a cat.
166
438605
3086
這張圖的資訊不只是一隻貓。
07:33
There's geodata.
167
441691
2092
還有地理資訊在裡頭。
07:35
When you take a picture with your smartphone
168
443783
2212
當你以智慧型手機
07:37
or digital camera,
169
445995
1299
或數位相機拍照，
07:39
there's a lot of additional information
170
447294
1654
會有許多額外的資訊
07:40
saved along in that image.
171
448948
1616
儲存在照片裡。
07:42
In fact, this image also contains
172
450564
1932
事實上，這張照片還包含
07:44
the equivalent of this,
173
452496
1805
相當於這個的資訊，
07:46
more specifically, this.
174
454301
3079
更具體地說是這個。
07:49
Sure, it's not likely that someone's going to try
175
457380
1956
當然，不太可能有人嘗試
07:51
to track me down and do me harm
176
459336
2285
根據這張貓照片的相關資訊
07:53
based upon image data associated
177
461621
1784
追蹤我以及傷害我。
07:55
with a picture I took of my cat,
178
463405
1948
07:57
but I start by assuming the worst will happen.
179
465353
3651
但我一開始就要假設
最壞的情況一定會發生，
最壞的情況一定會發生，
08:01
That's why, when we launched photos on Twitter,
180
469004
2338
這就是為什麼我們
在開放上傳照片到推特時，
在開放上傳照片到推特時，
08:03
we made the decision to strip that geodata out.
181
471342
3821
就決定把照片裡的地理資訊全刪掉。
08:07
(Applause)
182
475163
5847
（掌聲）
08:13
If I start by assuming the worst
183
481010
2613
如果一開始，我就假設
可能發生最壞的情況，
可能發生最壞的情況，
08:15
and work backwards,
184
483623
947
然後再往前倒推，
08:16
I can make sure that the protections we build
185
484570
2553
我可以確定我們建立的保護制度，
08:19
work for both expected
186
487123
1768
可以應付意料中
08:20
and unexpected use cases.
187
488891
2078
以及意料外的事件。
08:22
Given that I spend my days and nights
188
490969
2945
我日夜地
08:25
imagining the worst that could happen,
189
493914
2541
想像發生最壞情況的情形，
08:28
it wouldn't be surprising if
my worldview was gloomy.
my worldview was gloomy.
190
496455
4257
如果因此造成我憂鬱的世界觀，
也不會令人感到意外。
也不會令人感到意外。
08:32
(Laughter)
191
500712
1783
（笶聲）
08:34
It's not.
192
502495
1417
其實並非如此。
08:35
The vast majority of interactions I see --
193
503912
3876
我看到的絶大部份互動，
08:39
and I see a lot, believe me -- are positive,
194
507788
3901
我看了很多，相信我，
它們都是正面的。
它們都是正面的。
08:43
people reaching out to help
195
511689
1924
人們伸出援手互相幫忙，
08:45
or to connect or share information with each other.
196
513613
3448
彼此互相連絡或分享資訊。
08:49
It's just that for those of us dealing with scale,
197
517061
3323
只是我們要處理龐大的資訊量，
08:52
for those of us tasked with keeping people safe,
198
520384
3800
承擔保護使用者安全的責任，
08:56
we have to assume the worst will happen,
199
524184
2546
所以必須假設將發生最壞的情況，
08:58
because for us, a one-in-a-million chance
200
526730
4227
對我們來說，百萬分之一的可能性
09:02
is pretty good odds.
201
530957
2749
是相當高的機率。
09:05
Thank you.
202
533706
1864
謝謝。
09:07
(Applause)
203
535570
4000
（掌聲）
ABOUT THE SPEAKER
Del Harvey - Security mavenDel Harvey works to define policy and to ensure user safety and security in the challenging realm of modern social media.
Why you should listen
As Senior Director of Trust and Safety at Twitter, Del Harvey works to define policy and to ensure user safety and security in the challenging realm of modern social media. Prior to joining Twitter, she spent five years as the co-administrator and law enforcement liaison for a 501(c)3 non-profit charity, working with agencies ranging from local police departments to the FBI, U.S. Marshals and the Secret Service.
Del Harvey | Speaker | TED.com