2018 brought massive, hardware-level security vulnerabilities to the forefront. Here’s the five biggest vulnerabilities of the year, and how you can address them.
2018 was a year full of headaches for IT professionals, as security vulnerabilities became larger and more difficult to patch, since software mitigations for hardware vulnerabilities require some level of compromise. Here’s the five biggest security vulnerabilities of 2018, and what-if anything-you can do to address them in your organization.
1. Spectre and Meltdown dominated security decisions all year
On January 4, the Spectre and Meltdown vulnerabilities allowing applications to read kernel memory were disclosed, and posed security problems for IT professionals all year, as the duo represented largely hardware-level flaws, which can be mitigated-but not outright patched-through software. Though Intel processors (except for Atom processors before 2013, and the Itanium series) are the most vulnerable, microcode patches were necessary for AMD, OpenPOWER, and CPUs based on Arm designs. Other software mitigations do exist, though some require vendors to recompile their programs with protections in place.
The disclosure of these vulnerabilities sparked a renewed interest in side-channel attacks requiring manipulative or speculative execution. Months later, the BranchScope vulnerability was disclosed, focusing on the shared branch target predictor. The researchers behind that disclosure indicated that BranchScope provides the ability to read data which should be protected by the SGX secure enclave, as well as defeat ASLR.
Between the initial disclosure, Spectre-NG, Spectre 1.2, and SpectreRSB, a total of eight variants were discovered, in addition to related work like SgxPectre.
2. Record-breaking DDoS attacks with memcached
Malicious actors staged amplification attacks using flaws in memcached, reaching heights of 1.7 Tbps. The attack is initiated by a server spoofing their own IP address-specifying the attack target address as the origin address-and sending a 15-byte request packet, which is answered by a vulnerable memcached server with responses ranging from 134KB to 750KB. The size disparity between the request and response-as much as 51,200 times larger-made this attack particularly potent.
Proof-of-concept code, which can be easily adapted for attacks was published by various researchers, among them “Memcrashed.py,” which integrates with the Shodan search engine to find vulnerable servers from which you can launch an attack.
Fortunately, it is possible to stop memcached DDoS attacks, though users of memcached should change the defaults to prevent their systems from being abused. If UDP is not used in your deployment, you can disable the feature with the switch -U 0. Otherwise, limiting access to localhost with the switch -listen 127.0.0.1 is advisable.
3. Drupal CMS vulnerability allows attackers to commandeer your site
A failure to sanitize inputs resulted in the announcement of emergency patches for 1.1 million Drupal-powered websites in late March. The vulnerability relates to a conflict between how PHP handles arrays in URL parameters, and Drupal’s use of a hash (#) at the beginning of array keys to signify special keys that typically result in further computation, allowing attackers to inject code arbitrarily. The attack was nicknamed “Drupalgeddon 2: Electric Hashaloo” by Paragon Initative’s Scott Arciszewski.
In April, the same core issue was patched for a second time, relating to the URL handling of GET parameters not being sanitized to remove the # symbol, creating a remote code execution vulnerability.
Despite the highly publicized nature of the vulnerability, over 115,000 Drupal websites were still vulnerable to the issue, and various botnets were actively leveraging the vulnerability to deploy cryptojacking malware.
ZDNet’s Catalin Cimpanu broke a story in November detailing a new type of attack which leverages Drupalgeddon 2 and Dirty COW to install cryptojacking malware, which can proliferate due to the number of unpatched Drupal installations in the wild.
4. BGP attacks intercept DNS servers for address hijacking
Border Gateway Protocol (BGP), the glue that is used to determine the most efficient path between two systems on the internet, is primed to become the target of malicious actors going forward, as the protocol was designed in large part before considerations of malicious network activity were considered. There is no central authority for BGP routes, and routes are accepted at the ISP level, placing it outside the reach of typical enterprise deployments and far outside the reach of consumers.
In April, a BGP attack was waged against Amazon Route 53, the DNS service component of AWS. According to Oracle’s Internet Intelligence group, the attack originated from hardware located in a facility operated by eNet (AS10297) of Columbus, Ohio. The attackers redirected requests to MyEtherWallet.com to a server in Russia, which used a phishing site clone to harvest account information by reading existing cookies. The hackers gained 215 Ether from the attack, which equates to approximately $160,000 USD.
BGP has also been abused by governments in certain circumstances. In November 2018, reports indicated that the Iranian government used BGP attacks in an attempt to intercept Telegram traffic, and China has allegedly used BGP attacks through points of presence in North America, Europe, and Asia.
Work on securing BGP against these attacks is ongoing with NIST and DHS Science and Technology Directorate collaborating on Secure Inter-Domain Routing (SIDR), which aims to implement “BGP Route Origin Validation, using Resource Public Key Infrastructure, [which] can address and resolve the erroneous exchange of network routes.”
5. Australia’s Assistance and Access Bill undermines security
In Australia, the “Assistance and Access Bill 2018,” which provides the government “frameworks for voluntary and mandatory industry assistance to law enforcement and intelligence agencies,” basically provides government access to the contents of encrypted communication. It is essentially the definition of a self-inflicted wound, as the powers it provides the government stand to undermine confidence in Australian products, as well as the Australian outposts of technology companies.
The bill hastily passed on December 7 and was touted as necessary in the interest of “safeguarding national security,” though subtracting perpetrators. Australia has seen a grand total of seven deaths related to terrorist activities since 2000. Additionally, the bill permits demands to be issued in relation to “the interests of Australia’s foreign relations or the interests of Australia’s national economic well-being.”
While the bill appears to not provide the full firehose of user data unencrypted to government agencies, it does permit the government to compel companies to provide content from specific communication, though forbids the disclosure of any demands made to companies about compliance. Stilgherrian provides a balanced view of the final bill in his guide on ZDNet.